#!/usr/bin/env python # -*- coding: utf-8 -*- # Based upon makeunicodedata.py # (http://hg.python.org/cpython/file/c8192197d23d/Tools/unicode/makeunicodedata.py) # written by Fredrik Lundh (fredrik@pythonware.com) # # Copyright (C) 2011 Tom Schuster # # 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 3 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, see . from __future__ import print_function import csv import io import re import os import sys from contextlib import closing # ECMAScript 2016 # §11.2 White Space whitespace = [ # python doesn't support using control character names :( 0x9, # CHARACTER TABULATION 0xb, # LINE TABULATION 0xc, # FORM FEED ord(u'\N{SPACE}'), ord(u'\N{NO-BREAK SPACE}'), ord(u'\N{ZERO WIDTH NO-BREAK SPACE}'), # also BOM ] # §11.3 Line Terminators line_terminator = [ 0xa, # LINE FEED 0xd, # CARRIAGE RETURN ord(u'\N{LINE SEPARATOR}'), ord(u'\N{PARAGRAPH SEPARATOR}'), ] # These are also part of IdentifierPart §11.6 Names and Keywords compatibility_identifier_part = [ ord(u'\N{ZERO WIDTH NON-JOINER}'), ord(u'\N{ZERO WIDTH JOINER}'), ] FLAG_SPACE = 1 << 0 FLAG_UNICODE_ID_START = 1 << 1 FLAG_UNICODE_ID_CONTINUE_ONLY = 1 << 2 MAX_BMP = 0xffff public_domain = """ /* * Any copyright is dedicated to the Public Domain. * http://creativecommons.org/licenses/publicdomain/ */ """ mpl_license = """\ /* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*- * vim: set ts=8 sts=4 et sw=4 tw=99: * This Source Code Form is subject to the terms of the Mozilla Public * License, v. 2.0. If a copy of the MPL was not distributed with this * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ """ warning_message = """\ /* Generated by make_unicode.py DO NOT MODIFY */ """ unicode_version_message = """\ /* Unicode version: {0} */ """ def read_unicode_data(unicode_data): """ If you want to understand how this wonderful file format works checkout Unicode Standard Annex #44 - Unicode Character Database http://www.unicode.org/reports/tr44/ """ reader = csv.reader(unicode_data, delimiter=';') while True: row = reader.next() name = row[1] # We need to expand the UAX #44 4.2.3 Code Point Range if name.startswith('<') and name.endswith('First>'): next_row = reader.next() for i in range(int(row[0], 16), int(next_row[0], 16) + 1): row[0] = i row[1] = name[1:-8] yield row else: row[0] = int(row[0], 16) yield row def read_case_folding(case_folding): for line in case_folding: if line == '\n' or line.startswith('#'): continue row = line.split('; ') if row[1] in ['F', 'T']: continue row[0] = int(row[0], 16) row[2] = int(row[2], 16) yield row def read_derived_core_properties(derived_core_properties): for line in derived_core_properties: if line == '\n' or line.startswith('#'): continue row = line.split('#')[0].split(';') char_range = row[0].strip() char_property = row[1].strip() if '..' not in char_range: yield (int(char_range, 16), char_property) else: [start, end] = char_range.split('..') for char in range(int(start, 16), int(end, 16) + 1): yield (char, char_property) def int_ranges(ints): """ Yields consecutive ranges (inclusive) from integer values. """ from itertools import tee, izip_longest (a, b) = tee(sorted(ints)) start = next(b) for (curr, succ) in izip_longest(a, b): if curr + 1 != succ: yield (start, curr) start = succ def utf16_encode(code): NonBMPMin = 0x10000 LeadSurrogateMin = 0xD800 TrailSurrogateMin = 0xDC00 lead = (code - NonBMPMin) / 1024 + LeadSurrogateMin trail = ((code - NonBMPMin) % 1024) + TrailSurrogateMin return lead, trail def make_non_bmp_convert_macro(out_file, name, convert_map): # Find continuous range in convert_map. convert_list = [] entry = None for code in sorted(convert_map.keys()): lead, trail = utf16_encode(code) converted = convert_map[code] diff = converted - code if (entry and code == entry['code'] + entry['length'] and diff == entry['diff'] and lead == entry['lead']): entry['length'] += 1 continue entry = { 'code': code, 'diff': diff, 'length': 