Remove TeX control sequences from TeX source (for LanguageTool)

remove-tex-contro.py

#!/usr/bin/env python

__author__ = 'Y.OKAMURA <okamura=AT=informationsea.info>'
__copyright__ = 'Copyright (C) 2013 Y.OKAMURA'
__license__ = 'GPL3+'

import argparse
import re
import sys

def _main():
    parser = argparse.ArgumentParser(description="Remove TeX control sequences from TeX source")
    parser.add_argument('input', type=argparse.FileType('r'), default=sys.stdin, nargs='?')
    parser.add_argument('output', type=argparse.FileType('w'), default=sys.stdout, nargs='?')
    options = parser.parse_args()

    sections = re.compile(r'\\((sub)*section\*?|caption)\{([^\}]+)\}')
    removes = re.compile(r'(\\[^\{\}\s\\]+\s|%.+|\s[\{\}]\s|\\[^\{]+([\{\[][^\{\}]+[\}\]])+|\$[^\$]+\$|[\{\}])')

    begin = re.compile(r'\\(begin|end)\{([^\}]+)\}')
    dontremove = ['center', 'itemize', 'document']

    beginstack = list()

    for line in options.input:
        line = unicode(line, 'ISO-2022-JP')
        pos = 0
        while begin.search(line, pos):
            m = begin.search(line, pos)
            if m.group(1) == 'begin':
                beginstack.append(m.group(2))
            else:
                beginstack.pop()
            pos = m.end()
            
        if any([(x not in dontremove) for x in beginstack]):
            options.output.write('\n')
            continue
        
        pos = 0
        while sections.search(line, pos):
            m = sections.search(line, pos)
            line = line[:m.start()] + u' '*(m.start(3)-m.start()) + m.group(3) + u' '*(m.end(3)-m.end()) + line[m.end():]
            pos = m.end()-1

        pos = 0
        while removes.search(line, pos):
            m = removes.search(line, pos)
            line = line[:m.start()] + u' '*(m.end() - m.start()) + line[m.end():]
            pos = m.end()-1
            
        options.output.write(line.encode('utf-8'))
        
#        if begin.search(s):
#            print 'BEGIN: '+begin.search(s).group(1)
    
    
    pass
    
    
if __name__ == '__main__':
    _main()