ladybird/Libraries/LibJS/Runtime/AbstractOperations.cpp

1938 lines
90 KiB
C++
Raw Normal View History

/*
* Copyright (c) 2020-2023, Linus Groh <linusg@serenityos.org>
* Copyright (c) 2021-2023, Andreas Kling <andreas@ladybird.org>
*
* SPDX-License-Identifier: BSD-2-Clause
*/
#include <AK/CharacterTypes.h>
#include <AK/FloatingPointStringConversions.h>
#include <AK/Function.h>
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
#include <AK/Optional.h>
#include <AK/Utf16View.h>
#include <LibJS/Bytecode/Interpreter.h>
#include <LibJS/ModuleLoading.h>
#include <LibJS/Parser.h>
#include <LibJS/Runtime/AbstractOperations.h>
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
#include <LibJS/Runtime/Accessor.h>
#include <LibJS/Runtime/ArgumentsObject.h>
#include <LibJS/Runtime/Array.h>
#include <LibJS/Runtime/BoundFunction.h>
#include <LibJS/Runtime/Completion.h>
#include <LibJS/Runtime/DeclarativeEnvironment.h>
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
#include <LibJS/Runtime/ECMAScriptFunctionObject.h>
#include <LibJS/Runtime/ErrorTypes.h>
#include <LibJS/Runtime/FunctionEnvironment.h>
#include <LibJS/Runtime/FunctionObject.h>
#include <LibJS/Runtime/GlobalEnvironment.h>
#include <LibJS/Runtime/GlobalObject.h>
#include <LibJS/Runtime/Object.h>
#include <LibJS/Runtime/ObjectEnvironment.h>
#include <LibJS/Runtime/PromiseCapability.h>
#include <LibJS/Runtime/PromiseConstructor.h>
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
#include <LibJS/Runtime/PropertyDescriptor.h>
#include <LibJS/Runtime/PropertyKey.h>
#include <LibJS/Runtime/ProxyObject.h>
#include <LibJS/Runtime/Reference.h>
#include <LibJS/Runtime/StringPrototype.h>
#include <LibJS/Runtime/SuppressedError.h>
#include <LibJS/Runtime/Temporal/AbstractOperations.h>
#include <LibJS/Runtime/ValueInlines.h>
namespace JS {
// 7.2.1 RequireObjectCoercible ( argument ), https://tc39.es/ecma262/#sec-requireobjectcoercible
ThrowCompletionOr<Value> require_object_coercible(VM& vm, Value value)
{
if (value.is_nullish())
return vm.throw_completion<TypeError>(ErrorType::NotObjectCoercible, value.to_string_without_side_effects());
return value;
}
// 7.3.14 Call ( F, V [ , argumentsList ] ), https://tc39.es/ecma262/#sec-call
ThrowCompletionOr<Value> call_impl(VM& vm, Value function, Value this_value, ReadonlySpan<Value> arguments_list)
LibJS: Implement [[Call]] and [[Construct]] internal slots properly This patch implements: - Spec compliant [[Call]] and [[Construct]] internal slots, as virtual FunctionObject::internal_{call,construct}(). These effectively replace the old virtual FunctionObject::{call,construct}(), but with several advantages: - Clear and consistent naming, following the object internal methods - Use of completions - internal_construct() returns an Object, and not Value! This has been a source of confusion for a long time, since in the spec there's always an Object returned but the Value return type in LibJS meant that this could not be fully trusted and something could screw you over. - Arguments are passed explicitly in form of a MarkedValueList, allowing manipulation (BoundFunction). We still put them on the execution context as a lot of code depends on it (VM::arguments()), but not from the Call() / Construct() AOs anymore, which now allows for bypassing them and invoking [[Call]] / [[Construct]] directly. Nothing but Call() / Construct() themselves do that at the moment, but future additions to ECMA262 or already existing web specs might. - Spec compliant, standalone Call() and Construct() AOs: currently the closest we have is VM::{call,construct}(), but those try to cater to all the different function object subclasses at once, resulting in a horrible mess and calling AOs with functions they should never be called with; most prominently PrepareForOrdinaryCall and OrdinaryCallBindThis, which are only for ECMAScriptFunctionObject. As a result this also contains an implicit optimization: we no longer need to create a new function environment for NativeFunctions - which, worth mentioning, is what started this whole crusade in the first place :^)
2021-10-08 20:37:21 +01:00
{
// 1. If argumentsList is not present, set argumentsList to a new empty List.
// 2. If IsCallable(F) is false, throw a TypeError exception.
if (!function.is_function())
return vm.throw_completion<TypeError>(ErrorType::NotAFunction, function.to_string_without_side_effects());
LibJS: Implement [[Call]] and [[Construct]] internal slots properly This patch implements: - Spec compliant [[Call]] and [[Construct]] internal slots, as virtual FunctionObject::internal_{call,construct}(). These effectively replace the old virtual FunctionObject::{call,construct}(), but with several advantages: - Clear and consistent naming, following the object internal methods - Use of completions - internal_construct() returns an Object, and not Value! This has been a source of confusion for a long time, since in the spec there's always an Object returned but the Value return type in LibJS meant that this could not be fully trusted and something could screw you over. - Arguments are passed explicitly in form of a MarkedValueList, allowing manipulation (BoundFunction). We still put them on the execution context as a lot of code depends on it (VM::arguments()), but not from the Call() / Construct() AOs anymore, which now allows for bypassing them and invoking [[Call]] / [[Construct]] directly. Nothing but Call() / Construct() themselves do that at the moment, but future additions to ECMA262 or already existing web specs might. - Spec compliant, standalone Call() and Construct() AOs: currently the closest we have is VM::{call,construct}(), but those try to cater to all the different function object subclasses at once, resulting in a horrible mess and calling AOs with functions they should never be called with; most prominently PrepareForOrdinaryCall and OrdinaryCallBindThis, which are only for ECMAScriptFunctionObject. As a result this also contains an implicit optimization: we no longer need to create a new function environment for NativeFunctions - which, worth mentioning, is what started this whole crusade in the first place :^)
2021-10-08 20:37:21 +01:00
// 3. Return ? F.[[Call]](V, argumentsList).
return function.as_function().internal_call(this_value, arguments_list);
LibJS: Implement [[Call]] and [[Construct]] internal slots properly This patch implements: - Spec compliant [[Call]] and [[Construct]] internal slots, as virtual FunctionObject::internal_{call,construct}(). These effectively replace the old virtual FunctionObject::{call,construct}(), but with several advantages: - Clear and consistent naming, following the object internal methods - Use of completions - internal_construct() returns an Object, and not Value! This has been a source of confusion for a long time, since in the spec there's always an Object returned but the Value return type in LibJS meant that this could not be fully trusted and something could screw you over. - Arguments are passed explicitly in form of a MarkedValueList, allowing manipulation (BoundFunction). We still put them on the execution context as a lot of code depends on it (VM::arguments()), but not from the Call() / Construct() AOs anymore, which now allows for bypassing them and invoking [[Call]] / [[Construct]] directly. Nothing but Call() / Construct() themselves do that at the moment, but future additions to ECMA262 or already existing web specs might. - Spec compliant, standalone Call() and Construct() AOs: currently the closest we have is VM::{call,construct}(), but those try to cater to all the different function object subclasses at once, resulting in a horrible mess and calling AOs with functions they should never be called with; most prominently PrepareForOrdinaryCall and OrdinaryCallBindThis, which are only for ECMAScriptFunctionObject. As a result this also contains an implicit optimization: we no longer need to create a new function environment for NativeFunctions - which, worth mentioning, is what started this whole crusade in the first place :^)
2021-10-08 20:37:21 +01:00
}
ThrowCompletionOr<Value> call_impl(VM&, FunctionObject& function, Value this_value, ReadonlySpan<Value> arguments_list)
{
// 1. If argumentsList is not present, set argumentsList to a new empty List.
// 2. If IsCallable(F) is false, throw a TypeError exception.
// Note: Called with a FunctionObject ref
// 3. Return ? F.[[Call]](V, argumentsList).
return function.internal_call(this_value, arguments_list);
}
// 7.3.15 Construct ( F [ , argumentsList [ , newTarget ] ] ), https://tc39.es/ecma262/#sec-construct
ThrowCompletionOr<GC::Ref<Object>> construct_impl(VM&, FunctionObject& function, ReadonlySpan<Value> arguments_list, FunctionObject* new_target)
LibJS: Implement [[Call]] and [[Construct]] internal slots properly This patch implements: - Spec compliant [[Call]] and [[Construct]] internal slots, as virtual FunctionObject::internal_{call,construct}(). These effectively replace the old virtual FunctionObject::{call,construct}(), but with several advantages: - Clear and consistent naming, following the object internal methods - Use of completions - internal_construct() returns an Object, and not Value! This has been a source of confusion for a long time, since in the spec there's always an Object returned but the Value return type in LibJS meant that this could not be fully trusted and something could screw you over. - Arguments are passed explicitly in form of a MarkedValueList, allowing manipulation (BoundFunction). We still put them on the execution context as a lot of code depends on it (VM::arguments()), but not from the Call() / Construct() AOs anymore, which now allows for bypassing them and invoking [[Call]] / [[Construct]] directly. Nothing but Call() / Construct() themselves do that at the moment, but future additions to ECMA262 or already existing web specs might. - Spec compliant, standalone Call() and Construct() AOs: currently the closest we have is VM::{call,construct}(), but those try to cater to all the different function object subclasses at once, resulting in a horrible mess and calling AOs with functions they should never be called with; most prominently PrepareForOrdinaryCall and OrdinaryCallBindThis, which are only for ECMAScriptFunctionObject. As a result this also contains an implicit optimization: we no longer need to create a new function environment for NativeFunctions - which, worth mentioning, is what started this whole crusade in the first place :^)
2021-10-08 20:37:21 +01:00
{
// 1. If newTarget is not present, set newTarget to F.
if (!new_target)
new_target = &function;
// 2. If argumentsList is not present, set argumentsList to a new empty List.
// 3. Return ? F.[[Construct]](argumentsList, newTarget).
return function.internal_construct(arguments_list, *new_target);
LibJS: Implement [[Call]] and [[Construct]] internal slots properly This patch implements: - Spec compliant [[Call]] and [[Construct]] internal slots, as virtual FunctionObject::internal_{call,construct}(). These effectively replace the old virtual FunctionObject::{call,construct}(), but with several advantages: - Clear and consistent naming, following the object internal methods - Use of completions - internal_construct() returns an Object, and not Value! This has been a source of confusion for a long time, since in the spec there's always an Object returned but the Value return type in LibJS meant that this could not be fully trusted and something could screw you over. - Arguments are passed explicitly in form of a MarkedValueList, allowing manipulation (BoundFunction). We still put them on the execution context as a lot of code depends on it (VM::arguments()), but not from the Call() / Construct() AOs anymore, which now allows for bypassing them and invoking [[Call]] / [[Construct]] directly. Nothing but Call() / Construct() themselves do that at the moment, but future additions to ECMA262 or already existing web specs might. - Spec compliant, standalone Call() and Construct() AOs: currently the closest we have is VM::{call,construct}(), but those try to cater to all the different function object subclasses at once, resulting in a horrible mess and calling AOs with functions they should never be called with; most prominently PrepareForOrdinaryCall and OrdinaryCallBindThis, which are only for ECMAScriptFunctionObject. As a result this also contains an implicit optimization: we no longer need to create a new function environment for NativeFunctions - which, worth mentioning, is what started this whole crusade in the first place :^)
2021-10-08 20:37:21 +01:00
}
// 7.3.19 LengthOfArrayLike ( obj ), https://tc39.es/ecma262/#sec-lengthofarraylike
ThrowCompletionOr<size_t> length_of_array_like(VM& vm, Object const& object)
{
// OPTIMIZATION: For Array objects with a magical "length" property, it should always reflect the size of indexed property storage.
if (object.has_magical_length_property())
return object.indexed_properties().array_like_size();
// 1. Return (? ToLength(? Get(obj, "length"))).
return TRY(object.get(vm.names.length)).to_length(vm);
}
// 7.3.20 CreateListFromArrayLike ( obj [ , elementTypes ] ), https://tc39.es/ecma262/#sec-createlistfromarraylike
ThrowCompletionOr<GC::RootVector<Value>> create_list_from_array_like(VM& vm, Value value, Function<ThrowCompletionOr<void>(Value)> check_value)
{
// 1. If elementTypes is not present, set elementTypes to « Undefined, Null, Boolean, String, Symbol, Number, BigInt, Object ».
