04277e02d7
* All files with FSF copyright notices: Update copyright dates using scripts/update-copyrights. * locale/programs/charmap-kw.h: Regenerated. * locale/programs/locfile-kw.h: Likewise.
365 lines
14 KiB
Python
Executable File
365 lines
14 KiB
Python
Executable File
#!/usr/bin/python3
|
|
# -*- coding: utf-8 -*-
|
|
# Copyright (C) 2014-2019 Free Software Foundation, Inc.
|
|
# This file is part of the GNU C Library.
|
|
#
|
|
# The GNU C Library is free software; you can redistribute it and/or
|
|
# modify it under the terms of the GNU Lesser General Public
|
|
# License as published by the Free Software Foundation; either
|
|
# version 2.1 of the License, or (at your option) any later version.
|
|
#
|
|
# The GNU C Library 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
|
|
# Lesser General Public License for more details.
|
|
#
|
|
# You should have received a copy of the GNU Lesser General Public
|
|
# License along with the GNU C Library; if not, see
|
|
# <http://www.gnu.org/licenses/>.
|
|
|
|
'''glibc/localedata/charmaps/UTF-8 file generator script
|
|
|
|
This script generates a glibc/localedata/charmaps/UTF-8 file
|
|
from Unicode data.
|
|
|
|
Usage: python3 utf8_gen.py UnicodeData.txt EastAsianWidth.txt
|
|
|
|
It will output UTF-8 file
|
|
'''
|
|
|
|
import argparse
|
|
import sys
|
|
import re
|
|
import unicode_utils
|
|
|
|
# Auxiliary tables for Hangul syllable names, see the Unicode 3.0 book,
|
|
# sections 3.11 and 4.4.
|
|
|
|
JAMO_INITIAL_SHORT_NAME = (
|
|
'G', 'GG', 'N', 'D', 'DD', 'R', 'M', 'B', 'BB', 'S', 'SS', '', 'J', 'JJ',
|
|
'C', 'K', 'T', 'P', 'H'
|
|
)
|
|
|
|
JAMO_MEDIAL_SHORT_NAME = (
|
|
'A', 'AE', 'YA', 'YAE', 'EO', 'E', 'YEO', 'YE', 'O', 'WA', 'WAE', 'OE',
|
|
'YO', 'U', 'WEO', 'WE', 'WI', 'YU', 'EU', 'YI', 'I'
|
|
)
|
|
|
|
JAMO_FINAL_SHORT_NAME = (
|
|
'', 'G', 'GG', 'GS', 'N', 'NI', 'NH', 'D', 'L', 'LG', 'LM', 'LB', 'LS',
|
|
'LT', 'LP', 'LH', 'M', 'B', 'BS', 'S', 'SS', 'NG', 'J', 'C', 'K', 'T',
|
|
'P', 'H'
|
|
)
|
|
|
|
def process_range(start, end, outfile, name):
|
|
'''Writes a range of code points into the CHARMAP section of the
|
|
output file
|
|
|
|
'''
|
|
if 'Hangul Syllable' in name:
|
|
# from glibc/localedata/ChangeLog:
|
|
#
|
|
# 2000-09-24 Bruno Haible <haible@clisp.cons.org>
|
|
# * charmaps/UTF-8: Expand <Hangul Syllable> and <Private Use> ranges,
|
|
# so they become printable and carry a width. Comment out surrogate
|
|
# ranges. Add a WIDTH table
|
|
#
|
|
# So we expand the Hangul Syllables here:
|
|
for i in range(int(start, 16), int(end, 16)+1 ):
|
|
index2, index3 = divmod(i - 0xaC00, 28)
|
|
index1, index2 = divmod(index2, 21)
|
|
hangul_syllable_name = 'HANGUL SYLLABLE ' \
|
|
+ JAMO_INITIAL_SHORT_NAME[index1] \
|
|
+ JAMO_MEDIAL_SHORT_NAME[index2] \
|
|
+ JAMO_FINAL_SHORT_NAME[index3]
|
|
outfile.write('{:<11s} {:<12s} {:s}\n'.format(
|
|
unicode_utils.ucs_symbol(i), convert_to_hex(i),
|
|
hangul_syllable_name))
|
|
return
|
|
# UnicodeData.txt file has contains code point ranges like this:
|
|
#
|
|
# 3400;<CJK Ideograph Extension A, First>;Lo;0;L;;;;;N;;;;;
|
|
# 4DB5;<CJK Ideograph Extension A, Last>;Lo;0;L;;;;;N;;;;;
|
|
#
|
|
# The glibc UTF-8 file splits ranges like these into shorter
|
|
# ranges of 64 code points each:
|
|
#
|
|
# <U3400>..<U343F> /xe3/x90/x80 <CJK Ideograph Extension A>
|
|
# …
|
|
# <U4D80>..<U4DB5> /xe4/xb6/x80 <CJK Ideograph Extension A>
|
|
for i in range(int(start, 16), int(end, 16), 64 ):
|
|
if i > (int(end, 16)-64):
|
|
outfile.write('{:s}..{:s} {:<12s} {:s}\n'.format(
|
|
unicode_utils.ucs_symbol(i),
|
|
unicode_utils.ucs_symbol(int(end,16)),
|
|
convert_to_hex(i),
|
|
name))
|
|
break
|
|
outfile.write('{:s}..{:s} {:<12s} {:s}\n'.format(
|
|
unicode_utils.ucs_symbol(i),
|
|
unicode_utils.ucs_symbol(i+63),
|
|
convert_to_hex(i),
|
|
name))
|
|
|
|
def process_charmap(flines, outfile):
|
|
'''This function takes an array which contains *all* lines of
|
|
of UnicodeData.txt and write lines to outfile as used in the
|
|
|
|
CHARMAP
|
|
…
|
|
END CHARMAP
|
|
|
|
section of the UTF-8 file in glibc/localedata/charmaps/UTF-8.
