gramps/po/update_po.py
Nick Hall 2319e04d95 Add header to XML fragments file
A header containing a charset is required by xgettext.
2023-07-26 17:32:16 +01:00

511 lines
18 KiB
Python

#! /usr/bin/env python3
#
# update_po - a gramps tool to update translations
#
# Copyright (C) 2006-2006 Kees Bakker
# Copyright (C) 2006 Brian Matherly
# Copyright (C) 2008 Stephen George
# Copyright (C) 2012
# Copyright (C) 2020 Nick Hall
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
#
"""
update_po.py for Gramps translations.
Examples:
python update_po.py -t
Tests if 'gettext' and 'python' are well configured.
python update_po.py -h
Calls help and command line interface.
python update_po.py -p
Generates a new template/catalog (gramps.pot).
python update_po.py -m de.po
Merges 'de.po' file with 'gramps.pot'.
python update_po.py -k de.po
Checks 'de.po' file, tests to compile and generates a textual resume.
"""
from __future__ import print_function
import os
import sys
import shutil
from argparse import ArgumentParser
from tokenize import tokenize, STRING, COMMENT, NL, TokenError
# Windows OS
if sys.platform in ['linux', 'linux2', 'darwin', 'cygwin'] or shutil.which('msgmerge'):
msgmergeCmd = 'msgmerge'
msgfmtCmd = 'msgfmt'
msgattribCmd = 'msgattrib'
xgettextCmd = 'xgettext'
pythonCmd = os.path.join(sys.prefix, 'bin', 'python3')
elif sys.platform == 'win32':
# GetText Win 32 obtained from http://gnuwin32.sourceforge.net/packages/gettext.htm
# ....\gettext\bin\msgmerge.exe needs to be on the path
msgmergeCmd = os.path.join('C:', 'Program Files(x86)', 'gettext', 'bin', 'msgmerge.exe')
msgfmtCmd = os.path.join('C:', 'Program Files(x86)', 'gettext', 'bin', 'msgfmt.exe')
msgattribCmd = os.path.join('C:', 'Program Files(x86)', 'gettext', 'bin', 'msgattrib.exe')
xgettextCmd = os.path.join('C:', 'Program Files(x86)', 'gettext', 'bin', 'xgettext.exe')
pythonCmd = os.path.join(sys.prefix, 'bin', 'python.exe')
# Others OS
elif sys.platform in ['linux', 'linux2', 'darwin', 'cygwin']:
msgmergeCmd = 'msgmerge'
msgfmtCmd = 'msgfmt'
msgattribCmd = 'msgattrib'
xgettextCmd = 'xgettext'
pythonCmd = os.path.join(sys.prefix, 'bin', 'python3')
else:
print("Found platform %s, OS %s" % (sys.platform, os.name))
print ("Update PO ERROR: unknown system, don't know msgmerge, ... commands")
sys.exit(0)
# List of available languages, useful for grouped actions
# need files with po extension
LANG = [file for file in os.listdir('.') if file.endswith('.po')]
# add a special 'all' argument (for 'check' and 'merge' arguments)
LANG.append("all")
# visual polish on the languages list
LANG.sort()
def tests():
"""
Testing installed programs.
We made tests (-t flag) by displaying versions of tools if properly
installed. Cannot run all commands without 'gettext' and 'python'.
"""
try:
print ("\n====='msgmerge'=(merge our translation)================\n")
os.system('''%(program)s -V''' % {'program': msgmergeCmd})
except:
print ('Please, install %(program)s for updating your translation'
% {'program': msgmergeCmd})
try:
print ("\n==='msgfmt'=(format our translation for installation)==\n")
os.system('''%(program)s -V''' % {'program': msgfmtCmd})
except:
print ('Please, install %(program)s for checking your translation'
% {'program': msgfmtCmd})
try:
print ("\n===='msgattrib'==(list groups of messages)=============\n")
os.system('''%(program)s -V''' % {'program': msgattribCmd})
except:
print ('Please, install %(program)s for listing groups of messages'
% {'program': msgattribCmd})
try:
print("\n===='xgettext' =(generate a new template)==============\n")
os.system('''%(program)s -V''' % {'program': xgettextCmd})
except:
print ('Please, install %(program)s for generating a new template'
% {'program': xgettextCmd})
try:
print("\n=================='python'=============================\n")
os.system('''%(program)s -V''' % {'program': pythonCmd})
except:
print ('Please, install python')
def main():
"""
The utility for handling translation stuff.