// 2. If Type(obj) is not Object, throw a TypeError exception.
if (!value.is_object())
return vm.throw_completion<TypeError>(ErrorType::NotAnObject, value.to_string_without_side_effects());
auto& array_like = value.as_object();
// 3. Let len be ? LengthOfArrayLike(obj).
auto length = TRY(length_of_array_like(vm, array_like));
// 4. Let list be a new empty List.
auto list = GC::RootVector<Value> { vm.heap() };
list.ensure_capacity(length);
// 5. Let index be 0.
// 6. Repeat, while index < len,
for (size_t i = 0; i < length; ++i) {
// a. Let indexName be ! ToString(𝔽(index)).
auto index_name = PropertyKey { i };
// b. Let next be ? Get(obj, indexName).
auto next = TRY(array_like.get(index_name));
// c. If Type(next) is not an element of elementTypes, throw a TypeError exception.
if (check_value)
TRY(check_value(next));
// d. Append next as the last element of list.
list.unchecked_append(next);
}
// 7. Return list.
return ThrowCompletionOr(move(list));
}
// 7.3.23 SpeciesConstructor ( O, defaultConstructor ), https://tc39.es/ecma262/#sec-speciesconstructor
ThrowCompletionOr<FunctionObject*> species_constructor(VM& vm, Object const& object, FunctionObject& default_constructor)
{
// 1. Let C be ? Get(O, "constructor").
auto constructor = TRY(object.get(vm.names.constructor));
// 2. If C is undefined, return defaultConstructor.
if (constructor.is_undefined())
return &default_constructor;
// 3. If Type(C) is not Object, throw a TypeError exception.
if (!constructor.is_object())
return vm.throw_completion<TypeError>(ErrorType::NotAConstructor, constructor.to_string_without_side_effects());
// 4. Let S be ? Get(C, @@species).
auto species = TRY(constructor.as_object().get(vm.well_known_symbol_species()));
// 5. If S is either undefined or null, return defaultConstructor.
if (species.is_nullish())
return &default_constructor;
// 6. If IsConstructor(S) is true, return S.
if (species.is_constructor())
return &species.as_function();
// 7. Throw a TypeError exception.
return vm.throw_completion<TypeError>(ErrorType::NotAConstructor, species.to_string_without_side_effects());
}
// 7.3.25 GetFunctionRealm ( obj ), https://tc39.es/ecma262/#sec-getfunctionrealm
ThrowCompletionOr<Realm*> get_function_realm(VM& vm, FunctionObject const& function)
{
// 1. If obj has a [[Realm]] internal slot, then
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
if (function.realm()) {
// a. Return obj.[[Realm]].
return function.realm();
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
}
// 2. If obj is a bound function exotic object, then
if (is<BoundFunction>(function)) {
auto& bound_function = static_cast<BoundFunction const&>(function);
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// a. Let target be obj.[[BoundTargetFunction]].
auto& target = bound_function.bound_target_function();
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// b. Return ? GetFunctionRealm(target).
return get_function_realm(vm, target);
}
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 3. If obj is a Proxy exotic object, then
if (is<ProxyObject>(function)) {
auto& proxy = static_cast<ProxyObject const&>(function);
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// a. If obj.[[ProxyHandler]] is null, throw a TypeError exception.
if (proxy.is_revoked())
return vm.throw_completion<TypeError>(ErrorType::ProxyRevoked);
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// b. Let proxyTarget be obj.[[ProxyTarget]].
auto& proxy_target = proxy.target();
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// c. Return ? GetFunctionRealm(proxyTarget).
VERIFY(proxy_target.is_function());
return get_function_realm(vm, static_cast<FunctionObject const&>(proxy_target));
}
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 4. Return the current Realm Record.
return vm.current_realm();
}
// 8.5.2.1 InitializeBoundName ( name, value, environment ), https://tc39.es/ecma262/#sec-initializeboundname
ThrowCompletionOr<void> initialize_bound_name(VM& vm, FlyString const& name, Value value, Environment* environment)
{
// 1. If environment is not undefined, then
if (environment) {
// FIXME: The normal is not included in the explicit resource management spec yet, so there is no spec link for it.
// a. Perform ! environment.InitializeBinding(name, value, normal).
MUST(environment->initialize_binding(vm, name, value, Environment::InitializeBindingHint::Normal));
// b. Return unused.
return {};
}
// 2. Else,
else {
// a. Let lhs be ? ResolveBinding(name).
auto lhs = TRY(vm.resolve_binding(name));
// b. Return ? PutValue(lhs, value).
return TRY(lhs.put_value(vm, value));
}
VERIFY_NOT_REACHED();
}
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 10.1.6.2 IsCompatiblePropertyDescriptor ( Extensible, Desc, Current ), https://tc39.es/ecma262/#sec-iscompatiblepropertydescriptor
bool is_compatible_property_descriptor(bool extensible, PropertyDescriptor const& descriptor, Optional<PropertyDescriptor> const& current)
{
// 1. Return ValidateAndApplyPropertyDescriptor(undefined, "", Extensible, Desc, Current).
return validate_and_apply_property_descriptor(nullptr, FlyString {}, extensible, descriptor, current);
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
}
// 10.1.6.3 ValidateAndApplyPropertyDescriptor ( O, P, extensible, Desc, current ), https://tc39.es/ecma262/#sec-validateandapplypropertydescriptor
bool validate_and_apply_property_descriptor(Object* object, PropertyKey const& property_key, bool extensible, PropertyDescriptor const& descriptor, Optional<PropertyDescriptor> const& current)
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
{
// 1. Assert: IsPropertyKey(P) is true.
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 2. If current is undefined, then
if (!current.has_value()) {
// a. If extensible is false, return false.
if (!extensible)
return false;
// b. If O is undefined, return true.
if (object == nullptr)
return true;
// c. If IsAccessorDescriptor(Desc) is true, then
if (descriptor.is_accessor_descriptor()) {
// i. Create an own accessor property named P of object O whose [[Get]], [[Set]], [[Enumerable]], and [[Configurable]] attributes are set to the value of the corresponding field in Desc if Desc has that field, or to the attribute's default value otherwise.
auto accessor = Accessor::create(object->vm(), descriptor.get.value_or(nullptr), descriptor.set.value_or(nullptr));
object->storage_set(property_key, { accessor, descriptor.attributes() });
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
}
// d. Else,
else {
// i. Create an own data property named P of object O whose [[Value]], [[Writable]], [[Enumerable]], and [[Configurable]] attributes are set to the value of the corresponding field in Desc if Desc has that field, or to the attribute's default value otherwise.
auto value = descriptor.value.value_or(js_undefined());
object->storage_set(property_key, { value, descriptor.attributes() });
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
}
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// e. Return true.
return true;
}
// 3. Assert: current is a fully populated Property Descriptor.
// 4. If Desc does not have any fields, return true.
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
if (descriptor.is_empty())
return true;
// 5. If current.[[Configurable]] is false, then
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
if (!*current->configurable) {
// a. If Desc has a [[Configurable]] field and Desc.[[Configurable]] is true, return false.
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
if (descriptor.configurable.has_value() && *descriptor.configurable)
return false;
// b. If Desc has an [[Enumerable]] field and SameValue(Desc.[[Enumerable]], current.[[Enumerable]]) is false, return false.
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
if (descriptor.enumerable.has_value() && *descriptor.enumerable != *current->enumerable)
return false;
// c. If IsGenericDescriptor(Desc) is false and SameValue(IsAccessorDescriptor(Desc), IsAccessorDescriptor(current)) is false, return false.
if (!descriptor.is_generic_descriptor() && (descriptor.is_accessor_descriptor() != current->is_accessor_descriptor()))
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
return false;
// d. If IsAccessorDescriptor(current) is true, then
if (current->is_accessor_descriptor()) {
// i. If Desc has a [[Get]] field and SameValue(Desc.[[Get]], current.[[Get]]) is false, return false.
if (descriptor.get.has_value() && *descriptor.get != *current->get)
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
return false;
// ii. If Desc has a [[Set]] field and SameValue(Desc.[[Set]], current.[[Set]]) is false, return false.
if (descriptor.set.has_value() && *descriptor.set != *current->set)
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
return false;
}
// e. Else if current.[[Writable]] is false, then
else if (!*current->writable) {
// i. If Desc has a [[Writable]] field and Desc.[[Writable]] is true, return false.
if (descriptor.writable.has_value() && *descriptor.writable)
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
return false;
// ii. If Desc has a [[Value]] field and SameValue(Desc.[[Value]], current.[[Value]]) is false, return false.
if (descriptor.value.has_value() && (*descriptor.value != *current->value))
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
return false;
}
}
// 6. If O is not undefined, then
if (object != nullptr) {
// a. If IsDataDescriptor(current) is true and IsAccessorDescriptor(Desc) is true, then
if (current->is_data_descriptor() && descriptor.is_accessor_descriptor()) {
// i. If Desc has a [[Configurable]] field, let configurable be Desc.[[Configurable]], else let configurable be current.[[Configurable]].
auto configurable = descriptor.configurable.value_or(*current->configurable);
// ii. If Desc has a [[Enumerable]] field, let enumerable be Desc.[[Enumerable]], else let enumerable be current.[[Enumerable]].
auto enumerable = descriptor.enumerable.value_or(*current->enumerable);
// iii. Replace the property named P of object O with an accessor property having [[Configurable]] and [[Enumerable]] attributes set to configurable and enumerable, respectively, and each other attribute set to its corresponding value in Desc if present, otherwise to its default value.
auto accessor = Accessor::create(object->vm(), descriptor.get.value_or(nullptr), descriptor.set.value_or(nullptr));
PropertyAttributes attributes;
attributes.set_enumerable(enumerable);
attributes.set_configurable(configurable);
object->storage_set(property_key, { accessor, attributes });
}
// b. Else if IsAccessorDescriptor(current) is true and IsDataDescriptor(Desc) is true, then
else if (current->is_accessor_descriptor() && descriptor.is_data_descriptor()) {
// i. If Desc has a [[Configurable]] field, let configurable be Desc.[[Configurable]], else let configurable be current.[[Configurable]].
auto configurable = descriptor.configurable.value_or(*current->configurable);
// ii. If Desc has a [[Enumerable]] field, let enumerable be Desc.[[Enumerable]], else let enumerable be current.[[Enumerable]].
auto enumerable = descriptor.enumerable.value_or(*current->enumerable);
// iii. Replace the property named P of object O with a data property having [[Configurable]] and [[Enumerable]] attributes set to configurable and enumerable, respectively, and each other attribute set to its corresponding value in Desc if present, otherwise to its default value.
auto value = descriptor.value.value_or(js_undefined());
PropertyAttributes attributes;
attributes.set_writable(descriptor.writable.value_or(false));
attributes.set_enumerable(enumerable);
attributes.set_configurable(configurable);
object->storage_set(property_key, { value, attributes });
}
// c. Else,
else {
// i. For each field of Desc, set the corresponding attribute of the property named P of object O to the value of the field.
Value value;
if (descriptor.is_accessor_descriptor() || (current->is_accessor_descriptor() && !descriptor.is_data_descriptor())) {
auto getter = descriptor.get.value_or(current->get.value_or(nullptr));
auto setter = descriptor.set.value_or(current->set.value_or(nullptr));
value = Accessor::create(object->vm(), getter, setter);
} else {
value = descriptor.value.value_or(current->value.value_or({}));
}
PropertyAttributes attributes;
attributes.set_writable(descriptor.writable.value_or(current->writable.value_or(false)));
attributes.set_enumerable(descriptor.enumerable.value_or(current->enumerable.value_or(false)));
attributes.set_configurable(descriptor.configurable.value_or(current->configurable.value_or(false)));
object->storage_set(property_key, { value, attributes });
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
}
}
// 7. Return true.
