Add deeper XML checking to update-translation script

- Catch problems such as mismatched formatting characters. Remove
  messages that can give problems at runtime.

- Also remove unfinished/untranslated messages, they just take up space
  in the ts and waste parsing time.

Fixes #4774.
This commit is contained in:
Wladimir J. van der Laan 2014-08-28 13:09:19 +02:00
parent 93f97aab62
commit da59f28335

View File

@ -14,13 +14,14 @@ It will do the following automatically:
TODO:
- auto-add new translations to the build system according to the translation process
- remove 'unfinished' translation items
'''
from __future__ import division, print_function
import subprocess
import re
import sys
import os
import io
import xml.etree.ElementTree as ET
# Name of transifex tool
TX = 'tx'
@ -40,24 +41,143 @@ def fetch_all_translations():
print('Error while fetching translations', file=sys.stderr)
exit(1)
def postprocess_translations():
print('Postprocessing...')
def find_format_specifiers(s):
'''Find all format specifiers in a string.'''
pos = 0
specifiers = []
while True:
percent = s.find('%', pos)
if percent < 0:
break
specifiers.append(s[percent+1])
pos = percent+2
return specifiers
def split_format_specifiers(specifiers):
'''Split format specifiers between numeric (Qt) and others (strprintf)'''
numeric = []
other = []
for s in specifiers:
if s in {'1','2','3','4','5','6','7','8','9'}:
numeric.append(s)
else:
other.append(s)
# numeric (Qt) can be present in any order, others (strprintf) must be in specified order
return set(numeric),other
def sanitize_string(s):
'''Sanitize string for printing'''
return s.replace('\n',' ')
def check_format_specifiers(source, translation, errors):
source_f = split_format_specifiers(find_format_specifiers(source))
# assert that no source messages contain both Qt and strprintf format specifiers
# if this fails, go change the source as this is hacky and confusing!
assert(not(source_f[0] and source_f[1]))
try:
translation_f = split_format_specifiers(find_format_specifiers(translation))
except IndexError:
errors.append("Parse error in translation '%s'" % sanitize_string(translation))
return False
else:
if source_f != translation_f:
errors.append("Mismatch between '%s' and '%s'" % (sanitize_string(source), sanitize_string(translation)))
return False
return True
def all_ts_files(suffix=''):
for filename in os.listdir(LOCALE_DIR):
# process only language files, and do not process source language
if not filename.endswith('.ts') or filename == SOURCE_LANG:
if not filename.endswith('.ts'+suffix) or filename == SOURCE_LANG+suffix:
continue
if suffix: # remove provided suffix
filename = filename[0:-len(suffix)]
filepath = os.path.join(LOCALE_DIR, filename)
with open(filepath, 'rb') as f:
yield(filename, filepath)
FIX_RE = re.compile(b'[\x00-\x09\x0b\x0c\x0e-\x1f]')
def remove_invalid_characters(s):
'''Remove invalid characters from translation string'''
return FIX_RE.sub(b'', s)
# Override cdata escape function to make our output match Qt's (optional, just for cleaner diffs for
# comparison, disable by default)
_orig_escape_cdata = None
def escape_cdata(text):
text = _orig_escape_cdata(text)
text = text.replace("'", '&apos;')
text = text.replace('"', '&quot;')
return text
def postprocess_translations(reduce_diff_hacks=False):
print('Checking and postprocessing...')
if reduce_diff_hacks:
global _orig_escape_cdata
_orig_escape_cdata = ET._escape_cdata
ET._escape_cdata = escape_cdata
for (filename,filepath) in all_ts_files():
os.rename(filepath, filepath+'.orig')
have_errors = False
for (filename,filepath) in all_ts_files('.orig'):
# pre-fixups to cope with transifex output
parser = ET.XMLParser(encoding='utf-8') # need to override encoding because 'utf8' is not understood only 'utf-8'
with open(filepath + '.orig', 'rb') as f:
data = f.read()
# remove non-allowed control characters
data = re.sub('[\x00-\x09\x0b\x0c\x0e-\x1f]', '', data)
data = data.split('\n')
# strip locations from non-origin translation
# location tags are used to guide translators, they are not necessary for compilation
# TODO: actually process XML instead of relying on Transifex's one-tag-per-line output format
data = [line for line in data if not '<location' in line]
with open(filepath, 'wb') as f:
f.write('\n'.join(data))
# remove control characters; this must be done over the entire file otherwise the XML parser will fail
data = remove_invalid_characters(data)
tree = ET.parse(io.BytesIO(data), parser=parser)
# iterate over all messages in file
root = tree.getroot()
for context in root.findall('context'):
for message in context.findall('message'):
numerus = message.get('numerus') == 'yes'
source = message.find('source').text
translation_node = message.find('translation')
# pick all numerusforms
if numerus:
translations = [i.text for i in translation_node.findall('numerusform')]
else:
translations = [translation_node.text]
for translation in translations:
if translation is None:
continue
errors = []
valid = check_format_specifiers(source, translation, errors)
for error in errors:
print('%s: %s' % (filename, error))
if not valid: # set type to unfinished and clear string if invalid
translation_node.clear()
translation_node.set('type', 'unfinished')
have_errors = True
# Remove location tags
for location in message.findall('location'):
message.remove(location)
# Remove entire message if it is an unfinished translation
if translation_node.get('type') == 'unfinished':
context.remove(message)
# write fixed-up tree
# if diff reduction requested, replace some XML to 'sanitize' to qt formatting
if reduce_diff_hacks:
out = io.BytesIO()
tree.write(out, encoding='utf-8')
out = out.getvalue()
out = out.replace(b' />', b'/>')
with open(filepath, 'wb') as f:
f.write(out)
else:
tree.write(filepath, encoding='utf-8')
return have_errors
if __name__ == '__main__':
check_at_repository_root()