1, 'lead': lead, 'trail': trail, } convert_list.append(entry) # Generate macro call for each range. lines = [] for entry in convert_list: from_code = entry['code'] to_code = entry['code'] + entry['length'] - 1 diff = entry['diff'] lead = entry['lead'] from_trail = entry['trail'] to_trail = entry['trail'] + entry['length'] - 1 lines.append(' macro(0x{:x}, 0x{:x}, 0x{:x}, 0x{:x}, 0x{:x}, {:d})'.format( from_code, to_code, lead, from_trail, to_trail, diff)) out_file.write('#define FOR_EACH_NON_BMP_{}(macro) \\\n'.format(name)) out_file.write(' \\\n'.join(lines)) out_file.write('\n') def for_each_non_bmp_group(group_set): # Find continuous range in group_set. group_list = [] entry = None for code in sorted(group_set.keys()): if entry and code == entry['code'] + entry['length']: entry['length'] += 1 continue entry = { 'code': code, 'length': 1 } group_list.append(entry) for entry in group_list: yield (entry['code'], entry['code'] + entry['length'] - 1) def process_derived_core_properties(derived_core_properties): id_start = set() id_continue = set() for (char, prop) in read_derived_core_properties(derived_core_properties): if prop == 'ID_Start': id_start.add(char) if prop == 'ID_Continue': id_continue.add(char) return (id_start, id_continue) def process_unicode_data(unicode_data, derived_core_properties): dummy = (0, 0, 0) table = [dummy] cache = {dummy: 0} index = [0] * (MAX_BMP + 1) same_upper_map = {} same_upper_dummy = (0, 0, 0) same_upper_table = [same_upper_dummy] same_upper_cache = {same_upper_dummy: 0} same_upper_index = [0] * (MAX_BMP + 1) test_table = {} test_space_table = [] non_bmp_lower_map = {} non_bmp_upper_map = {} non_bmp_id_start_set = {} non_bmp_id_cont_set = {} non_bmp_space_set = {} (id_start, id_continue) = process_derived_core_properties(derived_core_properties) for row in read_unicode_data(unicode_data): code = row[0] name = row[1] category = row[2] alias = row[-5] uppercase = row[-3] lowercase = row[-2] flags = 0 if uppercase: upper = int(uppercase, 16) if upper not in same_upper_map: same_upper_map[upper] = [code] else: same_upper_map[upper].append(code) else: upper = code if lowercase: lower = int(lowercase, 16) else: lower = code if code > MAX_BMP: if code != lower: non_bmp_lower_map[code] = lower if code != upper: non_bmp_upper_map[code] = upper if category == 'Zs': non_bmp_space_set[code] = 1 test_space_table.append(code) if code in id_start: non_bmp_id_start_set[code] = 1 if code in id_continue: non_bmp_id_cont_set[code] = 1 continue # we combine whitespace and lineterminators because in pratice we don't need them separated if category == 'Zs' or code in whitespace or code in line_terminator: flags |= FLAG_SPACE test_space_table.append(code) # §11.6 (IdentifierStart) if code in id_start: flags |= FLAG_UNICODE_ID_START # §11.6 (IdentifierPart) elif code in id_continue or code in compatibility_identifier_part: flags |= FLAG_UNICODE_ID_CONTINUE_ONLY test_table[code] = (upper, lower, name, alias) up_d = upper - code low_d = lower - code assert up_d > -65535 and up_d < 65535 assert low_d > -65535 and low_d < 65535 upper = up_d & 0xffff lower = low_d & 0xffff item = (upper, lower, flags) i = cache.get(item) if i is None: assert item not in table cache[item] = i = len(table) table.append(item) index[code] = i for code in range(0, MAX_BMP + 1): entry = test_table.get(code) if not entry: continue (upper, lower, name, alias) = entry if upper not in same_upper_map: continue same_upper_ds = [v - code for v in same_upper_map[upper]] assert len(same_upper_ds) <= 3 assert all([v > -65535 and v < 65535 for v in same_upper_ds]) same_upper = [v & 0xffff for v in same_upper_ds] same_upper_0 = same_upper[0] if len(same_upper) >= 1 else 0 same_upper_1 = same_upper[1] if len(same_upper) >= 2 else 0 same_upper_2 = same_upper[2] if len(same_upper) >= 3 else 0 item = (same_upper_0, same_upper_1, same_upper_2) i = same_upper_cache.get(item) if i is None: assert item not in