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
return true;
}
2021-08-08 11:04:30 +01:00
// 10.1.14 GetPrototypeFromConstructor ( constructor, intrinsicDefaultProto ), https://tc39.es/ecma262/#sec-getprototypefromconstructor
ThrowCompletionOr<Object*> get_prototype_from_constructor(VM& vm, FunctionObject const& constructor, GC::Ref<Object> (Intrinsics::*intrinsic_default_prototype)())
{
// 1. Assert: intrinsicDefaultProto is this specification's name of an intrinsic object. The corresponding object must be an intrinsic that is intended to be used as the [[Prototype]] value of an object.
// 2. Let proto be ? Get(constructor, "prototype").
auto prototype = TRY(constructor.get(vm.names.prototype));
// 3. If Type(proto) is not Object, then
if (!prototype.is_object()) {
// a. Let realm be ? GetFunctionRealm(constructor).
auto* realm = TRY(get_function_realm(vm, constructor));
// b. Set proto to realm's intrinsic object named intrinsicDefaultProto.
prototype = (realm->intrinsics().*intrinsic_default_prototype)();
}
// 4. Return proto.
return &prototype.as_object();
}
// 9.1.2.2 NewDeclarativeEnvironment ( E ), https://tc39.es/ecma262/#sec-newdeclarativeenvironment
// 4.1.2.1 NewDeclarativeEnvironment ( E ), https://tc39.es/proposal-explicit-resource-management/#sec-declarative-environment-records-initializebinding-n-v
GC::Ref<DeclarativeEnvironment> new_declarative_environment(Environment& environment)
{
auto& heap = environment.heap();
// 1. Let env be a new Declarative Environment Record containing no bindings.
// 2. Set env.[[OuterEnv]] to E.
// 3. Set env.[[DisposeCapability]] to NewDisposeCapability().
// 4. Return env.
return heap.allocate<DeclarativeEnvironment>(&environment);
}
// 9.1.2.3 NewObjectEnvironment ( O, W, E ), https://tc39.es/ecma262/#sec-newobjectenvironment
GC::Ref<ObjectEnvironment> new_object_environment(Object& object, bool is_with_environment, Environment* environment)
{
auto& heap = object.heap();
// 1. Let env be a new Object Environment Record.
// 2. Set env.[[BindingObject]] to O.
// 3. Set env.[[IsWithEnvironment]] to W.
// 4. Set env.[[OuterEnv]] to E.
// 5. Return env.
return heap.allocate<ObjectEnvironment>(object, is_with_environment ? ObjectEnvironment::IsWithEnvironment::Yes : ObjectEnvironment::IsWithEnvironment::No, environment);
}
// 9.1.2.4 NewFunctionEnvironment ( F, newTarget ), https://tc39.es/ecma262/#sec-newfunctionenvironment
// 4.1.2.2 NewFunctionEnvironment ( F, newTarget ), https://tc39.es/proposal-explicit-resource-management/#sec-newfunctionenvironment
GC::Ref<FunctionEnvironment> new_function_environment(ECMAScriptFunctionObject& function, Object* new_target)
{
auto& heap = function.heap();
// 1. Let env be a new function Environment Record containing no bindings.
auto env = heap.allocate<FunctionEnvironment>(function.environment());
// 2. Set env.[[FunctionObject]] to F.
env->set_function_object(function);
// 3. If F.[[ThisMode]] is lexical, set env.[[ThisBindingStatus]] to lexical.
if (function.this_mode() == ECMAScriptFunctionObject::ThisMode::Lexical)
env->set_this_binding_status(FunctionEnvironment::ThisBindingStatus::Lexical);
// 4. Else, set env.[[ThisBindingStatus]] to uninitialized.
else
env->set_this_binding_status(FunctionEnvironment::ThisBindingStatus::Uninitialized);
// 5. Set env.[[NewTarget]] to newTarget.
env->set_new_target(new_target ?: js_undefined());
// 6. Set env.[[OuterEnv]] to F.[[Environment]].
// 7. Set env.[[DisposeCapability]] to NewDisposeCapability().
// NOTE: Done in step 1 via the FunctionEnvironment constructor.
// 8. Return env.
return env;
}
// 9.2.1.1 NewPrivateEnvironment ( outerPrivEnv ), https://tc39.es/ecma262/#sec-newprivateenvironment
GC::Ref<PrivateEnvironment> new_private_environment(VM& vm, PrivateEnvironment* outer)
2021-10-11 20:29:31 +02:00
{
// 1. Let names be a new empty List.
// 2. Return the PrivateEnvironment Record { [[OuterPrivateEnvironment]]: outerPrivEnv, [[Names]]: names }.
return vm.heap().allocate<PrivateEnvironment>(outer);
2021-10-11 20:29:31 +02:00
}
// 9.4.3 GetThisEnvironment ( ), https://tc39.es/ecma262/#sec-getthisenvironment
GC::Ref<Environment> get_this_environment(VM& vm)
{
// 1. Let env be the running execution context's LexicalEnvironment.
// 2. Repeat,
for (auto* env = vm.lexical_environment(); env; env = env->outer_environment()) {
// a. Let exists be env.HasThisBinding().
// b. If exists is true, return env.
if (env->has_this_binding())
return *env;
// c. Let outer be env.[[OuterEnv]].
// d. Assert: outer is not null.
// e. Set env to outer.
}
VERIFY_NOT_REACHED();
}
// 9.14 CanBeHeldWeakly ( v ), https://tc39.es/proposal-symbols-as-weakmap-keys/#sec-canbeheldweakly-abstract-operation
bool can_be_held_weakly(Value value)
{
// 1. If Type(v) is Object, return true.
if (value.is_object())
return true;
// 2. If Type(v) is Symbol, then
if (value.is_symbol()) {
// a. For each element e of the GlobalSymbolRegistry List (see 19.4.2.2), do
// i. If SameValue(e.[[Symbol]], v) is true, return false.
// b. Return true.
return !value.as_symbol().is_global();
}
// 3. Return false.
return false;
}
// 13.3.7.2 GetSuperConstructor ( ), https://tc39.es/ecma262/#sec-getsuperconstructor
Object* get_super_constructor(VM& vm)
{
// 1. Let envRec be GetThisEnvironment().
auto env = get_this_environment(vm);
// 2. Assert: envRec is a function Environment Record.
// 3. Let activeFunction be envRec.[[FunctionObject]].
// 4. Assert: activeFunction is an ECMAScript function object.
auto& active_function = as<FunctionEnvironment>(*env).function_object();
// 5. Let superConstructor be ! activeFunction.[[GetPrototypeOf]]().
2021-10-03 15:57:26 +01:00
auto* super_constructor = MUST(active_function.internal_get_prototype_of());
// 6. Return superConstructor.
return super_constructor;
}
// 19.2.1.1 PerformEval ( x, strictCaller, direct ), https://tc39.es/ecma262/#sec-performeval
ThrowCompletionOr<Value> perform_eval(VM& vm, Value x, CallerMode strict_caller, EvalMode direct)
{
// 1. Assert: If direct is false, then strictCaller is also false.
VERIFY(direct == EvalMode::Direct || strict_caller == CallerMode::NonStrict);
// 2. If Type(x) is not String, return x.
if (!x.is_string())
return x;
auto& code_string = x.as_string();
// 3. Let evalRealm be the current Realm Record.
auto& eval_realm = *vm.running_execution_context().realm;
// 4. NOTE: In the case of a direct eval, evalRealm is the realm of both the caller of eval and of the eval function itself.
// 5. Perform ? HostEnsureCanCompileStrings(evalRealm, « », x, direct).
TRY(vm.host_ensure_can_compile_strings(eval_realm, {}, code_string.utf8_string_view(), direct));
// 6. Let inFunction be false.
bool in_function = false;
// 7. Let inMethod be false.
bool in_method = false;
// 8. Let inDerivedConstructor be false.
bool in_derived_constructor = false;
// 9. Let inClassFieldInitializer be false.
bool in_class_field_initializer = false;
// 10. If direct is true, then
if (direct == EvalMode::Direct) {
// a. Let thisEnvRec be GetThisEnvironment().
auto this_environment_record = get_this_environment(vm);
// b. If thisEnvRec is a function Environment Record, then
if (is<FunctionEnvironment>(*this_environment_record)) {
auto& this_function_environment_record = static_cast<FunctionEnvironment&>(*this_environment_record);
// i. Let F be thisEnvRec.[[FunctionObject]].
auto& function = this_function_environment_record.function_object();
// ii. Set inFunction to true.
in_function = true;
// iii. Set inMethod to thisEnvRec.HasSuperBinding().
in_method = this_function_environment_record.has_super_binding();
// iv. If F.[[ConstructorKind]] is derived, set inDerivedConstructor to true.
if (function.constructor_kind() == ECMAScriptFunctionObject::ConstructorKind::Derived)
in_derived_constructor = true;
// v. Let classFieldInitializerName be F.[[ClassFieldInitializerName]].
auto& class_field_initializer_name = function.class_field_initializer_name();
// vi. If classFieldInitializerName is not empty, set inClassFieldInitializer to true.
if (!class_field_initializer_name.has<Empty>())
in_class_field_initializer = true;
}
}
// 11. Perform the following substeps in an implementation-defined order, possibly interleaving parsing and error detection:
// a. Let script be ParseText(StringToCodePoints(x), Script).
// c. If script Contains ScriptBody is false, return undefined.
// d. Let body be the ScriptBody of script.
// NOTE: We do these next steps by passing initial state to the parser.
// e. If inFunction is false, and body Contains NewTarget, throw a SyntaxError exception.
// f. If inMethod is false, and body Contains SuperProperty, throw a SyntaxError exception.
// g. If inDerivedConstructor is false, and body Contains SuperCall, throw a SyntaxError exception.
// h. If inClassFieldInitializer is true, and ContainsArguments of body is true, throw a SyntaxError exception.
Parser::EvalInitialState initial_state {
.in_eval_function_context = in_function,
.allow_super_property_lookup = in_method,
.allow_super_constructor_call = in_derived_constructor,
.in_class_field_initializer = in_class_field_initializer,
};
Parser parser { Lexer { code_string.utf8_string_view() }, Program::Type::Script, move(initial_state) };
auto program = parser.parse_program(strict_caller == CallerMode::Strict);
// b. If script is a List of errors, throw a SyntaxError exception.
if (parser.has_errors()) {
auto& error = parser.errors()[0];
return vm.throw_completion<SyntaxError>(error.to_string());
}
bool strict_eval = false;
// 12. If strictCaller is true, let strictEval be true.
if (strict_caller == CallerMode::Strict)
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
strict_eval = true;
// 13. Else, let strictEval be IsStrict of script.
else
strict_eval = program->is_strict_mode();
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// 14. Let runningContext be the running execution context.
// 15. NOTE: If direct is true, runningContext will be the execution context that performed the direct eval. If direct is false, runningContext will be the execution context for the invocation of the eval function.
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
auto& running_context = vm.running_execution_context();
Environment* lexical_environment;
Environment* variable_environment;
2021-10-11 20:29:31 +02:00
PrivateEnvironment* private_environment;
// 16. If direct is true, then
if (direct == EvalMode::Direct) {
// a. Let lexEnv be NewDeclarativeEnvironment(runningContext's LexicalEnvironment).
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
lexical_environment = new_declarative_environment(*running_context.lexical_environment);
// b. Let varEnv be runningContext's VariableEnvironment.
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
variable_environment = running_context.variable_environment;
// c. Let privateEnv be runningContext's PrivateEnvironment.
2021-10-11 20:29:31 +02:00
private_environment = running_context.private_environment;
}
// 17. Else,
else {
// a. Let lexEnv be NewDeclarativeEnvironment(evalRealm.[[GlobalEnv]]).
lexical_environment = new_declarative_environment(eval_realm.global_environment());
// b. Let varEnv be evalRealm.[[GlobalEnv]].
variable_environment = &eval_realm.global_environment();
// c. Let privateEnv be null.
