Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Skip to content

Commit 9a206d0

Browse files
committed
Improve script generating unaccent rules
Script now use the standard Unicode transliterator Latin-ASCII. Author: Leonard Benedetti
1 parent 3aff33a commit 9a206d0

File tree

2 files changed

+762
-56
lines changed

2 files changed

+762
-56
lines changed

contrib/unaccent/generate_unaccent_rules.py

+107-37
Original file line numberDiff line numberDiff line change
@@ -1,20 +1,33 @@
1-
#!/usr/bin/python
1+
#!/usr/bin/python2
2+
# -*- coding: utf-8 -*-
23
#
34
# This script builds unaccent.rules on standard output when given the
4-
# contents of UnicodeData.txt[1] on standard input. Optionally includes
5-
# ligature expansion, if --expand-ligatures is given on the command line.
5+
# contents of UnicodeData.txt [1] and Latin-ASCII.xml [2] given as
6+
# arguments. Optionally includes ligature expansion and Unicode CLDR
7+
# Latin-ASCII transliterator, enabled by default, this can be disabled
8+
# with "--no-ligatures-expansion" command line option.
69
#
710
# The approach is to use the Unicode decomposition data to identify
811
# precomposed codepoints that are equivalent to a ligature of several
912
# letters, or a base letter with any number of diacritical marks.
10-
# There is also a small set of special cases for codepoints that we
11-
# traditionally support even though Unicode doesn't consider them to
12-
# be ligatures or letters with marks.
1313
#
14-
# [1] http://unicode.org/Public/7.0.0/ucd/UnicodeData.txt
14+
# This approach handles most letters with diacritical marks and some
15+
# ligatures. However, several characters (notably a majority of
16+
# ligatures) don't have decomposition. To handle all these cases, one can
17+
# use a standard Unicode transliterator available in Common Locale Data
18+
# Repository (CLDR): Latin-ASCII. This transliterator associates Unicode
19+
# characters to ASCII-range equivalent. Unless "--no-ligatures-expansion"
20+
# option is enabled, the XML file of this transliterator [2] -- given as a
21+
# command line argument -- will be parsed and used.
22+
#
23+
# [1] http://unicode.org/Public/8.0.0/ucd/UnicodeData.txt
24+
# [2] http://unicode.org/cldr/trac/export/12304/tags/release-28/common/transforms/Latin-ASCII.xml
25+
1526

1627
import re
28+
import argparse
1729
import sys
30+
import xml.etree.ElementTree as ET
1831

1932
def print_record(codepoint, letter):
2033
print (unichr(codepoint) + "\t" + letter).encode("UTF-8")
@@ -63,15 +76,73 @@ def get_plain_letters(codepoint, table):
6376
assert(is_ligature(codepoint, table))
6477
return [get_plain_letter(table[id], table) for id in codepoint.combining_ids]
6578

66-
def main(expand_ligatures):
79+
def parse_cldr_latin_ascii_transliterator(latinAsciiFilePath):
80+
"""Parse the XML file and return a set of tuples (src, trg), where "src"
81+
is the original character and "trg" the substitute."""
82+
charactersSet = set()
83+
84+
# RegEx to parse rules
85+
rulePattern = re.compile(ur'^(?:(.)|(\\u[0-9a-fA-F]{4})) \u2192 (?:\'(.+)\'|(.+)) ;')
86+
87+
# construct tree from XML
88+
transliterationTree = ET.parse(latinAsciiFilePath)
89+
transliterationTreeRoot = transliterationTree.getroot()
90+
91+
for rule in transliterationTreeRoot.findall("./transforms/transform/tRule"):
92+
matches = rulePattern.search(rule.text)
93+
94+
# The regular expression capture four groups corresponding
95+
# to the characters.
96+
#
97+
# Group 1: plain "src" char. Empty if group 2 is not.
98+
# Group 2: unicode-espaced "src" char (e.g. "\u0110"). Empty if group 1 is not.
99+
#
100+
# Group 3: plain "trg" char. Empty if group 4 is not.
101+
# Group 4: plain "trg" char between quotes. Empty if group 3 is not.
102+
if matches is not None:
103+
src = matches.group(1) if matches.group(1) is not None else matches.group(2).decode('unicode-escape')
104+
trg = matches.group(3) if matches.group(3) is not None else matches.group(4)
105+
106+
# "'" and """ are escaped
107+
trg = trg.replace("\\'", "'").replace('\\"', '"')
108+
109+
# the parser of unaccent only accepts non-whitespace characters
110+
# for "src" and "trg" (see unaccent.c)
111+
if not src.isspace() and not trg.isspace():
112+
charactersSet.add((ord(src), trg))
113+
114+
return charactersSet
115+
116+
def special_cases():
117+
"""Returns the special cases which are not handled by other methods"""
118+
charactersSet = set()
119+
120+
# Cyrillic
121+
charactersSet.add((0x0401, u"\u0415")) # CYRILLIC CAPITAL LETTER IO
122+
charactersSet.add((0x0451, u"\u0435")) # CYRILLIC SMALL LETTER IO
123+
124+
# Symbols of "Letterlike Symbols" Unicode Block (U+2100 to U+214F)
125+
charactersSet.add((0x2103, u"\xb0C")) # DEGREE CELSIUS
126+
charactersSet.add((0x2109, u"\xb0F")) # DEGREE FAHRENHEIT
127+
charactersSet.add((0x2117, "(P)")) # SOUND RECORDING COPYRIGHT
128+
129+
return charactersSet
130+
131+
def main(args):
67132
# http://www.unicode.org/reports/tr44/tr44-14.html#Character_Decomposition_Mappings
68133
decomposition_type_pattern = re.compile(" *<[^>]*> *")
69134