same_upper_table same_upper_cache[item] = i = len(same_upper_table) same_upper_table.append(item) same_upper_index[code] = i return ( table, index, same_upper_table, same_upper_index, non_bmp_lower_map, non_bmp_upper_map, non_bmp_space_set, non_bmp_id_start_set, non_bmp_id_cont_set, test_table, test_space_table, ) def process_case_folding(case_folding): folding_map = {} rev_folding_map = {} folding_dummy = (0, 0, 0, 0) folding_table = [folding_dummy] folding_cache = {folding_dummy: 0} folding_index = [0] * (MAX_BMP + 1) folding_tests = [] folding_codes = set() non_bmp_folding_map = {} non_bmp_rev_folding_map = {} for row in read_case_folding(case_folding): code = row[0] mapping = row[2] folding_map[code] = mapping if code > MAX_BMP: non_bmp_folding_map[code] = mapping non_bmp_rev_folding_map[mapping] = code if mapping not in rev_folding_map: rev_folding_map[mapping] = [code] else: rev_folding_map[mapping].append(code) folding_codes.add(code) folding_codes.add(mapping) for code in sorted(folding_codes): if code in folding_map: folding = folding_map[code] else: folding = code if code in rev_folding_map: rev_folding = rev_folding_map[code] elif folding in rev_folding_map: rev_folding = [c for c in rev_folding_map[folding] if c != code] else: rev_folding = [] assert len(rev_folding) <= 3 if folding != code or len(rev_folding): item = [code] if folding != code: item.append(folding) folding_tests.append(item + rev_folding) if code > MAX_BMP: continue folding_d = folding - code rev_folding_ds = [v - code for v in rev_folding] assert folding_d > -65535 and folding_d < 65535 assert all([v > -65535 and v < 65535 for v in rev_folding]) folding = folding_d & 0xffff rev_folding = [v & 0xffff for v in rev_folding_ds] rev_folding_0 = rev_folding[0] if len(rev_folding) >= 1 else 0 rev_folding_1 = rev_folding[1] if len(rev_folding) >= 2 else 0 rev_folding_2 = rev_folding[2] if len(rev_folding) >= 3 else 0 item = (folding, rev_folding_0, rev_folding_1, rev_folding_2) i = folding_cache.get(item) if i is None: assert item not in folding_table folding_cache[item] = i = len(folding_table) folding_table.append(item) folding_index[code] = i return ( folding_table, folding_index, non_bmp_folding_map, non_bmp_rev_folding_map, folding_tests ) def make_non_bmp_file(version, non_bmp_lower_map, non_bmp_upper_map, non_bmp_folding_map, non_bmp_rev_folding_map): file_name = 'UnicodeNonBMP.h'; with io.open(file_name, mode='wb') as non_bmp_file: non_bmp_file.write(mpl_license) non_bmp_file.write('\n') non_bmp_file.write(warning_message) non_bmp_file.write(unicode_version_message.format(version)) non_bmp_file.write(""" #ifndef vm_UnicodeNonBMP_h #define vm_UnicodeNonBMP_h // |macro| receives the following arguments // macro(FROM, TO, LEAD, TRAIL_FROM, TRAIL_TO, DIFF) // FROM: code point where the range starts // TO: code point where the range ends // LEAD: common lead surrogate of FROM and TO // TRAIL_FROM: trail surrogate of FROM // TRAIL_FROM: trail surrogate of TO // DIFF: the difference between the code point in the range and // converted code point """) make_non_bmp_convert_macro(non_bmp_file, 'LOWERCASE', non_bmp_lower_map) non_bmp_file.write('\n') make_non_bmp_convert_macro(non_bmp_file, 'UPPERCASE', non_bmp_upper_map) non_bmp_file.write('\n') make_non_bmp_convert_macro(non_bmp_file, 'CASE_FOLDING', non_bmp_folding_map) non_bmp_file.write('\n') make_non_bmp_convert_macro(non_bmp_file, 'REV_CASE_FOLDING', non_bmp_rev_folding_map) non_bmp_file.write(""" #endif /* vm_UnicodeNonBMP_h */ """) def make_bmp_mapping_test(version, test_table): file_name = '../tests/ecma_5/String/string-upper-lower-mapping.js' with io.open(file_name, mode='wb') as test_mapping: test_mapping.write(warning_message) test_mapping.write(unicode_version_message.format(version)) test_mapping.write(public_domain) test_mapping.write('var mapping = [\n') for code in range(0, MAX_BMP + 1): entry = test_table.get(code) if entry: (upper, lower, name, alias) = entry