Mercurial > hg
view i18n/check-translation.py @ 21859:be4270d27a7e
progress: use 'encoding.trim' to trim output line correctly
Before this patch, 'progress' extension trims output line by directly
slicing byte sequence, but it may split at intermediate multi-byte
sequence.
This patch uses 'encoding.trim' to trim output line correctly, even if
it contains multi-byte characters.
"rm -f loop.pyc" before changing "loop.py" in "test-progress.t"
ensures that re-compilation of "loop.py", even if "loop.py" and
"loop.pyc" have same timestamp in seconds.
author | FUJIWARA Katsunori <foozy@lares.dti.ne.jp> |
---|---|
date | Sun, 06 Jul 2014 02:56:41 +0900 |
parents | 6afbfb9b1af1 |
children | 35c2ea4ca26f |
line wrap: on
line source
#!/usr/bin/env python # # check-translation.py - check Mercurial specific translation problems import polib import re checkers = [] def checker(level, msgidpat): def decorator(func): if msgidpat: match = re.compile(msgidpat).search else: match = lambda msgid: True checkers.append((func, level)) func.match = match return func return decorator def match(checker, pe): """Examine whether POEntry "pe" is target of specified checker or not """ if not checker.match(pe.msgid): return # examine suppression by translator comment nochecker = 'no-%s-check' % checker.__name__ for tc in pe.tcomment.split(): if nochecker == tc: return return True #################### def fatalchecker(msgidpat=None): return checker('fatal', msgidpat) @fatalchecker(r'\$\$') def promptchoice(pe): """Check translation of the string given to "ui.promptchoice()" >>> pe = polib.POEntry( ... msgid ='prompt$$missing &sep$$missing &$$followed by &none', ... msgstr='prompt missing &sep$$missing amp$$followed by none&') >>> match(promptchoice, pe) True >>> for e in promptchoice(pe): print e number of choices differs between msgid and msgstr msgstr has invalid choice missing '&' msgstr has invalid '&' followed by none """ idchoices = [c.rstrip(' ') for c in pe.msgid.split('$$')[1:]] strchoices = [c.rstrip(' ') for c in pe.msgstr.split('$$')[1:]] if len(idchoices) != len(strchoices): yield "number of choices differs between msgid and msgstr" indices = [(c, c.find('&')) for c in strchoices] if [c for c, i in indices if i == -1]: yield "msgstr has invalid choice missing '&'" if [c for c, i in indices if len(c) == i + 1]: yield "msgstr has invalid '&' followed by none" #################### def warningchecker(msgidpat=None): return checker('warning', msgidpat) @warningchecker() def taildoublecolons(pe): """Check equality of tail '::'-ness between msgid and msgstr >>> pe = polib.POEntry( ... msgid ='ends with ::', ... msgstr='ends with ::') >>> for e in taildoublecolons(pe): print e >>> pe = polib.POEntry( ... msgid ='ends with ::', ... msgstr='ends without double-colons') >>> for e in taildoublecolons(pe): print e tail '::'-ness differs between msgid and msgstr >>> pe = polib.POEntry( ... msgid ='ends without double-colons', ... msgstr='ends with ::') >>> for e in taildoublecolons(pe): print e tail '::'-ness differs between msgid and msgstr """ if pe.msgid.endswith('::') != pe.msgstr.endswith('::'): yield "tail '::'-ness differs between msgid and msgstr" @warningchecker() def indentation(pe): """Check equality of initial indentation between msgid and msgstr This may report unexpected warning, because this doesn't aware the syntax of rst document and the context of msgstr. >>> pe = polib.POEntry( ... msgid =' indented text', ... msgstr=' narrowed indentation') >>> for e in indentation(pe): print e initial indentation width differs betweeen msgid and msgstr """ idindent = len(pe.msgid) - len(pe.msgid.lstrip()) strindent = len(pe.msgstr) - len(pe.msgstr.lstrip()) if idindent != strindent: yield "initial indentation width differs betweeen msgid and msgstr" #################### def check(pofile, fatal=True, warning=False): targetlevel = { 'fatal': fatal, 'warning': warning } targetcheckers = [(checker, level) for checker, level in checkers if targetlevel[level]] if not targetcheckers: return [] detected = [] for pe in pofile.translated_entries(): errors = [] for checker, level in targetcheckers: if match(checker, pe): errors.extend((level, checker.__name__, error) for error in checker(pe)) if errors: detected.append((pe, errors)) return detected ######################################## if __name__ == "__main__": import sys import optparse optparser = optparse.OptionParser("""%prog [options] pofile ... This checks Mercurial specific translation problems in specified '*.po' files. Each detected problems are shown in the format below:: filename:linenum:type(checker): problem detail ..... "type" is "fatal" or "warning". "checker" is the name of the function detecting corresponded error. Checking by checker "foo" on the specific msgstr can be suppressed by the "translator comment" like below. Multiple "no-xxxx-check" should be separated by whitespaces:: # no-foo-check msgid = "....." msgstr = "....." """) optparser.add_option("", "--warning", help="show also warning level problems", action="store_true") optparser.add_option("", "--doctest", help="run doctest of this tool, instead of check", action="store_true") (options, args) = optparser.parse_args() if options.doctest: import os if 'TERM' in os.environ: del os.environ['TERM'] import doctest failures, tests = doctest.testmod() sys.exit(failures and 1 or 0) # replace polib._POFileParser to show linenum of problematic msgstr class ExtPOFileParser(polib._POFileParser): def process(self, symbol, linenum): super(ExtPOFileParser, self).process(symbol, linenum) if symbol == 'MS': # msgstr self.current_entry.linenum = linenum polib._POFileParser = ExtPOFileParser detected = [] warning = options.warning for f in args: detected.extend((f, pe, errors) for pe, errors in check(polib.pofile(f), warning=warning)) if detected: for f, pe, errors in detected: for level, checker, error in errors: sys.stderr.write('%s:%d:%s(%s): %s\n' % (f, pe.linenum, level, checker, error)) sys.exit(1)