2021-10-11 20:29:31 +02:00
private_environment = nullptr;
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
}
// 18. If strictEval is true, set varEnv to lexEnv.
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (strict_eval)
variable_environment = lexical_environment;
if (direct == EvalMode::Direct && !strict_eval) {
// NOTE: Non-strict direct eval() forces us to deoptimize variable accesses.
// Mark the variable environment chain as screwed since we will not be able
// to rely on cached environment coordinates from this point on.
variable_environment->set_permanently_screwed_by_eval();
}
// 19. If runningContext is not already suspended, suspend runningContext.
// NOTE: Done by the push on step 27.
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// 20. Let evalContext be a new ECMAScript code execution context.
auto eval_context = ExecutionContext::create();
// 21. Set evalContext's Function to null.
// NOTE: This was done in the construction of eval_context.
// 22. Set evalContext's Realm to evalRealm.
eval_context->realm = &eval_realm;
// 23. Set evalContext's ScriptOrModule to runningContext's ScriptOrModule.
eval_context->script_or_module = running_context.script_or_module;
// 24. Set evalContext's VariableEnvironment to varEnv.
eval_context->variable_environment = variable_environment;
// 25. Set evalContext's LexicalEnvironment to lexEnv.
eval_context->lexical_environment = lexical_environment;
// 26. Set evalContext's PrivateEnvironment to privateEnv.
eval_context->private_environment = private_environment;
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// NOTE: This isn't in the spec, but we require it.
eval_context->is_strict_mode = strict_eval;
// 27. Push evalContext onto the execution context stack; evalContext is now the running execution context.
TRY(vm.push_execution_context(*eval_context, {}));
// NOTE: We use a ScopeGuard to automatically pop the execution context when any of the `TRY`s below return a throw completion.
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
ScopeGuard pop_guard = [&] {
// 31. Suspend evalContext and remove it from the execution context stack.
// 32. Resume the context that is now on the top of the execution context stack as the running execution context.
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
vm.pop_execution_context();
};
// 28. Let result be Completion(EvalDeclarationInstantiation(body, varEnv, lexEnv, privateEnv, strictEval)).
TRY(eval_declaration_instantiation(vm, program, variable_environment, lexical_environment, private_environment, strict_eval));
Optional<Value> eval_result;
// 29. If result.[[Type]] is normal, then
// a. Set result to the result of evaluating body.
auto executable_result = Bytecode::Generator::generate_from_ast_node(vm, program, {});
if (executable_result.is_error())
return vm.throw_completion<InternalError>(ErrorType::NotImplemented, TRY_OR_THROW_OOM(vm, executable_result.error().to_string()));
auto executable = executable_result.release_value();
executable->name = "eval"_fly_string;
if (Bytecode::g_dump_bytecode)
executable->dump();
auto result_or_error = vm.bytecode_interpreter().run_executable(*executable, {});
if (result_or_error.value.is_error())
return result_or_error.value.release_error();
eval_result = result_or_error.return_register_value;
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// 30. If result.[[Type]] is normal and result.[[Value]] is empty, then
// a. Set result to NormalCompletion(undefined).
// NOTE: Step 31 and 32 is handled by `pop_guard` above.
// 33. Return ? result.
// NOTE: Step 33 is also performed with each use of `TRY` above.
return eval_result.value_or(js_undefined());
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
}
// 19.2.1.3 EvalDeclarationInstantiation ( body, varEnv, lexEnv, privateEnv, strict ), https://tc39.es/ecma262/#sec-evaldeclarationinstantiation
// 9.1.1.1 EvalDeclarationInstantiation ( body, varEnv, lexEnv, privateEnv, strict ), https://tc39.es/proposal-explicit-resource-management/#sec-evaldeclarationinstantiation
ThrowCompletionOr<void> eval_declaration_instantiation(VM& vm, Program const& program, Environment* variable_environment, Environment* lexical_environment, PrivateEnvironment* private_environment, bool strict)
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
{
auto& realm = *vm.current_realm();
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
GlobalEnvironment* global_var_environment = variable_environment->is_global_environment() ? static_cast<GlobalEnvironment*>(variable_environment) : nullptr;
// 1. Let varNames be the VarDeclaredNames of body.
// 2. Let varDeclarations be the VarScopedDeclarations of body.
// 3. If strict is false, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (!strict) {
// a. If varEnv is a global Environment Record, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (global_var_environment) {
// i. For each element name of varNames, do
TRY(program.for_each_var_declared_identifier([&](auto const& identifier) -> ThrowCompletionOr<void> {
// 1. If varEnv.HasLexicalDeclaration(name) is true, throw a SyntaxError exception.
if (global_var_environment->has_lexical_declaration(identifier.string()))
return vm.throw_completion<SyntaxError>(ErrorType::TopLevelVariableAlreadyDeclared, identifier.string());
// 2. NOTE: eval will not create a global var declaration that would be shadowed by a global lexical declaration.
return {};
}));
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
}
// b. Let thisEnv be lexEnv.
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
auto* this_environment = lexical_environment;
// c. Assert: The following loop will terminate.
// d. Repeat, while thisEnv is not the same as varEnv,
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
while (this_environment != variable_environment) {
// i. If thisEnv is not an object Environment Record, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (!is<ObjectEnvironment>(*this_environment)) {
// 1. NOTE: The environment of with statements cannot contain any lexical declaration so it doesn't need to be checked for var/let hoisting conflicts.
// 2. For each element name of varNames, do
TRY(program.for_each_var_declared_identifier([&](auto const& identifier) -> ThrowCompletionOr<void> {
auto const& name = identifier.string();
// a. If ! thisEnv.HasBinding(name) is true, then
if (MUST(this_environment->has_binding(name))) {
// i. Throw a SyntaxError exception.
return vm.throw_completion<SyntaxError>(ErrorType::TopLevelVariableAlreadyDeclared, name);
// FIXME: ii. NOTE: Annex B.3.4 defines alternate semantics for the above step.
// In particular it only throw the syntax error if it is not an environment from a catchclause.
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
}
// b. NOTE: A direct eval will not hoist var declaration over a like-named lexical declaration.
return {};
}));
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
}
// ii. Set thisEnv to thisEnv.[[OuterEnv]].
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
this_environment = this_environment->outer_environment();
VERIFY(this_environment);
}
}
// 4. Let privateIdentifiers be a new empty List.
// 5. Let pointer be privateEnv.
// 6. Repeat, while pointer is not null,
// a. For each Private Name binding of pointer.[[Names]], do
// i. If privateIdentifiers does not contain binding.[[Description]], append binding.[[Description]] to privateIdentifiers.
// b. Set pointer to pointer.[[OuterPrivateEnvironment]].
// 7. If AllPrivateIdentifiersValid of body with argument privateIdentifiers is false, throw a SyntaxError exception.
2021-10-11 20:29:31 +02:00
// FIXME: Add Private identifiers check here.
// 8. Let functionsToInitialize be a new empty List.
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
Vector<FunctionDeclaration const&> functions_to_initialize;
// 9. Let declaredFunctionNames be a new empty List.
HashTable<FlyString> declared_function_names;
// 10. For each element d of varDeclarations, in reverse List order, do
TRY(program.for_each_var_function_declaration_in_reverse_order([&](FunctionDeclaration const& function) -> ThrowCompletionOr<void> {
// a. If d is neither a VariableDeclaration nor a ForBinding nor a BindingIdentifier, then
// i. Assert: d is either a FunctionDeclaration, a GeneratorDeclaration, an AsyncFunctionDeclaration, or an AsyncGeneratorDeclaration.
// Note: This is done by for_each_var_function_declaration_in_reverse_order.
// ii. NOTE: If there are multiple function declarations for the same name, the last declaration is used.
// iii. Let fn be the sole element of the BoundNames of d.
// iv. If fn is not an element of declaredFunctionNames, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (declared_function_names.set(function.name()) != AK::HashSetResult::InsertedNewEntry)
return {};
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// 1. If varEnv is a global Environment Record, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (global_var_environment) {
// a. Let fnDefinable be ? varEnv.CanDeclareGlobalFunction(fn).
auto function_definable = TRY(global_var_environment->can_declare_global_function(function.name()));
// b. If fnDefinable is false, throw a TypeError exception.
if (!function_definable)
return vm.throw_completion<TypeError>(ErrorType::CannotDeclareGlobalFunction, function.name());
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
}
// 2. Append fn to declaredFunctionNames.
// Note: Already done in step iv.
// 3. Insert d as the first element of functionsToInitialize.
// NOTE: Since prepending is much slower, we just append
// and iterate in reverse order in step 17 below.
functions_to_initialize.append(function);
return {};
}));
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// 11. NOTE: Annex B.3.2.3 adds additional steps at this point.
// B.3.2.3 Changes to EvalDeclarationInstantiation, https://tc39.es/ecma262/#sec-web-compat-evaldeclarationinstantiation
// 11. If strict is false, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (!strict) {
// a. Let declaredFunctionOrVarNames be the list-concatenation of declaredFunctionNames and declaredVarNames.
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// The spec here uses 'declaredVarNames' but that has not been declared yet.
HashTable<FlyString> hoisted_functions;
// b. For each FunctionDeclaration f that is directly contained in the StatementList of a Block, CaseClause, or DefaultClause Contained within body, do
TRY(program.for_each_function_hoistable_with_annexB_extension([&](FunctionDeclaration& function_declaration) -> ThrowCompletionOr<void> {
// i. Let F be StringValue of the BindingIdentifier of f.
auto function_name = function_declaration.name();
// ii. If replacing the FunctionDeclaration f with a VariableStatement that has F as a BindingIdentifier would not produce any Early Errors for body, then
// Note: This is checked during parsing and for_each_function_hoistable_with_annexB_extension so it always passes here.
// 1. Let bindingExists be false.
// 2. Let thisEnv be lexEnv.
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
auto* this_environment = lexical_environment;
// 3. Assert: The following loop will terminate.
// 4. Repeat, while thisEnv is not the same as varEnv,
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
while (this_environment != variable_environment) {
// a. If thisEnv is not an object Environment Record, then
if (!is<ObjectEnvironment>(*this_environment)) {
// i. If ! thisEnv.HasBinding(F) is true, then
if (MUST(this_environment->has_binding(function_name))) {
// i. Let bindingExists be true.
// Note: When bindingExists is true we skip all the other steps.
return {};
}
}
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// b. Set thisEnv to thisEnv.[[OuterEnv]].
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
this_environment = this_environment->outer_environment();
VERIFY(this_environment);
}
// Note: At this point bindingExists is false.
// 5. If bindingExists is false and varEnv is a global Environment Record, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (global_var_environment) {
// a. If varEnv.HasLexicalDeclaration(F) is false, then
if (!global_var_environment->has_lexical_declaration(function_name)) {
// i. Let fnDefinable be ? varEnv.CanDeclareGlobalVar(F).
if (!TRY(global_var_environment->can_declare_global_var(function_name)))
return {};
}
// b. Else,
else {
// i. Let fnDefinable be false.
return {};
}
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
}
// 6. Else,
// a. Let fnDefinable be true.
// Note: At this point fnDefinable is true.
// 7. If bindingExists is false and fnDefinable is true, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// a. If declaredFunctionOrVarNames does not contain F, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (!declared_function_names.contains(function_name) && !hoisted_functions.contains(function_name)) {
// i. If varEnv is a global Environment Record, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (global_var_environment) {
// i. Perform ? varEnv.CreateGlobalVarBinding(F, true).
TRY(global_var_environment->create_global_var_binding(function_name, true));
}
// ii. Else,
else {
// i. Let bindingExists be ! varEnv.HasBinding(F).
// ii. If bindingExists is false, then
if (!MUST(variable_environment->has_binding(function_name))) {
// i. Perform ! varEnv.CreateMutableBinding(F, true).
MUST(variable_environment->create_mutable_binding(vm, function_name, true));
// ii. Perform ! varEnv.InitializeBinding(F, undefined, normal).