test_mapping.write(' [' + hex(upper) + ', ' + hex(lower) + '], /* ' + name + (' (' + alias + ')' if alias else '') + ' */\n') else: test_mapping.write(' [' + hex(code) + ', ' + hex(code) + '],\n') test_mapping.write('];') test_mapping.write(""" assertEq(mapping.length, 0x10000); for (var i = 0; i <= 0xffff; i++) { var char = String.fromCharCode(i); var info = mapping[i]; assertEq(char.toUpperCase().charCodeAt(0), info[0]); assertEq(char.toLowerCase().charCodeAt(0), info[1]); } if (typeof reportCompare === "function") reportCompare(true, true); """) def make_non_bmp_mapping_test(version, non_bmp_upper_map, non_bmp_lower_map): file_name = '../tests/ecma_6/String/string-code-point-upper-lower-mapping.js' with io.open(file_name, mode='wb') as test_non_bmp_mapping: test_non_bmp_mapping.write(warning_message) test_non_bmp_mapping.write(unicode_version_message.format(version)) test_non_bmp_mapping.write(public_domain) for code in sorted(non_bmp_upper_map.keys()): test_non_bmp_mapping.write("""\ assertEq(String.fromCodePoint(0x{:x}).toUpperCase().codePointAt(0), 0x{:x}); """.format(code, non_bmp_upper_map[code])) for code in sorted(non_bmp_lower_map.keys()): test_non_bmp_mapping.write("""\ assertEq(String.fromCodePoint(0x{:x}).toLowerCase().codePointAt(0), 0x{:x}); """.format(code, non_bmp_lower_map[code])) test_non_bmp_mapping.write(""" if (typeof reportCompare === "function") reportCompare(true, true); """) def make_space_test(version, test_space_table): file_name = '../tests/ecma_5/String/string-space-trim.js' with io.open(file_name, mode='wb') as test_space: test_space.write(warning_message) test_space.write(unicode_version_message.format(version)) test_space.write(public_domain) test_space.write('var onlySpace = String.fromCharCode(' + ', '.join(map(lambda c: hex(c), test_space_table)) + ');\n') test_space.write(""" assertEq(onlySpace.trim(), ""); assertEq((onlySpace + 'aaaa').trim(), 'aaaa'); assertEq(('aaaa' + onlySpace).trim(), 'aaaa'); assertEq((onlySpace + 'aaaa' + onlySpace).trim(), 'aaaa'); if (typeof reportCompare === "function") reportCompare(true, true); """) def make_icase_test(version, folding_tests): file_name = '../tests/ecma_6/RegExp/unicode-ignoreCase.js' with io.open(file_name, mode='wb') as test_icase: test_icase.write(warning_message) test_icase.write(unicode_version_message.format(version)) test_icase.write(public_domain) test_icase.write(""" var BUGNUMBER = 1135377; var summary = "Implement RegExp unicode flag -- ignoreCase flag."; print(BUGNUMBER + ": " + summary); function test(code, ...equivs) { var codeRe = new RegExp(String.fromCodePoint(code) + "+", "iu"); var ans = String.fromCodePoint(code) + equivs.map(c => String.fromCodePoint(c)).join(""); assertEqArray(codeRe.exec("<" + ans + ">"), [ans]); codeRe = new RegExp("[" + String.fromCodePoint(code) + "]+", "iu"); assertEqArray(codeRe.exec("<" + ans + ">"), [ans]); } """) for args in folding_tests: test_icase.write('test(' + ','.join([hex(c) for c in args]) + ');\n') test_icase.write(""" if (typeof reportCompare === "function") reportCompare(true, true); """) def make_unicode_file(version, table, index, same_upper_table, same_upper_index, folding_table, folding_index, non_bmp_space_set, non_bmp_id_start_set, non_bmp_id_cont_set): index1, index2, shift = splitbins(index) # Don't forget to update CharInfo in Unicode.h if you need to change this assert shift == 6 same_upper_index1, same_upper_index2, same_upper_shift = splitbins(same_upper_index) # Don't forget to update CodepointsWithSameUpperCaseInfo in Unicode.h if you need to change this assert same_upper_shift == 6 folding_index1, folding_index2, folding_shift = splitbins(folding_index) # Don't forget to update CaseFoldInfo in Unicode.h if you need to change this assert folding_shift == 6 # verify correctness for char in index: test = table[index[char]] idx = index1[char >> shift] idx = index2[(idx << shift) + (char & ((1 << shift) - 1))] assert