You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
bazarr/libs/charset_normalizer/constant.py

496 lines
19 KiB

from codecs import BOM_UTF8, BOM_UTF16_BE, BOM_UTF16_LE, BOM_UTF32_BE, BOM_UTF32_LE
from encodings.aliases import aliases
from re import IGNORECASE, compile as re_compile
from typing import Dict, List, Set, Union
from .assets import FREQUENCIES
# Contain for each eligible encoding a list of/item bytes SIG/BOM
ENCODING_MARKS: Dict[str, Union[bytes, List[bytes]]] = {
"utf_8": BOM_UTF8,
"utf_7": [
b"\x2b\x2f\x76\x38",
b"\x2b\x2f\x76\x39",
b"\x2b\x2f\x76\x2b",
b"\x2b\x2f\x76\x2f",
b"\x2b\x2f\x76\x38\x2d",
],
"gb18030": b"\x84\x31\x95\x33",
"utf_32": [BOM_UTF32_BE, BOM_UTF32_LE],
"utf_16": [BOM_UTF16_BE, BOM_UTF16_LE],
}
TOO_SMALL_SEQUENCE: int = 32
TOO_BIG_SEQUENCE: int = int(10e6)
UTF8_MAXIMAL_ALLOCATION: int = 1112064
UNICODE_RANGES_COMBINED: Dict[str, range] = {
"Control character": range(31 + 1),
"Basic Latin": range(32, 127 + 1),
"Latin-1 Supplement": range(128, 255 + 1),
"Latin Extended-A": range(256, 383 + 1),
"Latin Extended-B": range(384, 591 + 1),
"IPA Extensions": range(592, 687 + 1),
"Spacing Modifier Letters": range(688, 767 + 1),
"Combining Diacritical Marks": range(768, 879 + 1),
"Greek and Coptic": range(880, 1023 + 1),
"Cyrillic": range(1024, 1279 + 1),
"Cyrillic Supplement": range(1280, 1327 + 1),
"Armenian": range(1328, 1423 + 1),
"Hebrew": range(1424, 1535 + 1),
"Arabic": range(1536, 1791 + 1),
"Syriac": range(1792, 1871 + 1),
"Arabic Supplement": range(1872, 1919 + 1),
"Thaana": range(1920, 1983 + 1),
"NKo": range(1984, 2047 + 1),
"Samaritan": range(2048, 2111 + 1),
"Mandaic": range(2112, 2143 + 1),
"Syriac Supplement": range(2144, 2159 + 1),
"Arabic Extended-A": range(2208, 2303 + 1),
"Devanagari": range(2304, 2431 + 1),
"Bengali": range(2432, 2559 + 1),
"Gurmukhi": range(2560, 2687 + 1),
"Gujarati": range(2688, 2815 + 1),
"Oriya": range(2816, 2943 + 1),
"Tamil": range(2944, 3071 + 1),
"Telugu": range(3072, 3199 + 1),
"Kannada": range(3200, 3327 + 1),
"Malayalam": range(3328, 3455 + 1),
"Sinhala": range(3456, 3583 + 1),
"Thai": range(3584, 3711 + 1),
"Lao": range(3712, 3839 + 1),
"Tibetan": range(3840, 4095 + 1),
"Myanmar": range(4096, 4255 + 1),
"Georgian": range(4256, 4351 + 1),
"Hangul Jamo": range(4352, 4607 + 1),
"Ethiopic": range(4608, 4991 + 1),
"Ethiopic Supplement": range(4992, 5023 + 1),
"Cherokee": range(5024, 5119 + 1),
"Unified Canadian Aboriginal Syllabics": range(5120, 5759 + 1),
"Ogham": range(5760, 5791 + 1),
"Runic": range(5792, 5887 + 1),
"Tagalog": range(5888, 5919 + 1),
"Hanunoo": range(5920, 5951 + 1),
"Buhid": range(5952, 5983 + 1),
"Tagbanwa": range(5984, 6015 + 1),
"Khmer": range(6016, 6143 + 1),
"Mongolian": range(6144, 6319 + 1),
"Unified Canadian Aboriginal Syllabics Extended": range(6320, 6399 + 1),
"Limbu": range(6400, 6479 + 1),
"Tai Le": range(6480, 6527 + 1),
"New Tai Lue": range(6528, 6623 + 1),
"Khmer Symbols": range(6624, 6655 + 1),
"Buginese": range(6656, 6687 + 1),
"Tai Tham": range(6688, 6831 + 1),
"Combining Diacritical Marks Extended": range(6832, 6911 + 1),
"Balinese": range(6912, 7039 + 1),
"Sundanese": range(7040, 7103 + 1),