MUST(variable_environment->initialize_binding(vm, function_name, js_undefined(), Environment::InitializeBindingHint::Normal));
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
}
}
}
// iii. Append F to declaredFunctionOrVarNames.
hoisted_functions.set(function_name);
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// b. When the FunctionDeclaration f is evaluated, perform the following steps in place of the FunctionDeclaration Evaluation algorithm provided in 15.2.6:
// i. Let genv be the running execution context's VariableEnvironment.
// ii. Let benv be the running execution context's LexicalEnvironment.
// iii. Let fobj be ! benv.GetBindingValue(F, false).
// iv. Perform ? genv.SetMutableBinding(F, fobj, false).
// v. Return unused.
function_declaration.set_should_do_additional_annexB_steps();
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
return {};
}));
}
// 12. Let declaredVarNames be a new empty List.
HashTable<FlyString> declared_var_names;
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// 13. For each element d of varDeclarations, do
TRY(program.for_each_var_scoped_variable_declaration([&](VariableDeclaration const& declaration) {
// a. If d is a VariableDeclaration, a ForBinding, or a BindingIdentifier, then
// Note: This is handled by for_each_var_scoped_variable_declaration.
// i. For each String vn of the BoundNames of d, do
return declaration.for_each_bound_identifier([&](auto const& identifier) -> ThrowCompletionOr<void> {
auto const& name = identifier.string();
// 1. If vn is not an element of declaredFunctionNames, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (!declared_function_names.contains(name)) {
// a. If varEnv is a global Environment Record, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (global_var_environment) {
// i. Let vnDefinable be ? varEnv.CanDeclareGlobalVar(vn).
auto variable_definable = TRY(global_var_environment->can_declare_global_var(name));
// ii. If vnDefinable is false, throw a TypeError exception.
if (!variable_definable)
return vm.throw_completion<TypeError>(ErrorType::CannotDeclareGlobalVariable, name);
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
}
// b. If vn is not an element of declaredVarNames, then
// i. Append vn to declaredVarNames.
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
declared_var_names.set(name);
}
return {};
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
});
}));
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// 14. NOTE: No abnormal terminations occur after this algorithm step unless varEnv is a global Environment Record and the global object is a Proxy exotic object.
// 15. Let lexDeclarations be the LexicallyScopedDeclarations of body.
// 16. For each element d of lexDeclarations, do
TRY(program.for_each_lexically_scoped_declaration([&](Declaration const& declaration) {
// a. NOTE: Lexically declared names are only instantiated here but not initialized.
// b. For each element dn of the BoundNames of d, do
return declaration.for_each_bound_identifier([&](auto const& identifier) -> ThrowCompletionOr<void> {
auto const& name = identifier.string();
// i. If IsConstantDeclaration of d is true, then
if (declaration.is_constant_declaration()) {
// 1. Perform ? lexEnv.CreateImmutableBinding(dn, true).
TRY(lexical_environment->create_immutable_binding(vm, name, true));
}
// ii. Else,
else {
// 1. Perform ? lexEnv.CreateMutableBinding(dn, false).
TRY(lexical_environment->create_mutable_binding(vm, name, false));
}
return {};
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
});
}));
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// 17. For each Parse Node f of functionsToInitialize, do
// NOTE: We iterate in reverse order since we appended the functions
// instead of prepending. We append because prepending is much slower
// and we only use the created vector here.
for (auto& declaration : functions_to_initialize.in_reverse()) {
// a. Let fn be the sole element of the BoundNames of f.
// b. Let fo be InstantiateFunctionObject of f with arguments lexEnv and privateEnv.
auto function = ECMAScriptFunctionObject::create(realm, declaration.name(), declaration.source_text(), declaration.body(), declaration.parameters(), declaration.function_length(), declaration.local_variables_names(), lexical_environment, private_environment, declaration.kind(), declaration.is_strict_mode(),
declaration.parsing_insights());
// c. If varEnv is a global Environment Record, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (global_var_environment) {
// i. Perform ? varEnv.CreateGlobalFunctionBinding(fn, fo, true).
TRY(global_var_environment->create_global_function_binding(declaration.name(), function, true));
}
// d. Else,
else {
// i. Let bindingExists be ! varEnv.HasBinding(fn).
auto binding_exists = MUST(variable_environment->has_binding(declaration.name()));
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// ii. If bindingExists is false, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (!binding_exists) {
// 1. NOTE: The following invocation cannot return an abrupt completion because of the validation preceding step 14.
// 2. Perform ! varEnv.CreateMutableBinding(fn, true).
MUST(variable_environment->create_mutable_binding(vm, declaration.name(), true));
// 3. Perform ! varEnv.InitializeBinding(fn, fo, normal).
MUST(variable_environment->initialize_binding(vm, declaration.name(), function, Environment::InitializeBindingHint::Normal));
}
// iii. Else,
else {
// 1. Perform ! varEnv.SetMutableBinding(fn, fo, false).
MUST(variable_environment->set_mutable_binding(vm, declaration.name(), function, false));
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
}
}
}
// 18. For each String vn of declaredVarNames, do
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
for (auto& var_name : declared_var_names) {
// a. If varEnv is a global Environment Record, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (global_var_environment) {
// i. Perform ? varEnv.CreateGlobalVarBinding(vn, true).
TRY(global_var_environment->create_global_var_binding(var_name, true));
}
// b. Else,
else {
// i. Let bindingExists be ! varEnv.HasBinding(vn).
auto binding_exists = MUST(variable_environment->has_binding(var_name));
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
// ii. If bindingExists is false, then
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
if (!binding_exists) {
// 1. NOTE: The following invocation cannot return an abrupt completion because of the validation preceding step 14.
// 2. Perform ! varEnv.CreateMutableBinding(vn, true).
MUST(variable_environment->create_mutable_binding(vm, var_name, true));
// 3. Perform ! varEnv.InitializeBinding(vn, undefined, normal).
MUST(variable_environment->initialize_binding(vm, var_name, js_undefined(), Environment::InitializeBindingHint::Normal));
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
}
}
}
// 19. Return unused.
LibJS: Make scoping follow the spec Before this we used an ad-hoc combination of references and 'variables' stored in a hashmap. This worked in most cases but is not spec like. Additionally hoisting, dynamically naming functions and scope analysis was not done properly. This patch fixes all of that by: - Implement BindingInitialization for destructuring assignment. - Implementing a new ScopePusher which tracks the lexical and var scoped declarations. This hoists functions to the top level if no lexical declaration name overlaps. Furthermore we do checking of redeclarations in the ScopePusher now requiring less checks all over the place. - Add methods for parsing the directives and statement lists instead of having that code duplicated in multiple places. This allows declarations to pushed to the appropriate scope more easily. - Remove the non spec way of storing 'variables' in DeclarativeEnvironment and make Reference follow the spec instead of checking both the bindings and 'variables'. - Remove all scoping related things from the Interpreter. And instead use environments as specified by the spec. This also includes fixing that NativeFunctions did not produce a valid FunctionEnvironment which could cause issues with callbacks and eval. All FunctionObjects now have a valid NewFunctionEnvironment implementation. - Remove execute_statements from Interpreter and instead use ASTNode::execute everywhere this simplifies AST.cpp as you no longer need to worry about which method to call. - Make ScopeNodes setup their own environment. This uses four different methods specified by the spec {Block, Function, Eval, Global}DeclarationInstantiation with the annexB extensions. - Implement and use NamedEvaluation where specified. Additionally there are fixes to things exposed by these changes to eval, {for, for-in, for-of} loops and assignment. Finally it also fixes some tests in test-js which where passing before but not now that we have correct behavior :^).
2021-09-22 12:44:56 +02:00
return {};
}
// 10.4.4.6 CreateUnmappedArgumentsObject ( argumentsList ), https://tc39.es/ecma262/#sec-createunmappedargumentsobject
Object* create_unmapped_arguments_object(VM& vm, ReadonlySpan<Value> arguments)
{
auto& realm = *vm.current_realm();
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 1. Let len be the number of elements in argumentsList.
auto length = arguments.size();
// 2. Let obj be OrdinaryObjectCreate(%Object.prototype%, « [[ParameterMap]] »).
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 3. Set obj.[[ParameterMap]] to undefined.
auto object = Object::create(realm, realm.intrinsics().object_prototype());
object->set_has_parameter_map();
// 4. Perform ! DefinePropertyOrThrow(obj, "length", PropertyDescriptor { [[Value]]: 𝔽(len), [[Writable]]: true, [[Enumerable]]: false, [[Configurable]]: true }).
MUST(object->define_property_or_throw(vm.names.length, { .value = Value(length), .writable = true, .enumerable = false, .configurable = true }));
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 5. Let index be 0.
// 6. Repeat, while index < len,
for (size_t index = 0; index < length; ++index) {
// a. Let val be argumentsList[index].
auto value = arguments[index];
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// b. Perform ! CreateDataPropertyOrThrow(obj, ! ToString(𝔽(index)), val).
MUST(object->create_data_property_or_throw(index, value));
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// c. Set index to index + 1.
}
// 7. Perform ! DefinePropertyOrThrow(obj, @@iterator, PropertyDescriptor { [[Value]]: %Array.prototype.values%, [[Writable]]: true, [[Enumerable]]: false, [[Configurable]]: true }).
auto array_prototype_values = realm.intrinsics().array_prototype_values_function();
MUST(object->define_property_or_throw(vm.well_known_symbol_iterator(), { .value = array_prototype_values, .writable = true, .enumerable = false, .configurable = true }));
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 8. Perform ! DefinePropertyOrThrow(obj, "callee", PropertyDescriptor { [[Get]]: %ThrowTypeError%, [[Set]]: %ThrowTypeError%, [[Enumerable]]: false, [[Configurable]]: false }).
auto throw_type_error = realm.intrinsics().throw_type_error_function();
MUST(object->define_property_or_throw(vm.names.callee, { .get = throw_type_error, .set = throw_type_error, .enumerable = false, .configurable = false }));
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 9. Return obj.
return object;
}
// 10.4.4.7 CreateMappedArgumentsObject ( func, formals, argumentsList, env ), https://tc39.es/ecma262/#sec-createmappedargumentsobject
Object* create_mapped_arguments_object(VM& vm, FunctionObject& function, NonnullRefPtr<FunctionParameters const> const& formals, ReadonlySpan<Value> arguments, Environment& environment)
{
auto& realm = *vm.current_realm();
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 1. Assert: formals does not contain a rest parameter, any binding patterns, or any initializers. It may contain duplicate identifiers.
// 2. Let len be the number of elements in argumentsList.
VERIFY(arguments.size() <= NumericLimits<i32>::max());
i32 length = static_cast<i32>(arguments.size());
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 3. Let obj be MakeBasicObject(« [[Prototype]], [[Extensible]], [[ParameterMap]] »).
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 4. Set obj.[[GetOwnProperty]] as specified in 10.4.4.1.
// 5. Set obj.[[DefineOwnProperty]] as specified in 10.4.4.2.
// 6. Set obj.[[Get]] as specified in 10.4.4.3.
// 7. Set obj.[[Set]] as specified in 10.4.4.4.
// 8. Set obj.[[Delete]] as specified in 10.4.4.5.
// 9. Set obj.[[Prototype]] to %Object.prototype%.
auto object = realm.create<ArgumentsObject>(realm, environment);
// 14. Let index be 0.
// 15. Repeat, while index < len,
for (i32 index = 0; index < length; ++index) {
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// a. Let val be argumentsList[index].
auto value = arguments[index];
// b. Perform ! CreateDataPropertyOrThrow(obj, ! ToString(𝔽(index)), val).
MUST(object->create_data_property_or_throw(index, value));
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// c. Set index to index + 1.
}
// 16. Perform ! DefinePropertyOrThrow(obj, "length", PropertyDescriptor { [[Value]]: 𝔽(len), [[Writable]]: true, [[Enumerable]]: false, [[Configurable]]: true }).
MUST(object->define_property_or_throw(vm.names.length, { .value = Value(length), .writable = true, .enumerable = false, .configurable = true }));
// 17. Let mappedNames be a new empty List.