test == table[idx] # verify correctness for char in same_upper_index: test = same_upper_table[same_upper_index[char]] idx = same_upper_index1[char >> same_upper_shift] idx = same_upper_index2[(idx << same_upper_shift) + (char & ((1 << same_upper_shift) - 1))] assert test == same_upper_table[idx] # verify correctness for char in folding_index: test = folding_table[folding_index[char]] idx = folding_index1[char >> folding_shift] idx = folding_index2[(idx << folding_shift) + (char & ((1 << folding_shift) - 1))] assert test == folding_table[idx] comment = """ /* * So how does indexing work? * First let's have a look at a char16_t, 16-bits: * [................] * Step 1: * Extracting the upper 11 bits from the char16_t. * upper = char >> 5 ([***********.....]) * Step 2: * Using these bits to get an reduced index from index1. * index = index1[upper] * Step 3: * Combining the index and the bottom 5 bits of the original char16_t. * real_index = index2[(index << 5) + (char & ((1 << 5) - 1))] ([...********+++++]) * * The advantage here is that the biggest number in index1 doesn't need 10 bits, * but 7 and we save some memory. * * Step 4: * Get the character informations by looking up real_index in js_charinfo. * * Pseudocode of generation: * * let table be the mapping of char16_t => js_charinfo_index * let index1 be an empty array * let index2 be an empty array * let cache be a hash map * * while shift is less then maximal amount you can shift 0xffff before it's 0 * let chunks be table split in chunks of size 2**shift * * for every chunk in chunks * if chunk is in cache * let index be cache[chunk] * else * let index be the max key of index2 + 1 * for element in chunk * push element to index2 * put index as chunk in cache * * push index >> shift to index1 * * increase shift * stop if you found the best shift */ """ def dump(data, name, file): file.write('const uint8_t unicode::' + name + '[] = {\n') line = pad = ' ' * 4 lines = [] for entry in data: assert entry < 256 s = str(entry) s = s.rjust(3) if len(line + s) + 5 > 99: lines.append(line.rstrip()) line = pad + s + ', ' else: line = line + s + ', ' lines.append(line.rstrip()) file.write('\n'.join(lines)) file.write('\n};\n') file_name = 'Unicode.cpp' with io.open(file_name, 'wb') as data_file: data_file.write(warning_message) data_file.write(unicode_version_message.format(version)) data_file.write(public_domain) data_file.write('#include "vm/Unicode.h"\n\n') data_file.write('using namespace js;\n') data_file.write('using namespace js::unicode;\n') data_file.write(comment) data_file.write('const CharacterInfo unicode::js_charinfo[] = {\n') for d in table: data_file.write(' {') data_file.write(', '.join((str(e) for e in d))) data_file.write('},\n') data_file.write('};\n') data_file.write('\n') dump(index1, 'index1', data_file) data_file.write('\n') dump(index2, 'index2', data_file) data_file.write('\n') data_file.write('const CodepointsWithSameUpperCaseInfo unicode::js_codepoints_with_same_upper_info[] = {\n') for d in same_upper_table: data_file.write(' {') data_file.write(', '.join((str(e) for e in d))) data_file.write('},\n') data_file.write('};\n') data_file.write('\n') dump(same_upper_index1, 'codepoints_with_same_upper_index1', data_file) data_file.write('\n') dump(same_upper_index2, 'codepoints_with_same_upper_index2', data_file) data_file.write('\n') data_file.write('const FoldingInfo unicode::js_foldinfo[] = {\n') for d in folding_table: data_file.write(' {') data_file.write(', '.join((str(e) for e in d))) data_file.write('},\n') data_file.write('};\n') data_file.write('\n') dump(folding_index1, 'folding_index1', data_file) data_file.write('\n') dump(folding_index2, 'folding_index2', data_file) data_file.write('\n') # If the following assert fails, it means space character is added to # non-BMP area. In that case the following code should be uncommented # and the corresponding code should be added to frontend. assert