What is need by Gramps, nothing more.
"""
parser = ArgumentParser(
description='This program generates a new template and '
'also provides some common features.',
)
parser.add_argument("-t", "--test",
action="store_true", dest="test", default=True,
help="test if 'python' and 'gettext' are properly installed")
parser.add_argument("-x", "--xml",
action="store_true", dest="xml", default=False,
help="extract messages from xml based file formats")
parser.add_argument("-g", "--glade",
action="store_true", dest="glade", default=False,
help="extract messages from glade file format only")
parser.add_argument("-c", "--clean",
action="store_true", dest="clean", default=False,
help="remove created files")
parser.add_argument("-p", "--pot",
action="store_true", dest="catalog", default=False,
help="create a new catalog")
update = parser.add_argument_group('Update', 'Maintenance around translations')
# need at least one argument (sv.po, de.po, etc ...)
# lang.po files maintenance
update.add_argument("-m", dest="merge",
choices=LANG,
help="merge lang.po files with last catalog")
update.add_argument("-k", dest="check",
choices=LANG,
help="check lang.po files")
# testing stage
trans = parser.add_argument_group('Translation', 'Display content of translations file')
# need one argument (eg, de.po)
trans.add_argument("-u", dest="untranslated",
choices=[file for file in os.listdir('.') if file.endswith('.po')],
help="list untranslated messages")
trans.add_argument("-f", dest="fuzzy",
choices=[file for file in os.listdir('.') if file.endswith('.po')],
help="list fuzzy messages")
args = parser.parse_args()
namespace, extra = parser.parse_known_args()
if args.test:
tests()
if args.xml:
extract_xml()
if args.glade:
create_filesfile()
extract_glade()
if os.path.isfile('tmpfiles'):
os.unlink('tmpfiles')
if args.catalog:
retrieve()
if args.clean:
clean()
if args.merge:
#retrieve() windows os?
if sys.argv[2:] == ['all']:
sys.argv[2:] = LANG
merge(sys.argv[2:])
if args.check:
#retrieve() windows os?
if sys.argv[2:] == ['all']:
sys.argv[2:] = LANG
check(sys.argv[2:])
if args.untranslated:
untranslated(sys.argv[2:])
if args.fuzzy:
fuzzy(sys.argv[2:])
def create_filesfile():
"""
Create a file with all files that we should translate.
These are all python files not in POTFILES.skip added with those in
POTFILES.in
"""
dir = os.getcwd()
topdir = os.path.normpath(os.path.join(dir, '..', 'gramps'))
lentopdir = len(topdir)
with open('POTFILES.in') as f:
infiles = dict(['../' + file.strip(), None] for file in f if file.strip()
and not file[0]=='#')
with open('POTFILES.skip') as f:
notinfiles = dict(['../' + file.strip(), None] for file in f if file
and not file[0]=='#')
for (dirpath, dirnames, filenames) in os.walk(topdir):
root, subdir = os.path.split(dirpath)
if subdir.startswith("."):
#don't continue in this dir
dirnames[:] = []
continue
for dirname in dirnames:
# Skip hidden and system directories:
if dirname.startswith(".") or dirname in ["po", "locale"]:
dirnames.remove(dirname)
#add the files which are python or glade files
# if the directory does not exist or is a link, do nothing
if not os.path.isdir(dirpath) or os.path.islink(dirpath):
continue
for filename in os.listdir(dirpath):
name = os.path.split(filename)[1]
if name.endswith('.py') or name.endswith('.glade'):
full_filename = os.path.join(dirpath, filename)
#Skip the file if in POTFILES.skip
if full_filename[lentopdir:] in notinfiles:
infiles['../gramps' + full_filename[lentopdir:]] = None
#now we write out all the files in form ../gramps/filename
with open('tmpfiles', 'w') as f:
for file in sorted(infiles.keys()):
f.write(file)
f.write('\n')
def listing(name, extensionlist):
"""
List files according to extensions.