HashTable<FlyString> mapped_names;
// 18. Set index to numberOfParameters - 1.
// 19. Repeat, while index ≥ 0,
VERIFY(formals->size() <= NumericLimits<i32>::max());
for (i32 index = static_cast<i32>(formals->size()) - 1; index >= 0; --index) {
// a. Let name be parameterNames[index].
auto const& name = formals->parameters()[index].binding.get<NonnullRefPtr<Identifier const>>()->string();
// b. If name is not an element of mappedNames, then
if (mapped_names.contains(name))
continue;
// i. Add name as an element of the list mappedNames.
mapped_names.set(name);
// ii. If index < len, then
if (index < length) {
// 1. Let g be MakeArgGetter(name, env).
// 2. Let p be MakeArgSetter(name, env).
// 3. Perform ! map.[[DefineOwnProperty]](! ToString(𝔽(index)), PropertyDescriptor { [[Set]]: p, [[Get]]: g, [[Enumerable]]: false, [[Configurable]]: true }).
object->parameter_map().define_native_accessor(
realm,
PropertyKey { index },
[&environment, name](VM& vm) -> ThrowCompletionOr<Value> {
return MUST(environment.get_binding_value(vm, name, false));
},
[&environment, name](VM& vm) {
MUST(environment.set_mutable_binding(vm, name, vm.argument(0), false));
return js_undefined();
},
Attribute::Configurable);
}
}
// 20. Perform ! DefinePropertyOrThrow(obj, @@iterator, PropertyDescriptor { [[Value]]: %Array.prototype.values%, [[Writable]]: true, [[Enumerable]]: false, [[Configurable]]: true }).
auto array_prototype_values = realm.intrinsics().array_prototype_values_function();
MUST(object->define_property_or_throw(vm.well_known_symbol_iterator(), { .value = array_prototype_values, .writable = true, .enumerable = false, .configurable = true }));
// 21. Perform ! DefinePropertyOrThrow(obj, "callee", PropertyDescriptor { [[Value]]: func, [[Writable]]: true, [[Enumerable]]: false, [[Configurable]]: true }).
MUST(object->define_property_or_throw(vm.names.callee, { .value = &function, .writable = true, .enumerable = false, .configurable = true }));
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 22. Return obj.
return object;
}
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
// 7.1.21 CanonicalNumericIndexString ( argument ), https://tc39.es/ecma262/#sec-canonicalnumericindexstring
CanonicalIndex canonical_numeric_index_string(PropertyKey const& property_key, CanonicalIndexMode mode)
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
{
// NOTE: If the property name is a number type (An implementation-defined optimized
// property key type), it can be treated as a string property that has already been
// converted successfully into a canonical numeric index.
VERIFY(property_key.is_string() || property_key.is_number());
if (property_key.is_number())
return CanonicalIndex(CanonicalIndex::Type::Index, property_key.as_number());
if (mode != CanonicalIndexMode::DetectNumericRoundtrip)
return CanonicalIndex(CanonicalIndex::Type::Undefined, 0);
auto& argument = property_key.as_string();
// Handle trivial cases without a full round trip test
// We do not need to check for argument == "0" at this point because we
// already covered it with the is_number() == true path.
if (argument.is_empty())
return CanonicalIndex(CanonicalIndex::Type::Undefined, 0);
u32 current_index = 0;
auto const* characters = argument.bytes_as_string_view().characters_without_null_termination();
auto const length = argument.bytes_as_string_view().length();
if (characters[current_index] == '-') {
current_index++;
if (current_index == length)
return CanonicalIndex(CanonicalIndex::Type::Undefined, 0);
}
if (characters[current_index] == '0') {
current_index++;
if (current_index == length)
return CanonicalIndex(CanonicalIndex::Type::Numeric, 0);
if (characters[current_index] != '.')
return CanonicalIndex(CanonicalIndex::Type::Undefined, 0);
current_index++;
if (current_index == length)
return CanonicalIndex(CanonicalIndex::Type::Undefined, 0);
}
// Short circuit a few common cases
if (argument == "Infinity"sv || argument == "-Infinity"sv || argument == "NaN"sv)
return CanonicalIndex(CanonicalIndex::Type::Numeric, 0);
// Short circuit any string that doesn't start with digits
if (char first_non_zero = characters[current_index]; first_non_zero < '0' || first_non_zero > '9')
return CanonicalIndex(CanonicalIndex::Type::Undefined, 0);
// 2. Let n be ! ToNumber(argument).
auto maybe_double = argument.bytes_as_string_view().to_number<double>(AK::TrimWhitespace::No);
if (!maybe_double.has_value())
return CanonicalIndex(CanonicalIndex::Type::Undefined, 0);
// FIXME: We return 0 instead of n but it might not observable?
// 3. If SameValue(! ToString(n), argument) is true, return n.
if (number_to_string(*maybe_double) == argument)
return CanonicalIndex(CanonicalIndex::Type::Numeric, 0);
// 4. Return undefined.
return CanonicalIndex(CanonicalIndex::Type::Undefined, 0);
LibJS: Rewrite most of Object for spec compliance :^) This is a huge patch, I know. In hindsight this perhaps could've been done slightly more incremental, but I started and then fixed everything until it worked, and here we are. I tried splitting of some completely unrelated changes into separate commits, however. Anyway. This is a rewrite of most of Object, and by extension large parts of Array, Proxy, Reflect, String, TypedArray, and some other things. What we already had worked fine for about 90% of things, but getting the last 10% right proved to be increasingly difficult with the current code that sort of grew organically and is only very loosely based on the spec - this became especially obvious when we started fixing a large number of test262 failures. Key changes include: - 1:1 matching function names and parameters of all object-related functions, to avoid ambiguity. Previously we had things like put(), which the spec doesn't have - as a result it wasn't always clear which need to be used. - Better separation between object abstract operations and internal methods - the former are always the same, the latter can be overridden (and are therefore virtual). The internal methods (i.e. [[Foo]] in the spec) are now prefixed with 'internal_' for clarity - again, it was previously not always clear which AO a certain method represents, get() could've been both Get and [[Get]] (I don't know which one it was closer to right now). Note that some of the old names have been kept until all code relying on them is updated, but they are now simple wrappers around the closest matching standard abstract operation. - Simplifications of the storage layer: functions that write values to storage are now prefixed with 'storage_' to make their purpose clear, and as they are not part of the spec they should not contain any steps specified by it. Much functionality is now covered by the layers above it and was removed (e.g. handling of accessors, attribute checks). - PropertyAttributes has been greatly simplified, and is being replaced by PropertyDescriptor - a concept similar to the current implementation, but more aligned with the actual spec. See the commit message of the previous commit where it was introduced for details. - As a bonus, and since I had to look at the spec a whole lot anyway, I introduced more inline comments with the exact steps from the spec - this makes it super easy to verify correctness. - East-const all the things. As a result of all of this, things are much more correct but a bit slower now. Retaining speed wasn't a consideration at all, I have done no profiling of the new code - there might be low hanging fruits, which we can then harvest separately. Special thanks to Idan for helping me with this by tracking down bugs, updating everything outside of LibJS to work with these changes (LibWeb, Spreadsheet, HackStudio), as well as providing countless patches to fix regressions I introduced - there still are very few (we got it down to 5), but we also get many new passing test262 tests in return. :^) Co-authored-by: Idan Horowitz <idan.horowitz@gmail.com>
2021-07-04 18:14:16 +01:00
}
// 22.1.3.19.1 GetSubstitution ( matched, str, position, captures, namedCaptures, replacementTemplate ), https://tc39.es/ecma262/#sec-getsubstitution
ThrowCompletionOr<String> get_substitution(VM& vm, Utf16View const& matched, Utf16View const& str, size_t position, Span<Value> captures, Value named_captures, Value replacement_template)
{
// 1. Let stringLength be the length of str.
auto string_length = str.length_in_code_units();
// 2. Assert: position ≤ stringLength.
VERIFY(position <= string_length);
// 3. Let result be the empty String.
Utf16Data result;
// 4. Let templateRemainder be replacementTemplate.
auto replace_template_string = TRY(replacement_template.to_utf16_string(vm));
auto template_remainder = replace_template_string.view();
// 5. Repeat, while templateRemainder is not the empty String,
while (!template_remainder.is_empty()) {
// a. NOTE: The following steps isolate ref (a prefix of templateRemainder), determine refReplacement (its replacement), and then append that replacement to result.
Utf16View ref;
Utf16View ref_replacement;
Optional<Utf16String> capture_string;
// b. If templateRemainder starts with "$$", then
if (template_remainder.starts_with(u"$$")) {
// i. Let ref be "$$".
ref = u"$$";
// ii. Let refReplacement be "$".
ref_replacement = u"$";
}
// c. Else if templateRemainder starts with "$`", then
else if (template_remainder.starts_with(u"$`")) {
// i. Let ref be "$`".
ref = u"$`";
// ii. Let refReplacement be the substring of str from 0 to position.
ref_replacement = str.substring_view(0, position);
}
// d. Else if templateRemainder starts with "$&", then
else if (template_remainder.starts_with(u"$&")) {
// i. Let ref be "$&".
ref = u"$&";
// ii. Let refReplacement be matched.
ref_replacement = matched;
}
// e. Else if templateRemainder starts with "$'" (0x0024 (DOLLAR SIGN) followed by 0x0027 (APOSTROPHE)), then
else if (template_remainder.starts_with(u"$'")) {
// i. Let ref be "$'".
ref = u"$'";
// ii. Let matchLength be the length of matched.
auto match_length = matched.length_in_code_units();
// iii. Let tailPos be position + matchLength.
auto tail_pos = position + match_length;
// iv. Let refReplacement be the substring of str from min(tailPos, stringLength).
ref_replacement = str.substring_view(min(tail_pos, string_length));
// v. NOTE: tailPos can exceed stringLength only if this abstract operation was invoked by a call to the intrinsic @@replace method of %RegExp.prototype% on an object whose "exec" property is not the intrinsic %RegExp.prototype.exec%.
}
// f. Else if templateRemainder starts with "$" followed by 1 or more decimal digits, then
else if (template_remainder.starts_with(u"$") && template_remainder.length_in_code_units() > 1 && is_ascii_digit(template_remainder.code_unit_at(1))) {
// i. If templateRemainder starts with "$" followed by 2 or more decimal digits, let digitCount be 2. Otherwise, let digitCount be 1.
size_t digit_count = 1;
if (template_remainder.length_in_code_units() > 2 && is_ascii_digit(template_remainder.code_point_at(2)))
digit_count = 2;
// ii. Let digits be the substring of templateRemainder from 1 to 1 + digitCount.
auto digits = template_remainder.substring_view(1, digit_count);
// iii. Let index be (StringToNumber(digits)).
auto utf8_digits = MUST(digits.to_utf8());
auto index = static_cast<size_t>(string_to_number(utf8_digits));
// iv. Assert: 0 ≤ index ≤ 99.
VERIFY(index <= 99);
// v. Let captureLen be the number of elements in captures.
auto capture_length = captures.size();
// vi. If index > captureLen and digitCount = 2, then
if (index > capture_length && digit_count == 2) {
// 1. NOTE: When a two-digit replacement pattern specifies an index exceeding the count of capturing groups, it is treated as a one-digit replacement pattern followed by a literal digit.
// 2. Set digitCount to 1.
digit_count = 1;
// 3. Set digits to the substring of digits from 0 to 1.
digits = digits.substring_view(0, 1);
// 4. Set index to (StringToNumber(digits)).
utf8_digits = MUST(digits.to_utf8());
index = static_cast<size_t>(string_to_number(utf8_digits));
}
// vii. Let ref be the substring of templateRemainder from 0 to 1 + digitCount.
ref = template_remainder.substring_view(0, 1 + digit_count);
// viii. If 1 ≤ index ≤ captureLen, then
if (1 <= index && index <= capture_length) {
// 1. Let capture be captures[index - 1].
auto capture = captures[index - 1];
// 2. If capture is undefined, then
if (capture.is_undefined()) {
// a. Let refReplacement be the empty String.
ref_replacement = {};
}
// 3. Else,
else {
// a. Let refReplacement be capture.
capture_string = TRY(capture.to_utf16_string(vm));
ref_replacement = capture_string->view();
}
}
// ix. Else,
else {
// 1. Let refReplacement be ref.
ref_replacement = ref;
}
}
// g. Else if templateRemainder starts with "$<", then
else if (template_remainder.starts_with(u"$<")) {
// i. Let gtPos be StringIndexOf(templateRemainder, ">", 0).