"Batak": range(7104, 7167 + 1),
"Lepcha": range(7168, 7247 + 1),
"Ol Chiki": range(7248, 7295 + 1),
"Cyrillic Extended C": range(7296, 7311 + 1),
"Sundanese Supplement": range(7360, 7375 + 1),
"Vedic Extensions": range(7376, 7423 + 1),
"Phonetic Extensions": range(7424, 7551 + 1),
"Phonetic Extensions Supplement": range(7552, 7615 + 1),
"Combining Diacritical Marks Supplement": range(7616, 7679 + 1),
"Latin Extended Additional": range(7680, 7935 + 1),
"Greek Extended": range(7936, 8191 + 1),
"General Punctuation": range(8192, 8303 + 1),
"Superscripts and Subscripts": range(8304, 8351 + 1),
"Currency Symbols": range(8352, 8399 + 1),
"Combining Diacritical Marks for Symbols": range(8400, 8447 + 1),
"Letterlike Symbols": range(8448, 8527 + 1),
"Number Forms": range(8528, 8591 + 1),
"Arrows": range(8592, 8703 + 1),
"Mathematical Operators": range(8704, 8959 + 1),
"Miscellaneous Technical": range(8960, 9215 + 1),
"Control Pictures": range(9216, 9279 + 1),
"Optical Character Recognition": range(9280, 9311 + 1),
"Enclosed Alphanumerics": range(9312, 9471 + 1),
"Box Drawing": range(9472, 9599 + 1),
"Block Elements": range(9600, 9631 + 1),
"Geometric Shapes": range(9632, 9727 + 1),
"Miscellaneous Symbols": range(9728, 9983 + 1),
"Dingbats": range(9984, 10175 + 1),
"Miscellaneous Mathematical Symbols-A": range(10176, 10223 + 1),
"Supplemental Arrows-A": range(10224, 10239 + 1),
"Braille Patterns": range(10240, 10495 + 1),
"Supplemental Arrows-B": range(10496, 10623 + 1),
"Miscellaneous Mathematical Symbols-B": range(10624, 10751 + 1),
"Supplemental Mathematical Operators": range(10752, 11007 + 1),
"Miscellaneous Symbols and Arrows": range(11008, 11263 + 1),
"Glagolitic": range(11264, 11359 + 1),
"Latin Extended-C": range(11360, 11391 + 1),
"Coptic": range(11392, 11519 + 1),
"Georgian Supplement": range(11520, 11567 + 1),
"Tifinagh": range(11568, 11647 + 1),
"Ethiopic Extended": range(11648, 11743 + 1),
"Cyrillic Extended-A": range(11744, 11775 + 1),
"Supplemental Punctuation": range(11776, 11903 + 1),
"CJK Radicals Supplement": range(11904, 12031 + 1),
"Kangxi Radicals": range(12032, 12255 + 1),
"Ideographic Description Characters": range(12272, 12287 + 1),
"CJK Symbols and Punctuation": range(12288, 12351 + 1),
"Hiragana": range(12352, 12447 + 1),
"Katakana": range(12448, 12543 + 1),
"Bopomofo": range(12544, 12591 + 1),
"Hangul Compatibility Jamo": range(12592, 12687 + 1),
"Kanbun": range(12688, 12703 + 1),
"Bopomofo Extended": range(12704, 12735 + 1),
"CJK Strokes": range(12736, 12783 + 1),
"Katakana Phonetic Extensions": range(12784, 12799 + 1),
"Enclosed CJK Letters and Months": range(12800, 13055 + 1),
"CJK Compatibility": range(13056, 13311 + 1),
"CJK Unified Ideographs Extension A": range(13312, 19903 + 1),
"Yijing Hexagram Symbols": range(19904, 19967 + 1),
"CJK Unified Ideographs": range(19968, 40959 + 1),
"Yi Syllables": range(40960, 42127 + 1),
"Yi Radicals": range(42128, 42191 + 1),
"Lisu": range(42192, 42239 + 1),
"Vai": range(42240, 42559 + 1),
"Cyrillic Extended-B": range(42560, 42655 + 1),
"Bamum": range(42656, 42751 + 1),
"Modifier Tone Letters": range(42752, 42783 + 1),
"Latin Extended-D": range(42784, 43007 + 1),