Parsing from a textual file (gramps) is faster and easy for maintenance.
Like POTFILES.in and POTFILES.skip
"""
with open('tmpfiles') as f:
files = [file.strip() for file in f if file and not file[0]=='#']
with open(name, 'w') as temp:
for entry in files:
for ext in extensionlist:
if entry.endswith(ext):
temp.write(entry)
temp.write('\n')
break
def headers():
"""
Look at existing C file format headers.
Generated by 'intltool-extract' but want to get rid of this
dependency (perl, just a set of tools).
"""
headers = []
# in.h; extract_xml
if os.path.isfile('''fragments.pot'''):
headers.append('''fragments.pot''')
return headers
def extract_xml():
"""
Extract translation strings from XML based, mime and desktop files.
Uses custom ITS rules found in the po/its directory.
"""
if not os.path.isfile('gramps.pot'):
create_template()
for input_file in ['../data/holidays.xml',
'../data/tips.xml',
'../data/org.gramps_project.Gramps.xml.in',
'../data/org.gramps_project.Gramps.appdata.xml.in',
'../data/org.gramps_project.Gramps.desktop.in']:
os.system(('GETTEXTDATADIR=. %(xgettext)s -F -j '
'-o gramps.pot --from-code=UTF-8 %(inputfile)s')
% {'xgettext': xgettextCmd,
'inputfile': input_file}
)
print (input_file)
def create_template():
"""
Create a new file for template, if it does not exist.
"""
with open('gramps.pot', 'w') as template:
pass
def extract_glade():
"""
Extract messages from a temp file with all .glade
"""
if not os.path.isfile('gramps.pot'):
create_template()
listing('glade.txt', ['.glade'])
os.system('''%(xgettext)s -F --add-comments -j -L Glade '''
'''--from-code=UTF-8 -o gramps.pot --files-from=glade.txt'''
% {'xgettext': xgettextCmd}
)
def xml_fragments():
""" search through the file for xml fragments that contain the
'translate="yes">string<' pattern. These need to be added to the message
catalog """
with open('tmpfiles') as __f:
files = [file.strip() for file in __f if
file and not (file[0] == '#') and file.endswith('.py\n')]
print("Checking for XML fragments in Python files")
modop = int(len(files) / 20)
wfp = open("fragments.pot", 'w', encoding='utf-8')
wfp.write('msgid ""\n')
wfp.write('msgstr ""\n')
wfp.write('"Content-Type: text/plain; charset=UTF-8\\n"\n\n')
for indx, filename in enumerate(files):
if not indx % modop:
print(int(indx / len(files) * 100), end='\r')
fp = open(filename, 'rb')
try:
tokens = tokenize(fp.readline)
in_string = False
for _token, _text, _start, _end, _line in tokens:
if _text.startswith('"""') or _text.startswith("'''"):
_text = _text[3:]
elif _text.startswith('"') or _text.startswith("'"):
_text = _text[1:]
if _text.endswith('"""') or _text.endswith("'''"):
_text = _text[:-3]
elif _text.endswith('"') or _text.endswith("'"):
_text = _text[:-1]
if _token == STRING and not in_string:
in_string = True
line_no = _start[0]
text = _text
continue
elif _token == STRING and in_string:
text += _text
continue
elif _token == COMMENT or _token == NL and in_string:
# need to ignore comments and concatinate strings
_ml = True
continue
elif in_string:
in_string = False
end = 0