len(non_bmp_space_set.keys()) == 0 data_file.write("""\ bool js::unicode::IsIdentifierStartNonBMP(uint32_t codePoint) { """) for (from_code, to_code) in for_each_non_bmp_group(non_bmp_id_start_set): data_file.write("""\ if (codePoint >= 0x{:x} && codePoint <= 0x{:x}) return true; """.format(from_code, to_code)) data_file.write("""\ return false; } bool js::unicode::IsIdentifierPartNonBMP(uint32_t codePoint) { """) for (from_code, to_code) in for_each_non_bmp_group(non_bmp_id_cont_set): data_file.write("""\ if (codePoint >= 0x{:x} && codePoint <= 0x{:x}) return true; """.format(from_code, to_code)) data_file.write("""\ return false; } """) def getsize(data): """ return smallest possible integer size for the given array """ maxdata = max(data) assert maxdata < 2**32 if maxdata < 256: return 1 elif maxdata < 65536: return 2 else: return 4 def splitbins(t): """t -> (t1, t2, shift). Split a table to save space. t is a sequence of ints. This function can be useful to save space if many of the ints are the same. t1 and t2 are lists of ints, and shift is an int, chosen to minimize the combined size of t1 and t2 (in C code), and where for each i in range(len(t)), t[i] == t2[(t1[i >> shift] << shift) + (i & mask)] where mask is a bitmask isolating the last "shift" bits. """ def dump(t1, t2, shift, bytes): print("%d+%d bins at shift %d; %d bytes" % ( len(t1), len(t2), shift, bytes), file=sys.stderr) print("Size of original table:", len(t)*getsize(t), \ "bytes", file=sys.stderr) n = len(t)-1 # last valid index maxshift = 0 # the most we can shift n and still have something left if n > 0: while n >> 1: n >>= 1 maxshift += 1 del n bytes = sys.maxsize # smallest total size so far t = tuple(t) # so slices can be dict keys for shift in range(maxshift + 1): t1 = [] t2 = [] size = 2**shift bincache = {} for i in range(0, len(t), size): bin = t[i:i + size] index = bincache.get(bin) if index is None: index = len(t2) bincache[bin] = index t2.extend(bin) t1.append(index >> shift) # determine memory size b = len(t1) * getsize(t1) + len(t2) * getsize(t2) if b < bytes: best = t1, t2, shift bytes = b t1, t2, shift = best print("Best:", end=' ', file=sys.stderr) dump(t1, t2, shift, bytes) # exhaustively verify that the decomposition is correct mask = 2**shift - 1 for i in range(len(t)): assert t[i] == t2[(t1[i >> shift] << shift) + (i & mask)] return best def make_irregexp_tables(version, table, index, folding_table, folding_index, test_table): import string from functools import partial from itertools import chain, ifilter, imap MAX_ASCII = 0x7F MAX_LATIN1 = 0xFF LEAD_SURROGATE_MIN = 0xD800 TRAIL_SURROGATE_MAX = 0xDFFF def hex2(n): assert 0 <= n and n < 16**2 return '0x{:02X}'.format(n) def hex4(n): assert 0 <= n and n < 16**4 return '0x{:04X}'.format(n) def uhex4(n): assert 0 <= n and n < 16**4 return 'U+{:04X}'.format(n) def case_info(code): assert 0 <= code and code <= MAX_BMP (upper, lower, flags) = table[index[code]] return ((code + upper) & 0xffff, (code + lower) & 0xffff, flags) def is_space(code): (_, _, flags) = case_info(code) return bool(flags & FLAG_SPACE) def to_upper(code): (upper, _, _) = case_info(code) return upper def casefold(code): assert 0 <= code and code <= MAX_BMP (folding, _, _, _) = folding_table[folding_index[code]] return (code + folding) & 0xffff def casefolds_to_ascii(code): return casefold(code) <= MAX_ASCII def casefolds_to_latin1(code): return casefold(code) <= MAX_LATIN1 def casemaps_to_nonlatin1(code): upper = to_upper(code) return upper > MAX_LATIN1 def char_name(code): assert 0 <= code and code <= MAX_BMP if code not in test_table: return '' if code == LEAD_SURROGATE_MIN: return '' if code == TRAIL_SURROGATE_MAX: return '' (_, _, name, alias) = test_table[code] return name if not name.startswith('<') else alias def write_character_range(println, name, characters): char_ranges = list(int_ranges(characters)) println('') println('const int js::irregexp::k{}Ranges[] = {{'.format(name)) for (start, end) in char_ranges: s_name = char_name(start) e_name = char_name(end) println(' {}, {} + 1, // {}'.format(hex4(start), hex4(end), '{}..