|
|
|
|
Samples for input lines:
|
|
|
|
0010;<control>;Cc;0;BN;;;;;N;DATA LINK ESCAPE;;;;
|
|
3400;<CJK Ideograph Extension A, First>;Lo;0;L;;;;;N;;;;;
|
|
4DB5;<CJK Ideograph Extension A, Last>;Lo;0;L;;;;;N;;;;;
|
|
D800;<Non Private Use High Surrogate, First>;Cs;0;L;;;;;N;;;;;
|
|
DB7F;<Non Private Use High Surrogate, Last>;Cs;0;L;;;;;N;;;;;
|
|
100000;<Plane 16 Private Use, First>;Co;0;L;;;;;N;;;;;
|
|
10FFFD;<Plane 16 Private Use, Last>;Co;0;L;;;;;N;;;;;
|
|
|
|
Samples for output lines (Unicode-Value UTF-8-HEX Unicode-Char-Name):
|
|
|
|
<U0010> /x10 DATA LINK ESCAPE
|
|
<U3400>..<U343F> /xe3/x90/x80 <CJK Ideograph Extension A>
|
|
%<UD800> /xed/xa0/x80 <Non Private Use High Surrogate, First>
|
|
%<UDB7F> /xed/xad/xbf <Non Private Use High Surrogate, Last>
|
|
<U0010FFC0>..<U0010FFFD> /xf4/x8f/xbf/x80 <Plane 16 Private Use>
|
|
|
|
'''
|
|
fields_start = []
|
|
for line in flines:
|
|
fields = line.split(";")
|
|
# Some characters have “<control>” as their name. We try to
|
|
# use the “Unicode 1.0 Name” (10th field in
|
|
# UnicodeData.txt) for them.
|
|
#
|
|
# The Characters U+0080, U+0081, U+0084 and U+0099 have
|
|
# “<control>” as their name but do not even have aa
|
|
# ”Unicode 1.0 Name”. We could write code to take their
|
|
# alternate names from NameAliases.txt.
|
|
if fields[1] == "<control>" and fields[10]:
|
|
fields[1] = fields[10]
|
|
# Handling code point ranges like:
|
|
#
|
|
# 3400;<CJK Ideograph Extension A, First>;Lo;0;L;;;;;N;;;;;
|
|
# 4DB5;<CJK Ideograph Extension A, Last>;Lo;0;L;;;;;N;;;;;
|
|
if fields[1].endswith(', First>') and not 'Surrogate,' in fields[1]:
|
|
fields_start = fields
|
|
continue
|
|
if fields[1].endswith(', Last>') and not 'Surrogate,' in fields[1]:
|
|
process_range(fields_start[0], fields[0],
|
|
outfile, fields[1][:-7]+'>')
|
|
fields_start = []
|
|
continue
|
|
fields_start = []
|
|
if 'Surrogate,' in fields[1]:
|
|
# Comment out the surrogates in the UTF-8 file.
|
|
# One could of course skip them completely but
|
|
# the original UTF-8 file in glibc had them as
|
|
# comments, so we keep these comment lines.
|
|
outfile.write('%')
|
|
outfile.write('{:<11s} {:<12s} {:s}\n'.format(
|
|
unicode_utils.ucs_symbol(int(fields[0], 16)),
|
|
convert_to_hex(int(fields[0], 16)),
|
|
fields[1]))
|
|
|
|
def convert_to_hex(code_point):
|
|
'''Converts a code point to a hexadecimal UTF-8 representation
|
|
like /x**/x**/x**.'''