70135
table = {}
71136
all = []
72137

138+
# unordered set for ensure uniqueness
139+
charactersSet = set()
140+
141+
# read file UnicodeData.txt
142+
unicodeDataFile = open(args.unicodeDataFilePath, 'r')
143+
73144
# read everything we need into memory
74-
for line in sys.stdin.readlines():
145+
for line in unicodeDataFile:
75146
fields = line.split(";")
76147
if len(fields) > 5:
77148
# http://www.unicode.org/reports/tr44/tr44-14.html#UnicodeData.txt
@@ -89,35 +160,34 @@ def main(expand_ligatures):
89160
if codepoint.general_category.startswith('L') and \
90161
len(codepoint.combining_ids) > 1:
91162
if is_letter_with_marks(codepoint, table):
92-
print_record(codepoint.id,
93-
chr(get_plain_letter(codepoint, table).id))
94-
elif expand_ligatures and is_ligature(codepoint, table):
95-
print_record(codepoint.id,
163+
charactersSet.add((codepoint.id,
164+
chr(get_plain_letter(codepoint, table).id)))
165+
elif args.noLigaturesExpansion is False and is_ligature(codepoint, table):
166+
charactersSet.add((codepoint.id,
96167
"".join(unichr(combining_codepoint.id)
97168
for combining_codepoint \
98-
in get_plain_letters(codepoint, table)))
99-
100-
# some special cases
101-
print_record(0x00d8, "O") # LATIN CAPITAL LETTER O WITH STROKE
102-
print_record(0x00f8, "o") # LATIN SMALL LETTER O WITH STROKE
103-
print_record(0x0110, "D") # LATIN CAPITAL LETTER D WITH STROKE
104-
print_record(0x0111, "d") # LATIN SMALL LETTER D WITH STROKE
105-
print_record(0x0131, "i") # LATIN SMALL LETTER DOTLESS I
106-
print_record(0x0126, "H") # LATIN CAPITAL LETTER H WITH STROKE
107-
print_record(0x0127, "h") # LATIN SMALL LETTER H WITH STROKE
108-
print_record(0x0141, "L") # LATIN CAPITAL LETTER L WITH STROKE
109-
print_record(0x0142, "l") # LATIN SMALL LETTER L WITH STROKE
110-
print_record(0x0149, "'n") # LATIN SMALL LETTER N PRECEDED BY APOSTROPHE
111-
print_record(0x0166, "T") # LATIN CAPITAL LETTER T WITH STROKE
112-
print_record(0x0167, "t") # LATIN SMALL LETTER t WITH STROKE
113-
print_record(0x0401, u"\u0415") # CYRILLIC CAPITAL LETTER IO
114-
print_record(0x0451, u"\u0435") # CYRILLIC SMALL LETTER IO
115-
if expand_ligatures:
116-
print_record(0x00c6, "AE") # LATIN CAPITAL LETTER AE
117-
print_record(0x00df, "ss") # LATIN SMALL LETTER SHARP S
118-
print_record(0x00e6, "ae") # LATIN SMALL LETTER AE
119-
print_record(0x0152, "OE") # LATIN CAPITAL LIGATURE OE
120-
print_record(0x0153, "oe") # LATIN SMALL LIGATURE OE
169+
in get_plain_letters(codepoint, table))))
170+
171+
# add CLDR Latin-ASCII characters
172+
if not args.noLigaturesExpansion:
173+
charactersSet |= parse_cldr_latin_ascii_transliterator(args.latinAsciiFilePath)
174+
charactersSet |= special_cases()
175+
176+
# sort for more convenient display
177+
charactersList = sorted(charactersSet, key=lambda characterPair: characterPair[0])
178+
179+
for characterPair in charactersList:
180+
print_record(characterPair[0], characterPair[1])
121181

122182
if __name__ == "__main__":
123-
main(len(sys.argv) == 2 and sys.argv[1] == "--expand-ligatures")
183+
parser = argparse.ArgumentParser(description='This script builds unaccent.rules on standard output when given the contents of UnicodeData.txt and Latin-ASCII.xml given as arguments.')
184+
parser.add_argument("--unicode-data-file", help="Path to formatted text file corresponding to UnicodeData.txt. See <http://unicode.org/Public/8.0.0/ucd/UnicodeData.txt>.", type=str, required=True, dest='unicodeDataFilePath')
185+
parser.add_argument("--latin-ascii-file", help="Path to XML file from Unicode Common Locale Data Repository (CLDR) corresponding to Latin-ASCII transliterator (Latin-ASCII.xml). See <http://unicode.org/cldr/trac/export/12304/tags/release-28/common/transforms/Latin-ASCII.xml>.", type=str, dest='latinAsciiFilePath')
186+
parser.add_argument("--no-ligatures-expansion", help="Do not expand ligatures and do not use Unicode CLDR Latin-ASCII transliterator. By default, this option is not enabled and \"--latin-ascii-file\" argument is required. If this option is enabled, \"--latin-ascii-file\" argument is optional and ignored.", action="store_true", dest='noLigaturesExpansion')
187+
args = parser.parse_args()
188+
189+
if args.noLigaturesExpansion is False and args.latinAsciiFilePath is None:
190+
sys.stderr.write('You must specify the path to Latin-ASCII transliterator file with \"--latin-ascii-file\" option or use \"--no-ligatures-expansion\" option. Use \"-h\" option for help.')
191+
sys.exit(1)
192+
193+
main(args)

0 commit comments

Comments
 (0)