// NOTE: We can actually start at index 2 because we know the string starts with "$<".
auto greater_than_position = string_index_of(template_remainder, u">", 2);
// ii. If gtPos = -1 or namedCaptures is undefined, then
if (!greater_than_position.has_value() || named_captures.is_undefined()) {
// 1. Let ref be "$<".
ref = u"$<";
// 2. Let refReplacement be ref.
ref_replacement = ref;
}
// iii. Else,
else {
// 1. Let ref be the substring of templateRemainder from 0 to gtPos + 1.
ref = template_remainder.substring_view(0, *greater_than_position + 1);
// 2. Let groupName be the substring of templateRemainder from 2 to gtPos.
auto group_name_view = template_remainder.substring_view(2, *greater_than_position - 2);
auto group_name = MUST(group_name_view.to_utf8(Utf16View::AllowInvalidCodeUnits::Yes));
// 3. Assert: namedCaptures is an Object.
VERIFY(named_captures.is_object());
// 4. Let capture be ? Get(namedCaptures, groupName).
auto capture = TRY(named_captures.as_object().get(group_name));
// 5. If capture is undefined, then
if (capture.is_undefined()) {
// a. Let refReplacement be the empty String.
ref_replacement = {};
}
// 6. Else,
else {
// a. Let refReplacement be ? ToString(capture).
capture_string = TRY(capture.to_utf16_string(vm));
ref_replacement = capture_string->view();
}
}
}
// h. Else,
else {
// i. Let ref be the substring of templateRemainder from 0 to 1.
ref = template_remainder.substring_view(0, 1);
// ii. Let refReplacement be ref.
ref_replacement = ref;
}
// i. Let refLength be the length of ref.
auto ref_length = ref.length_in_code_units();
// k. Set result to the string-concatenation of result and refReplacement.
result.append(ref_replacement.data(), ref_replacement.length_in_code_points());
// j. Set templateRemainder to the substring of templateRemainder from refLength.
// NOTE: We do this step last because refReplacement may point to templateRemainder.
template_remainder = template_remainder.substring_view(ref_length);
}
// 6. Return result.
return MUST(Utf16View { result }.to_utf8(Utf16View::AllowInvalidCodeUnits::Yes));
}
void DisposeCapability::visit_edges(GC::Cell::Visitor& visitor) const
{
for (auto const& disposable_resource : disposable_resource_stack)
disposable_resource.visit_edges(visitor);
}
void DisposableResource::visit_edges(GC::Cell::Visitor& visitor) const
{
visitor.visit(resource_value);
visitor.visit(dispose_method);
}
// 2.1.3 NewDisposeCapability ( ), https://tc39.es/proposal-explicit-resource-management/#sec-newdisposecapability
DisposeCapability new_dispose_capability()
{
// 1. Let stack be a new empty List.
// 2. Return the DisposeCapability Record { [[DisposableResourceStack]]: stack }.
return DisposeCapability {};
}
// 2.1.4 AddDisposableResource ( disposeCapability, V, hint [ , method ] ), https://tc39.es/proposal-explicit-resource-management/#sec-adddisposableresource-disposable-v-hint-disposemethod
ThrowCompletionOr<void> add_disposable_resource(VM& vm, DisposeCapability& dispose_capability, Value value, Environment::InitializeBindingHint hint, GC::Ptr<FunctionObject> method)
{
Optional<DisposableResource> resource;
// 1. If method is not present then,
if (!method) {
// a. If V is either null or undefined and hint is sync-dispose, then
if (value.is_nullish() && hint == Environment::InitializeBindingHint::SyncDispose) {
// i. Return unused.
return {};
}
// b. NOTE: When V is either null or undefined and hint is async-dispose, we record that the resource was evaluated
// to ensure we will still perform an Await when resources are later disposed.
// c. Let resource be ? CreateDisposableResource(V, hint).
resource = TRY(create_disposable_resource(vm, value, hint));
}
// 2. Else,
else {
// a. Assert: V is undefined.
VERIFY(value.is_undefined());
// b. Let resource be ? CreateDisposableResource(undefined, hint, method).
resource = TRY(create_disposable_resource(vm, js_undefined(), hint, method));
}
// 3. Append resource to disposeCapability.[[DisposableResourceStack]].
dispose_capability.disposable_resource_stack.append(resource.release_value());
// 4. Return unused.
return {};
}
// 2.1.5 CreateDisposableResource ( V, hint [ , method ] ), https://tc39.es/proposal-explicit-resource-management/#sec-createdisposableresource
ThrowCompletionOr<DisposableResource> create_disposable_resource(VM& vm, Value value, Environment::InitializeBindingHint hint, GC::Ptr<FunctionObject> method)
{
// 1. If method is not present, then
if (!method) {
// a. If V is either null or undefined, then
if (value.is_nullish()) {
// i. Set V to undefined.
// ii. Set method to undefined.
}
// b. Else,
else {
// i. If V is not an Object, throw a TypeError exception.
if (!value.is_object())
return vm.throw_completion<TypeError>(ErrorType::NotAnObject, value);
// ii. Set method to ? GetDisposeMethod(V, hint).
method = TRY(get_dispose_method(vm, value, hint));
// iii. If method is undefined, throw a TypeError exception.
if (!method)
return vm.throw_completion<TypeError>(ErrorType::NoDisposeMethod, value);
}
}
// 2. Else,
else {
// a. If IsCallable(method) is false, throw a TypeError exception.
// NOTE: This is guaranteed to never occur due to its type.
}
// 3. Return the DisposableResource Record { [[ResourceValue]]: V, [[Hint]]: hint, [[DisposeMethod]]: method }.
return DisposableResource {
.resource_value = value.is_object() ? GC::Ptr { value.as_object() } : nullptr,
.hint = hint,
.dispose_method = method,
};
}
// 2.1.6 GetDisposeMethod ( V, hint ), https://tc39.es/proposal-explicit-resource-management/#sec-getdisposemethod
ThrowCompletionOr<GC::Ptr<FunctionObject>> get_dispose_method(VM& vm, Value value, Environment::InitializeBindingHint hint)
{
GC::Ptr<FunctionObject> method;
// 1. If hint is async-dispose, then
if (hint == Environment::InitializeBindingHint::AsyncDispose) {
// a. Let method be ? GetMethod(V, @@asyncDispose).
method = TRY(value.get_method(vm, vm.well_known_symbol_async_dispose()));
// b. If method is undefined, then
if (!method) {
// i. Set method to ? GetMethod(V, @@dispose).
method = TRY(value.get_method(vm, vm.well_known_symbol_dispose()));
// ii. If method is not undefined, then
if (method) {
auto& realm = *vm.current_realm();
// 1. Let closure be a new Abstract Closure with no parameters that captures method and performs the following steps when called:
auto closure = [&realm, method](VM& vm) -> ThrowCompletionOr<Value> {
// a. Let O be the this value.
auto object = vm.this_value();
// b. Let promiseCapability be ! NewPromiseCapability(%Promise%).
auto promise_capability = MUST(new_promise_capability(vm, realm.intrinsics().promise_constructor()));
// c. Let result be Completion(Call(method, O)).
// d. IfAbruptRejectPromise(result, promiseCapability).
TRY_OR_REJECT(vm, promise_capability, call(vm, method, object));
// e. Perform ? Call(promiseCapability.[[Resolve]], undefined, « undefined »).
TRY(call(vm, *promise_capability->resolve(), js_undefined(), js_undefined()));
// f. Return promiseCapability.[[Promise]].
return promise_capability->promise();
};
// 2. NOTE: This function is not observable to user code. It is used to ensure that a Promise returned
// from a synchronous @@dispose method will not be awaited and that any exception thrown will not be
// thrown synchronously.
// 3. Return CreateBuiltinFunction(closure, 0, "", « »).
return NativeFunction::create(realm, move(closure), 0);
}
}
}
// 2. Else,
else {
// a. Let method be ? GetMethod(V, @@dispose).
method = TRY(value.get_method(vm, vm.well_known_symbol_dispose()));
}
// 3. Return method.
return method;
}
// 2.1.7 Dispose ( V, hint, method ), https://tc39.es/proposal-explicit-resource-management/#sec-dispose
Completion dispose(VM& vm, Value value, Environment::InitializeBindingHint hint, GC::Ptr<FunctionObject> method)
{
Value result;
// 1. If method is undefined, let result be undefined.
if (!method) {
result = js_undefined();
}
// 2. Else, let result be ? Call(method, V).
else {
result = TRY(call(vm, *method, value));
}
// 3. If hint is async-dispose, then
if (hint == Environment::InitializeBindingHint::AsyncDispose) {
// a. Perform ? Await(result).
TRY(await(vm, result));
}
// 4. Return undefined.
return js_undefined();
}
// 2.1.8 DisposeResources ( disposeCapability, completion ), https://tc39.es/proposal-explicit-resource-management/#sec-disposeresources
Completion dispose_resources(VM& vm, DisposeCapability& dispose_capability, Completion completion)
{
// 1. Let needsAwait be false.
bool needs_await = false;
// 2. Let hasAwaited be false.
bool has_awaited = false;
// 3. For each element resource of disposeCapability.[[DisposableResourceStack]], in reverse list order, do
for (auto const& resource : dispose_capability.disposable_resource_stack.in_reverse()) {
// a. Let value be resource.[[ResourceValue]].
auto value = resource.resource_value;
// b. Let hint be resource.[[Hint]].
auto hint = resource.hint;
// c. Let method be resource.[[DisposeMethod]].
auto method = resource.dispose_method;
// d. If hint is sync-dispose and needsAwait is true and hasAwaited is false, then
if (hint == Environment::InitializeBindingHint::SyncDispose && needs_await && !has_awaited) {
// i. Perform ! Await(undefined).
MUST(await(vm, js_undefined()));
// ii. Set needsAwait to false.
needs_await = false;
}
// e. If method is not undefined, then
if (method) {
// i. Let result be Completion(Call(method, value)).
auto result = call(vm, *method, value);
// ii. If result is a normal completion and hint is async-dispose, then
if (!result.is_throw_completion() && hint == Environment::InitializeBindingHint::AsyncDispose) {
// 1. Set result to Completion(Await(result.[[Value]])).
result = await(vm, result.value());
// 2. Set hasAwaited to true.
has_awaited = true;
}
// iii. If result is a throw completion, then
else if (result.is_throw_completion()) {
// 1. If completion is a throw completion, then
if (completion.type() == Completion::Type::Throw) {
// a. Set result to result.[[Value]].
auto result_value = result.error().value();
// b. Let suppressed be completion.[[Value]].
auto suppressed = completion.value();
// c. Let error be a newly created SuppressedError object.
auto error = SuppressedError::create(*vm.current_realm());
// d. Perform CreateNonEnumerableDataPropertyOrThrow(error, "error", result).
error->create_non_enumerable_data_property_or_throw(vm.names.error, result_value);
// e. Perform CreateNonEnumerableDataPropertyOrThrow(error, "suppressed", suppressed).
error->create_non_enumerable_data_property_or_throw(vm.names.suppressed, suppressed);
// f. Set completion to ThrowCompletion(error).
completion = throw_completion(error);
}
// 2. Else,
else {
// a. Set completion to result.
completion = result.release_error();
}
}
}
// f. Else,
else {
// i. Assert: hint is async-dispose.
VERIFY(hint == Environment::InitializeBindingHint::AsyncDispose);
// ii. Set needsAwait to true.
needs_await = true;
// iii. NOTE: This can only indicate a case where either null or undefined was the initialized value of an
// await using declaration.
}
}
// 4. If needsAwait is true and hasAwaited is false, then
if (needs_await && !has_awaited) {
// a. Perform ! Await(undefined).