{}'.format(s_name, e_name) if start != end else s_name)) println(' {} + 1'.format(hex4(MAX_BMP))) println('};') println('const int js::irregexp::k{}RangeCount = {};'.format(name, len(char_ranges) * 2 + 1)) def write_character_test(println, test, consequent, default): # Latin1 characters which, when case-mapped through # String.prototype.toUpperCase(), canonicalize to a non-Latin1 character. # ES2017, §21.2.2.8.2 Runtime Semantics: Canonicalize casemapped_to_nonlatin1 = ifilter(casemaps_to_nonlatin1, xrange(0, MAX_LATIN1 + 1)) def casemap_closure(ch): upper = to_upper(ch) return (ch, [c for c in xrange(MAX_LATIN1 + 1, MAX_BMP + 1) if upper == to_upper(c)]) # Mapping from Latin1 characters to the list of case map equivalent # non-Latin1 characters. casemap_for_latin1 = dict(chain(imap(casemap_closure, casemapped_to_nonlatin1))) # Non-latin1 characters which, when Unicode case-folded, canonicalize to # a Latin1 character. # ES2017, §21.2.2.8.2 Runtime Semantics: Canonicalize casefolded_to_latin1 = ifilter(casefolds_to_latin1, xrange(MAX_LATIN1 + 1, MAX_BMP + 1)) println(' if (unicode) {') for ch in casefolded_to_latin1: casefolded = casefold(ch) # Skip if also handled below for case mapping. if casefolded in casemap_for_latin1 and ch in casemap_for_latin1[casefolded]: continue println(' // "{}" case folds to "{}".'.format(char_name(ch), char_name(casefolded))) println(' if ({})'.format(test(ch))) println(' return {};'.format(consequent(casefolded))) println(' }') println('') for (ch, casemapped_chars) in casemap_for_latin1.iteritems(): for casemapped in casemapped_chars: println(' // "{}" case maps to "{}".'.format(char_name(casemapped), char_name(ch))) println(' if ({})'.format(' || '.join(imap(test, casemapped_chars)))) println(' return {};'.format(consequent(ch))) println(' return {};'.format(default)) with io.open('../irregexp/RegExpCharacters-inl.h', 'wb') as chars_file: write = partial(print, file=chars_file, sep='', end='') println = partial(write, end='\n') write(warning_message) write(unicode_version_message.format(version)) println('#ifndef V8_JSREGEXPCHARACTERS_INL_H_') println('#define V8_JSREGEXPCHARACTERS_INL_H_') println('') println('namespace js {') println('') println('namespace irregexp {') println('') println('static inline bool') println('RangeContainsLatin1Equivalents(CharacterRange range, bool unicode)') println('{') write_character_test(println, lambda ch: 'range.Contains({})'.format(hex4(ch)), lambda _: 'true', 'false') println('}') println('') println('} } // namespace js::irregexp') println('') println('#endif // V8_JSREGEXPCHARACTERS_INL_H_') with io.open('../irregexp/RegExpCharacters.cpp', 'wb') as chars_file: write = partial(print, file=chars_file, sep='', end='') println = partial(write, end='\n') character_range = partial(write_character_range, println) # Characters in \s, 21.2.2.12 CharacterClassEscape. space_chars = filter(is_space, xrange(0, MAX_BMP + 1)) # Characters in \d, 21.2.2.12 CharacterClassEscape. digit_chars = map(ord, string.digits) assert all(ch <= MAX_ASCII for ch in digit_chars) # Characters in \w, 21.2.2.12 CharacterClassEscape. word_chars = map(ord, string.digits + string.ascii_letters + '_') assert all(ch <= MAX_ASCII for ch in word_chars) # Characters which case-fold to characters in \w. ignorecase_word_chars = (word_chars + filter(casefolds_to_ascii, xrange(MAX_ASCII + 1, MAX_BMP + 1))) # Surrogate characters. surrogate_chars = range(LEAD_SURROGATE_MIN, TRAIL_SURROGATE_MAX + 1) write(warning_message) write(unicode_version_message.format(version)) println('#include "irregexp/RegExpCharacters.h"') println('') println('#include "mozilla/Assertions.h"') println('') println('char16_t') println('js::irregexp::ConvertNonLatin1ToLatin1(char16_t c, bool unicode)') println('{') println(' MOZ_ASSERT(c > {}, "Character mustn\'t be Latin1");'.format(hex2(MAX_LATIN1))) write_character_test(println, lambda ch: 'c == {}'.format(hex4(ch)), hex2, '0') println('}') character_range('Space', space_chars) character_range('SpaceAndSurrogate', space_chars + surrogate_chars) character_range('Word', word_chars) character_range('IgnoreCaseWord', ignorecase_word_chars) character_range('WordAndSurrogate', word_chars + surrogate_chars) character_range('NegatedIgnoreCaseWordAndSurrogate', set(xrange(0, MAX_BMP + 1)) - set(ignorecase_word_chars + surrogate_chars)) character_range('Digit', digit_chars) character_range('DigitAndSurrogate', digit_chars + surrogate_chars) character_range('Surrogate', surrogate_chars) character_range('LineTerminator', line_terminator) def update_unicode(args): import urllib2 version = args.version if version is not None: baseurl = 'http://unicode.org/Public' if version == 'UNIDATA': url = '%s/%s' % (baseurl, version) else: url = '%s/%s/ucd' % (baseurl, version) print('Arguments:') if version is not None: print('\tVersion: %s' % version) print('\tDownload url: %s' % url) else: print('\tUsing local files.') print('\tAlways make sure you have the newest Unicode files!') print('') def download_or_open(fname): tfile_path = os.path.join(os.getcwd(), fname) if version is not None: print('Downloading %s...' % fname) unicode_data_url = '%s/%s' % (url, fname) with closing(urllib2.urlopen(unicode_data_url)) as reader: data = reader.read() tfile = io.open(tfile_path, 'w+b') tfile.write(data) tfile.flush() tfile.seek(0) else: if not os.path.isfile(tfile_path): raise RuntimeError('File not found: %s' % tfile_path) tfile = io.open(tfile_path, 'rb'); return tfile def version_from_file(f, fname): pat_version = re.compile(r"# %s-(?P\d+\.\d+\.\d+).txt" % fname) return pat_version.match(f.readline()).group("version") with download_or_open('UnicodeData.txt') as unicode_data, \ download_or_open('CaseFolding.txt') as case_folding, \ download_or_open('DerivedCoreProperties.txt') as derived_core_properties: unicode_version = version_from_file(derived_core_properties, 'DerivedCoreProperties') print('Processing...') ( table, index, same_upper_table, same_upper_index, non_bmp_lower_map, non_bmp_upper_map, non_bmp_space_set, non_bmp_id_start_set, non_bmp_id_cont_set, test_table, test_space_table ) = process_unicode_data(unicode_data, derived_core_properties) ( folding_table, folding_index, non_bmp_folding_map, non_bmp_rev_folding_map, folding_tests ) = process_case_folding(case_folding) print('Generating...') make_unicode_file(unicode_version, table, index, same_upper_table, same_upper_index, folding_table, folding_index, non_bmp_space_set, non_bmp_id_start_set, non_bmp_id_cont_set) make_non_bmp_file(unicode_version, non_bmp_lower_map, non_bmp_upper_map, non_bmp_folding_map, non_bmp_rev_folding_map) make_irregexp_tables(unicode_version, table, index, folding_table, folding_index, test_table) make_bmp_mapping_test(unicode_version, test_table) make_non_bmp_mapping_test(unicode_version, non_bmp_upper_map, non_bmp_lower_map) make_space_test(unicode_version, test_space_table) make_icase_test(unicode_version, folding_tests) if __name__ == '__main__': import argparse # This script must be run from js/src/vm to work correctly. if '/'.join(os.path.normpath(os.getcwd()).split(os.sep)[-3:]) != 'js/src/vm': raise RuntimeError('%s must be run from js/src/vm' % sys.argv[0]) parser = argparse.ArgumentParser(description='Update Unicode data.') parser.add_argument('--version', help='Optional Unicode version number. If specified, downloads the\ selected version from . If not specified\ uses the existing local files to generate the Unicode data. The\ number must match a published Unicode version, e.g. use\ "--version=8.0.0" to download Unicode 8 files. Alternatively use\ "--version=UNIDATA" to download the latest published version.') parser.set_defaults(func=update_unicode) args = parser.parse_args() args.func(args)