|
|
# Getting UTF8 of Unicode characters.
|
|
# In Python3, .encode('UTF-8') does not work for
|
|
# surrogates. Therefore, we use this conversion table
|
|
surrogates = {
|
|
0xD800: '/xed/xa0/x80',
|
|
0xDB7F: '/xed/xad/xbf',
|
|
0xDB80: '/xed/xae/x80',
|
|
0xDBFF: '/xed/xaf/xbf',
|
|
0xDC00: '/xed/xb0/x80',
|
|
0xDFFF: '/xed/xbf/xbf',
|
|
}
|
|
if code_point in surrogates:
|
|
return surrogates[code_point]
|
|
return ''.join([
|
|
'/x{:02x}'.format(c) for c in chr(code_point).encode('UTF-8')
|
|
])
|
|
|
|
def write_header_charmap(outfile):
|
|
'''Write the header on top of the CHARMAP section to the output file'''
|
|
outfile.write("<code_set_name> UTF-8\n")
|
|
outfile.write("<comment_char> %\n")
|
|
outfile.write("<escape_char> /\n")
|
|
outfile.write("<mb_cur_min> 1\n")
|
|
outfile.write("<mb_cur_max> 6\n\n")
|
|
outfile.write("% CHARMAP generated using utf8_gen.py\n")
|
|
outfile.write("% alias ISO-10646/UTF-8\n")
|
|
outfile.write("CHARMAP\n")
|
|
|
|
def write_header_width(outfile, unicode_version):
|
|
'''Writes the header on top of the WIDTH section to the output file'''
|
|
outfile.write('% Character width according to Unicode '
|
|
+ '{:s}.\n'.format(unicode_version))
|
|
outfile.write('% - Default width is 1.\n')
|
|
outfile.write('% - Double-width characters have width 2; generated from\n')
|
|
outfile.write('% "grep \'^[^;]*;[WF]\' EastAsianWidth.txt"\n')
|
|
outfile.write('% - Non-spacing characters have width 0; '
|
|
+ 'generated from PropList.txt or\n')
|
|
outfile.write('% "grep \'^[^;]*;[^;]*;[^;]*;[^;]*;NSM;\' '
|
|
+ 'UnicodeData.txt"\n')
|
|
outfile.write('% - Format control characters have width 0; '
|
|
+ 'generated from\n')
|
|
outfile.write("% \"grep '^[^;]*;[^;]*;Cf;' UnicodeData.txt\"\n")
|
|
# Not needed covered by Cf
|
|
# outfile.write("% - Zero width characters have width 0; generated from\n")
|
|
# outfile.write("% \"grep '^[^;]*;ZERO WIDTH ' UnicodeData.txt\"\n")
|
|
outfile.write("WIDTH\n")
|
|
|
|
def process_width(outfile, ulines, elines, plines):
|
|
'''ulines are lines from UnicodeData.txt, elines are lines from
|
|
EastAsianWidth.txt containing characters with width “W” or “F”,
|
|
plines are lines from PropList.txt which contain characters
|
|
with the property “Prepended_Concatenation_Mark”.
|
|
|
|
'''
|
|
width_dict = {}
|
|
for line in elines:
|
|
fields = line.split(";")
|
|
if not '..' in fields[0]:
|
|
code_points = (fields[0], fields[0])
|
|
else:
|
|
code_points = fields[0].split("..")
|
|
for key in range(int(code_points[0], 16),
|
|
int(code_points[1], 16)+1):
|
|
width_dict[key] = 2
|
|
|
|
for line in ulines:
|
|
fields = line.split(";")
|
|
if fields[4] == "NSM" or fields[2] in ("Cf", "Me", "Mn"):
|
|
width_dict[int(fields[0], 16)] = 0
|
|
|
|
for line in plines:
|
|
# Characters with the property “Prepended_Concatenation_Mark”
|
|
# should have the width 1:
|
|
fields = line.split(";")
|
|
if not '..' in fields[0]:
|
|
code_points = (fields[0], fields[0])
|
|
else:
|
|
code_points = fields[0].split("..")
|
|
for key in range(int(code_points[0], 16),
|
|
int(code_points[1], 16)+1):
|
|
del width_dict[key] # default width is 1
|
|
|
|
# handle special cases for compatibility
|
|
for key in list((0x00AD,)):
|
|
# https://www.cs.tut.fi/~jkorpela/shy.html
|
|
if key in width_dict:
|
|
del width_dict[key] # default width is 1
|
|
for key in list(range(0x1160, 0x1200)):
|
|
width_dict[key] = 0
|
|
for key in list(range(0x3248, 0x3250)):
