mirror of
				https://github.com/python/cpython.git
				synced 2025-10-31 05:31:20 +00:00 
			
		
		
		
	
		
			
				
	
	
		
			143 lines
		
	
	
	
		
			3.9 KiB
		
	
	
	
		
			Python
		
	
	
		
			Executable file
		
	
	
	
	
			
		
		
	
	
			143 lines
		
	
	
	
		
			3.9 KiB
		
	
	
	
		
			Python
		
	
	
		
			Executable file
		
	
	
	
	
| #! /usr/bin/env python
 | |
| 
 | |
| """Convert a LaTeX .toc file to some PDFTeX magic to create that neat outline.
 | |
| 
 | |
| The output file has an extension of '.bkm' instead of '.out', since hyperref
 | |
| already uses that extension.
 | |
| """
 | |
| 
 | |
| import getopt
 | |
| import os
 | |
| import re
 | |
| import string
 | |
| import sys
 | |
| 
 | |
| 
 | |
| # Ench item in an entry is a tuple of:
 | |
| #
 | |
| #   Section #,  Title String,  Page #,  List of Sub-entries
 | |
| #
 | |
| # The return value of parse_toc() is such a tuple.
 | |
| 
 | |
| cline_re = r"""^
 | |
| \\contentsline\ \{([a-z]*)}             # type of section in $1
 | |
| \{(?:\\numberline\ \{([0-9.A-Z]+)})?     # section number
 | |
| (.*)}                                   # title string
 | |
| \{(\d+)}$"""                            # page number
 | |
| 
 | |
| cline_rx = re.compile(cline_re, re.VERBOSE)
 | |
| 
 | |
| OUTER_TO_INNER = -1
 | |
| 
 | |
| _transition_map = {
 | |
|     ('chapter', 'section'): OUTER_TO_INNER,
 | |
|     ('section', 'subsection'): OUTER_TO_INNER,
 | |
|     ('subsection', 'subsubsection'): OUTER_TO_INNER,
 | |
|     ('subsubsection', 'subsection'): 1,
 | |
|     ('subsection', 'section'): 1,
 | |
|     ('section', 'chapter'): 1,
 | |
|     ('subsection', 'chapter'): 2,
 | |
|     ('subsubsection', 'section'): 2,
 | |
|     ('subsubsection', 'chapter'): 3,
 | |
|     }
 | |
| 
 | |
| INCLUDED_LEVELS = ("chapter", "section", "subsection", "subsubsection")
 | |
| 
 | |
| 
 | |
| def parse_toc(fp, bigpart=None):
 | |
|     toc = top = []
 | |
|     stack = [toc]
 | |
|     level = bigpart or 'chapter'
 | |
|     lineno = 0
 | |
|     while 1:
 | |
|         line = fp.readline()
 | |
|         if not line:
 | |
|             break
 | |
|         lineno = lineno + 1
 | |
|         m = cline_rx.match(line)
 | |
|         if m:
 | |
|             stype, snum, title, pageno = m.group(1, 2, 3, 4)
 | |
|             title = clean_title(title)
 | |
|             entry = (stype, snum, title, int(pageno), [])
 | |
|             if stype == level:
 | |
|                 toc.append(entry)
 | |
|             else:
 | |
|                 if stype not in INCLUDED_LEVELS:
 | |
|                     # we don't want paragraphs & subparagraphs
 | |
|                     continue
 | |
|                 direction = _transition_map[(level, stype)]
 | |
|                 if direction == OUTER_TO_INNER:
 | |
|                     toc = toc[-1][-1]
 | |
|                     stack.insert(0, toc)
 | |
|                     toc.append(entry)
 | |
|                 else:
 | |
|                     for i in range(direction):
 | |
|                         del stack[0]
 | |
|                         toc = stack[0]
 | |
|                     toc.append(entry)
 | |
|                 level = stype
 | |
|         else:
 | |
|             sys.stderr.write("l.%s: " + line)
 | |
|     return top
 | |
| 
 | |
| 
 | |
| hackscore_rx = re.compile(r"\\hackscore\s*{[^}]*}")
 | |
| raisebox_rx = re.compile(r"\\raisebox\s*{[^}]*}")
 | |
| title_rx = re.compile(r"\\([a-zA-Z])+\s+")
 | |
| title_trans = string.maketrans("", "")
 | |
| 
 | |
| def clean_title(title):
 | |
|     title = raisebox_rx.sub("", title)
 | |
|     title = hackscore_rx.sub(r"\\_", title)
 | |
|     pos = 0
 | |
|     while 1:
 | |
|         m = title_rx.search(title, pos)
 | |
|         if m:
 | |
|             start = m.start()
 | |
|             if title[start:start+15] != "\\textunderscore":
 | |
|                 title = title[:start] + title[m.end():]
 | |
|             pos = start + 1
 | |
|         else:
 | |
|             break
 | |
|     title = title.translate(title_trans, "{}")
 | |
|     return title
 | |
| 
 | |
| 
 | |
| def write_toc(toc, fp):
 | |
|     for entry in toc:
 | |
|         write_toc_entry(entry, fp, 0)
 | |
| 
 | |
| def write_toc_entry(entry, fp, layer):
 | |
|     stype, snum, title, pageno, toc = entry
 | |
|     s = "\\pdfoutline goto name{page%03d}" % pageno
 | |
|     if toc:
 | |
|         s = "%s count -%d" % (s, len(toc))
 | |
|     if snum:
 | |
|         title = "%s %s" % (snum, title)
 | |
|     s = "%s {%s}\n" % (s, title)
 | |
|     fp.write(s)
 | |
|     for entry in toc:
 | |
|         write_toc_entry(entry, fp, layer + 1)
 | |
| 
 | |
| 
 | |
| def process(ifn, ofn, bigpart=None):
 | |
|     toc = parse_toc(open(ifn), bigpart)
 | |
|     write_toc(toc, open(ofn, "w"))
 | |
| 
 | |
| 
 | |
| def main():
 | |
|     bigpart = None
 | |
|     opts, args = getopt.getopt(sys.argv[1:], "c:")
 | |
|     if opts:
 | |
|         bigpart = opts[0][1]
 | |
|     if not args:
 | |
|         usage()
 | |
|         sys.exit(2)
 | |
|     for filename in args:
 | |
|         base, ext = os.path.splitext(filename)
 | |
|         ext = ext or ".toc"
 | |
|         process(base + ext, base + ".bkm", bigpart)
 | |
| 
 | |
| 
 | |
| if __name__ == "__main__":
 | |
|     main()
 | 