MUST(await(vm, js_undefined()));
}
// 5. NOTE: After disposeCapability has been disposed, it will never be used again. The contents of
// disposeCapability.[[DisposableResourceStack]] can be discarded in implementations, such as by garbage
// collection, at this point.
// 6. Set disposeCapability.[[DisposableResourceStack]] to a new empty List.
dispose_capability.disposable_resource_stack.clear();
// 7. Return completion.
return completion;
}
// https://tc39.es/proposal-import-attributes/#sec-AllImportAttributesSupported
static bool all_import_attributes_supported(VM& vm, Vector<ImportAttribute> const& attributes)
{
// 1. Let supported be HostGetSupportedImportAttributes().
auto supported = vm.host_get_supported_import_attributes();
// 2. For each ImportAttribute Record attribute of attributes, do
for (auto const& attribute : attributes) {
// a. If supported does not contain attribute.[[Key]], return false.
if (!supported.contains_slow(attribute.key))
return false;
}
// 3. Return true.
return true;
}
// 13.3.10.2 EvaluateImportCall ( specifierExpression [ , optionsExpression ] ), https://tc39.es/proposal-import-attributes/#sec-evaluate-import-call
ThrowCompletionOr<Value> perform_import_call(VM& vm, Value specifier, Value options_value)
{
auto& realm = *vm.current_realm();
// 1. Let referrer be GetActiveScriptOrModule().
auto referrer = [&]() -> ImportedModuleReferrer {
auto active_script_or_module = vm.get_active_script_or_module();
// 2. If referrer is null, set referrer to the current Realm Record.
if (active_script_or_module.has<Empty>())
return GC::Ref<Realm> { realm };
if (active_script_or_module.has<GC::Ref<Script>>())
return active_script_or_module.get<GC::Ref<Script>>();
return GC::Ref<CyclicModule> { as<CyclicModule>(*active_script_or_module.get<GC::Ref<Module>>()) };
}();
// 7. Let promiseCapability be ! NewPromiseCapability(%Promise%).
auto promise_capability = MUST(new_promise_capability(vm, realm.intrinsics().promise_constructor()));
// 8. Let specifierString be Completion(ToString(specifier)).
// 9. IfAbruptRejectPromise(specifierString, promiseCapability).
auto specifier_string = TRY_OR_REJECT_WITH_VALUE(vm, promise_capability, specifier.to_string(vm));
// 10. Let attributes be a new empty List.
Vector<ImportAttribute> attributes;
// 11. If options is not undefined, then
if (!options_value.is_undefined()) {
// a. If Type(options) is not Object,
if (!options_value.is_object()) {
auto error = TypeError::create(realm, TRY_OR_THROW_OOM(vm, String::formatted(ErrorType::NotAnObject.message(), "ImportOptions")));
// i. Perform ! Call(promiseCapability.[[Reject]], undefined, « a newly created TypeError object »).
MUST(call(vm, *promise_capability->reject(), js_undefined(), error));
// ii. Return promiseCapability.[[Promise]].
return Value { promise_capability->promise() };
}
// b. Let attributesObj be Completion(Get(options, "with")).
// c. IfAbruptRejectPromise(attributesObj, promiseCapability).
auto attributes_obj = TRY_OR_REJECT_WITH_VALUE(vm, promise_capability, options_value.get(vm, vm.names.with));
// d. If attributesObj is not undefined, then
if (!attributes_obj.is_undefined()) {
// i. If Type(attributesObj) is not Object, then
if (!attributes_obj.is_object()) {
auto error = TypeError::create(realm, TRY_OR_THROW_OOM(vm, String::formatted(ErrorType::NotAnObject.message(), "ImportOptionsAssertions")));
// 1. Perform ! Call(promiseCapability.[[Reject]], undefined, « a newly created TypeError object »).
MUST(call(vm, *promise_capability->reject(), js_undefined(), error));
// 2. Return promiseCapability.[[Promise]].
return Value { promise_capability->promise() };
}
// ii. Let entries be Completion(EnumerableOwnProperties(attributesObj, key+value)).
// iii. IfAbruptRejectPromise(entries, promiseCapability).
auto entries = TRY_OR_REJECT_WITH_VALUE(vm, promise_capability, attributes_obj.as_object().enumerable_own_property_names(Object::PropertyKind::KeyAndValue));
// iv. For each entry of entries, do
for (auto const& entry : entries) {
// 1. Let key be ! Get(entry, "0").
auto key = MUST(entry.get(vm, PropertyKey(0)));
// 2. Let value be ! Get(entry, "1").
auto value = MUST(entry.get(vm, PropertyKey(1)));
// 3. If Type(value) is not String, then
if (!value.is_string()) {
auto error = TypeError::create(realm, TRY_OR_THROW_OOM(vm, String::formatted(ErrorType::NotAString.message(), "Import Assertion option value")));
// a. Perform ! Call(promiseCapability.[[Reject]], undefined, « a newly created TypeError object »).
MUST(call(vm, *promise_capability->reject(), js_undefined(), error));
// b. Return promiseCapability.[[Promise]].
return Value { promise_capability->promise() };
}
// 4. Append the ImportAttribute Record { [[Key]]: key, [[Value]]: value } to attributes.
attributes.empend(key.as_string().utf8_string(), value.as_string().utf8_string());
}
}
// e. If AllImportAttributesSupported(attributes) is false, then
if (!all_import_attributes_supported(vm, attributes)) {
auto error = TypeError::create(realm, TRY_OR_THROW_OOM(vm, String::formatted(ErrorType::NotAnObject.message(), "ImportOptionsAssertions")));
// i. Perform ! Call(promiseCapability.[[Reject]], undefined, « a newly created TypeError object »).
MUST(call(vm, *promise_capability->reject(), js_undefined(), error));
// ii. Return promiseCapability.[[Promise]].
return Value { promise_capability->promise() };
}
// f. Sort attributes according to the lexicographic order of their [[Key]] fields,
// treating the value of each such field as a sequence of UTF-16 code unit values.
// NOTE: This sorting is observable only in that hosts are prohibited from
// distinguishing among attributes by the order they occur in.
// NOTE: This is done when constructing the ModuleRequest.
}
// 12. Let moduleRequest be a new ModuleRequest Record { [[Specifier]]: specifierString, [[Attributes]]: attributes }.
ModuleRequest request { specifier_string, attributes };
// 13. Perform HostLoadImportedModule(referrer, moduleRequest, empty, promiseCapability).
vm.host_load_imported_module(referrer, move(request), nullptr, promise_capability);
// 13. Return promiseCapability.[[Promise]].
return Value { promise_capability->promise() };
}
// 14.5.2.1 GetOptionsObject ( options ), https://tc39.es/proposal-temporal/#sec-getoptionsobject
ThrowCompletionOr<GC::Ref<Object>> get_options_object(VM& vm, Value options)
{
auto& realm = *vm.current_realm();
// 1. If options is undefined, then
if (options.is_undefined()) {
// a. Return OrdinaryObjectCreate(null).
return Object::create(realm, nullptr);
}
// 2. If options is an Object, then
if (options.is_object()) {
// a. Return options.
return options.as_object();
}
// 3. Throw a TypeError exception.
return vm.throw_completion<TypeError>(ErrorType::NotAnObject, "Options");
}
// 14.5.2.2 GetOption ( options, property, type, values, default ), https://tc39.es/proposal-temporal/#sec-getoption
ThrowCompletionOr<Value> get_option(VM& vm, Object const& options, PropertyKey const& property, OptionType type, ReadonlySpan<StringView> values, OptionDefault const& default_)
{
VERIFY(property.is_string());
// 1. Let value be ? Get(options, property).
auto value = TRY(options.get(property));
// 2. If value is undefined, then
if (value.is_undefined()) {
// a. If default is REQUIRED, throw a RangeError exception.
if (default_.has<Required>())
return vm.throw_completion<RangeError>(ErrorType::OptionIsNotValidValue, "undefined"sv, property.as_string());
// b. Return default.
return default_.visit(
[](Required) -> Value { VERIFY_NOT_REACHED(); },
[](Empty) -> Value { return js_undefined(); },
[](bool default_) -> Value { return Value { default_ }; },
[](double default_) -> Value { return Value { default_ }; },
[&](StringView default_) -> Value { return PrimitiveString::create(vm, default_); });
}
// 3. If type is BOOLEAN, then
if (type == OptionType::Boolean) {
// a. Set value to ToBoolean(value).
value = Value { value.to_boolean() };
}
// 4. Else,
else {
// a. Assert: type is STRING.
VERIFY(type == OptionType::String);
// b. Set value to ? ToString(value).
value = TRY(value.to_primitive_string(vm));
}
// 5. If values is not EMPTY and values does not contain value, throw a RangeError exception.
if (!values.is_empty()) {
// NOTE: Every location in the spec that invokes GetOption with type=boolean also has values=undefined.
VERIFY(value.is_string());
if (auto value_string = value.as_string().utf8_string(); !values.contains_slow(value_string))
return vm.throw_completion<RangeError>(ErrorType::OptionIsNotValidValue, value_string, property.as_string());
}
// 6. Return value.
return value;
}
// 14.5.2.3 GetRoundingModeOption ( options, fallback ), https://tc39.es/proposal-temporal/#sec-temporal-getroundingmodeoption
ThrowCompletionOr<RoundingMode> get_rounding_mode_option(VM& vm, Object const& options, RoundingMode fallback)
{
// 1. Let allowedStrings be the List of Strings from the "String Identifier" column of Table 26.
static constexpr auto allowed_strings = to_array({ "ceil"sv, "floor"sv, "expand"sv, "trunc"sv, "halfCeil"sv, "halfFloor"sv, "halfExpand"sv, "halfTrunc"sv, "halfEven"sv });
// 2. Let stringFallback be the value from the "String Identifier" column of the row with fallback in its "Rounding Mode" column.
auto string_fallback = allowed_strings[to_underlying(fallback)];
// 3. Let stringValue be ? GetOption(options, "roundingMode", STRING, allowedStrings, stringFallback).
auto string_value = TRY(get_option(vm, options, vm.names.roundingMode, OptionType::String, allowed_strings, string_fallback));
// 4. Return the value from the "Rounding Mode" column of the row with stringValue in its "String Identifier" column.
return static_cast<RoundingMode>(allowed_strings.first_index_of(string_value.as_string().utf8_string_view()).value());
}
// 14.5.2.4 GetRoundingIncrementOption ( options ), https://tc39.es/proposal-temporal/#sec-temporal-getroundingincrementoption
ThrowCompletionOr<u64> get_rounding_increment_option(VM& vm, Object const& options)
{
// 1. Let value be ? Get(options, "roundingIncrement").
auto value = TRY(options.get(vm.names.roundingIncrement));
// 2. If value is undefined, return 1𝔽.
if (value.is_undefined())
return 1;
// 3. Let integerIncrement be ? ToIntegerWithTruncation(value).
auto integer_increment = TRY(Temporal::to_integer_with_truncation(vm, value, ErrorType::OptionIsNotValidValue, value, "roundingIncrement"sv));
// 4. If integerIncrement < 1 or integerIncrement > 10**9, throw a RangeError exception.
if (integer_increment < 1 || integer_increment > 1'000'000'000u)
return vm.throw_completion<RangeError>(ErrorType::OptionIsNotValidValue, value, "roundingIncrement");
// 5. Return integerIncrement.
return static_cast<u64>(integer_increment);
}
// AD-HOC
// FIXME: We should add a generic floor() method to our BigInt classes. But for now, since we know we are only dividing
// by powers of 10, we can implement a very situationally specific method to compute the floor of a division.
Crypto::SignedBigInteger big_floor(Crypto::SignedBigInteger const& numerator, Crypto::UnsignedBigInteger const& denominator)
{
auto result = numerator.divided_by(denominator);
if (result.remainder.is_zero())
return result.quotient;
if (!result.quotient.is_negative() && result.remainder.is_positive())
return result.quotient;
return result.quotient.minus(Crypto::SignedBigInteger { 1 });
}
}