|
|
# These are “A” which means we can decide whether to treat them
|
|
# as “W” or “N” based on context:
|
|
# http://www.unicode.org/mail-arch/unicode-ml/y2017-m08/0023.html
|
|
# For us, “W” seems better.
|
|
width_dict[key] = 2
|
|
for key in list(range(0x4DC0, 0x4E00)):
|
|
width_dict[key] = 2
|
|
|
|
same_width_lists = []
|
|
current_width_list = []
|
|
for key in sorted(width_dict):
|
|
if not current_width_list:
|
|
current_width_list = [key]
|
|
elif (key == current_width_list[-1] + 1
|
|
and width_dict[key] == width_dict[current_width_list[0]]):
|
|
current_width_list.append(key)
|
|
else:
|
|
same_width_lists.append(current_width_list)
|
|
current_width_list = [key]
|
|
if current_width_list:
|
|
same_width_lists.append(current_width_list)
|
|
|
|
for same_width_list in same_width_lists:
|
|
if len(same_width_list) == 1:
|
|
outfile.write('{:s}\t{:d}\n'.format(
|
|
unicode_utils.ucs_symbol(same_width_list[0]),
|
|
width_dict[same_width_list[0]]))
|
|
else:
|
|
outfile.write('{:s}...{:s}\t{:d}\n'.format(
|
|
unicode_utils.ucs_symbol(same_width_list[0]),
|
|
unicode_utils.ucs_symbol(same_width_list[-1]),
|
|
width_dict[same_width_list[0]]))
|
|
|
|
if __name__ == "__main__":
|
|
PARSER = argparse.ArgumentParser(
|
|
description='''
|
|
Generate a UTF-8 file from UnicodeData.txt, EastAsianWidth.txt, and PropList.txt.
|
|
''')
|
|
PARSER.add_argument(
|
|
'-u', '--unicode_data_file',
|
|
nargs='?',
|
|
type=str,
|
|
default='UnicodeData.txt',
|
|
help=('The UnicodeData.txt file to read, '
|
|
+ 'default: %(default)s'))
|
|
PARSER.add_argument(
|
|
'-e', '--east_asian_with_file',
|
|
nargs='?',
|
|
type=str,
|
|
default='EastAsianWidth.txt',
|
|
help=('The EastAsianWidth.txt file to read, '
|
|
+ 'default: %(default)s'))
|
|
PARSER.add_argument(
|
|
'-p', '--prop_list_file',
|
|
nargs='?',
|
|
type=str,
|
|
default='PropList.txt',
|
|
help=('The PropList.txt file to read, '
|
|
+ 'default: %(default)s'))
|
|
PARSER.add_argument(
|
|
'--unicode_version',
|
|
nargs='?',
|
|
required=True,
|
|
type=str,
|
|
help='The Unicode version of the input files used.')
|
|
ARGS = PARSER.parse_args()
|
|
|
|
with open(ARGS.unicode_data_file, mode='r') as UNIDATA_FILE:
|
|
UNICODE_DATA_LINES = UNIDATA_FILE.readlines()
|
|
with open(ARGS.east_asian_with_file, mode='r') as EAST_ASIAN_WIDTH_FILE:
|
|
EAST_ASIAN_WIDTH_LINES = []
|
|
for LINE in EAST_ASIAN_WIDTH_FILE:
|
|
# If characters from EastAasianWidth.txt which are from
|
|
# from reserved ranges (i.e. not yet assigned code points)
|
|
# are added to the WIDTH section of the UTF-8 file, then
|
|
# “make check” produces “Unknown Character” errors for
|
|
# these code points because such unassigned code points
|
|
# are not in the CHARMAP section of the UTF-8 file.
|
|
#
|
|
# Therefore, we skip all reserved code points when reading
|
|
# the EastAsianWidth.txt file.
|
|
if re.match(r'.*<reserved-.+>\.\.<reserved-.+>.*', LINE):
|
|
continue
|
|
if re.match(r'^[^;]*;[WF]', LINE):
|
|
EAST_ASIAN_WIDTH_LINES.append(LINE.strip())
|
|
with open(ARGS.prop_list_file, mode='r') as PROP_LIST_FILE:
|
|
PROP_LIST_LINES = []
|
|
for LINE in PROP_LIST_FILE:
|
|
if re.match(r'^[^;]*;[\s]*Prepended_Concatenation_Mark', LINE):
|
|
PROP_LIST_LINES.append(LINE.strip())
|
|
with open('UTF-8', mode='w') as OUTFILE:
|
|
# Processing UnicodeData.txt and write CHARMAP to UTF-8 file
|
|
write_header_charmap(OUTFILE)
|
|
process_charmap(UNICODE_DATA_LINES, OUTFILE)
|
|
OUTFILE.write("END CHARMAP\n\n")
|
|
# Processing EastAsianWidth.txt and write WIDTH to UTF-8 file
|
|
write_header_width(OUTFILE, ARGS.unicode_version)
|
|
process_width(OUTFILE,
|
|
UNICODE_DATA_LINES,
|
|
EAST_ASIAN_WIDTH_LINES,
|
|
PROP_LIST_LINES)
|
|
OUTFILE.write("END WIDTH\n")
|