# _find_message_in_xml(text)
while True:
fnd = text.find('translatable="yes">', end)
if fnd == -1:
break
end = text.find('<', fnd)
if end == -1:
print("\nBad xml fragment '%s' at %s line %d" %
(text[fnd:], filename, _start[0]))
break
msg = text[fnd + 19 : end]
if "%s" in msg or (msg.startswith('{') and
msg.endswith('}')):
print('\n#: %s:%d Are you sure you want to '
'translate the "%%s"???' %
(filename, line_no))
break
wfp.write('#: %s:%d\nmsgid "%s"\nmsgstr ""\n' %
(filename, line_no, msg))
except TokenError as e:
print('\n%s: %s, line %d, column %d' % (
e.args[0], filename, e.args[1][0], e.args[1][1]),
file=sys.stderr)
finally:
fp.close()
wfp.close()
def retrieve():
"""
Extract messages from all files used by Gramps (python, glade, xml)
"""
create_template()
create_filesfile()
xml_fragments()
listing('python.txt', ['.py', '.py.in'])
# additional keywords must always be kept in sync with those in genpot.sh
os.system('''%(xgettext)s -F --add-comments=Translators -j '''
'''--directory=./ -d gramps -L Python '''
'''-o gramps.pot --files-from=python.txt '''
'''--debug --keyword=_ --keyword=ngettext '''
'''--keyword=_T_ --keyword=trans_text:1,2c '''
'''--keyword=_:1,2c --keyword=_T_:1,2c '''
'''--keyword=sgettext --from-code=UTF-8''' % {'xgettext': xgettextCmd}
)
extract_glade()
extract_xml()
# C format header (.h extension)
for h in headers():
print ('xgettext for %s' % h)
os.system('''%(xgettext)s -F --add-comments=Translators -j '''
'''-o gramps.pot --keyword=N_ --from-code=UTF-8 %(head)s'''
% {'xgettext': xgettextCmd, 'head': h}
)
clean()
def clean():
"""
Remove created files (C format headers, temp listings)
"""
for h in headers():
if os.path.isfile(h):
os.unlink(h)
print ('Remove %(head)s' % {'head': h})
if os.path.isfile('python.txt'):
os.unlink('python.txt')
print ("Remove 'python.txt'")
if os.path.isfile('glade.txt'):
os.unlink('glade.txt')
print ("Remove 'glade.txt'")
if os.path.isfile('tmpfiles'):
os.unlink('tmpfiles')
print ("Remove 'tmpfiles'")
def merge(args):
"""
Merge messages with 'gramps.pot'
"""
for arg in args:
if arg == 'all':
continue
print ('Merge %(lang)s with current template' % {'lang': arg})
os.system('''%(msgmerge)s -U %(lang)s gramps.pot''' \
% {'msgmerge': msgmergeCmd, 'lang': arg})
print ("Updated file: '%(lang)s'." % {'lang': arg})
def check(args):
"""
Check the translation file
"""
for arg in args:
if arg == 'all':
continue
print ("Checked file: '%(lang.po)s'. See '%(txt)s.txt'." \
% {'lang.po': arg, 'txt': arg[:-3]})
os.system('''%(python)s ./check_po -s %(lang.po)s > %(lang)s.txt''' \
% {'python': pythonCmd, 'lang.po': arg, 'lang': arg[:-3]})
os.system('''%(msgfmt)s -c -v %(lang.po)s'''
% {'msgfmt': msgfmtCmd, 'lang.po': arg})
def untranslated(arg):
"""
List untranslated messages
"""
os.system('''%(msgattrib)s --untranslated %(lang.po)s''' % {'msgattrib': msgattribCmd, 'lang.po': arg[0]})
def fuzzy(arg):
"""
List fuzzy messages
"""
os.system('''%(msgattrib)s --only-fuzzy --no-obsolete %(lang.po)s''' % {'msgattrib': msgattribCmd, 'lang.po': arg[0]})
if __name__ == "__main__":
main()