"Syloti Nagri": range(43008, 43055 + 1),
"Common Indic Number Forms": range(43056, 43071 + 1),
"Phags-pa": range(43072, 43135 + 1),
"Saurashtra": range(43136, 43231 + 1),
"Devanagari Extended": range(43232, 43263 + 1),
"Kayah Li": range(43264, 43311 + 1),
"Rejang": range(43312, 43359 + 1),
"Hangul Jamo Extended-A": range(43360, 43391 + 1),
"Javanese": range(43392, 43487 + 1),
"Myanmar Extended-B": range(43488, 43519 + 1),
"Cham": range(43520, 43615 + 1),
"Myanmar Extended-A": range(43616, 43647 + 1),
"Tai Viet": range(43648, 43743 + 1),
"Meetei Mayek Extensions": range(43744, 43775 + 1),
"Ethiopic Extended-A": range(43776, 43823 + 1),
"Latin Extended-E": range(43824, 43887 + 1),
"Cherokee Supplement": range(43888, 43967 + 1),
"Meetei Mayek": range(43968, 44031 + 1),
"Hangul Syllables": range(44032, 55215 + 1),
"Hangul Jamo Extended-B": range(55216, 55295 + 1),
"High Surrogates": range(55296, 56191 + 1),
"High Private Use Surrogates": range(56192, 56319 + 1),
"Low Surrogates": range(56320, 57343 + 1),
"Private Use Area": range(57344, 63743 + 1),
"CJK Compatibility Ideographs": range(63744, 64255 + 1),
"Alphabetic Presentation Forms": range(64256, 64335 + 1),
"Arabic Presentation Forms-A": range(64336, 65023 + 1),
"Variation Selectors": range(65024, 65039 + 1),
"Vertical Forms": range(65040, 65055 + 1),
"Combining Half Marks": range(65056, 65071 + 1),
"CJK Compatibility Forms": range(65072, 65103 + 1),
"Small Form Variants": range(65104, 65135 + 1),
"Arabic Presentation Forms-B": range(65136, 65279 + 1),
"Halfwidth and Fullwidth Forms": range(65280, 65519 + 1),
"Specials": range(65520, 65535 + 1),
"Linear B Syllabary": range(65536, 65663 + 1),
"Linear B Ideograms": range(65664, 65791 + 1),
"Aegean Numbers": range(65792, 65855 + 1),
"Ancient Greek Numbers": range(65856, 65935 + 1),
"Ancient Symbols": range(65936, 65999 + 1),
"Phaistos Disc": range(66000, 66047 + 1),
"Lycian": range(66176, 66207 + 1),
"Carian": range(66208, 66271 + 1),
"Coptic Epact Numbers": range(66272, 66303 + 1),
"Old Italic": range(66304, 66351 + 1),
"Gothic": range(66352, 66383 + 1),
"Old Permic": range(66384, 66431 + 1),
"Ugaritic": range(66432, 66463 + 1),
"Old Persian": range(66464, 66527 + 1),
"Deseret": range(66560, 66639 + 1),
"Shavian": range(66640, 66687 + 1),
"Osmanya": range(66688, 66735 + 1),
"Osage": range(66736, 66815 + 1),
"Elbasan": range(66816, 66863 + 1),
"Caucasian Albanian": range(66864, 66927 + 1),
"Linear A": range(67072, 67455 + 1),
"Cypriot Syllabary": range(67584, 67647 + 1),
"Imperial Aramaic": range(67648, 67679 + 1),
"Palmyrene": range(67680, 67711 + 1),
"Nabataean": range(67712, 67759 + 1),
"Hatran": range(67808, 67839 + 1),
"Phoenician": range(67840, 67871 + 1),
"Lydian": range(67872, 67903 + 1),
"Meroitic Hieroglyphs": range(67968, 67999 + 1),
"Meroitic Cursive": range(68000, 68095 + 1),
"Kharoshthi": range(68096, 68191 + 1),
"Old South Arabian": range(68192, 68223 + 1),
"Old North Arabian": range(68224, 68255 + 1),
"Manichaean": range(68288, 68351 + 1),
"Avestan": range(68352, 68415 + 1),
"Inscriptional Parthian": range(68416, 68447 + 1),
"Inscriptional Pahlavi": range(68448, 68479 + 1),
"Psalter Pahlavi": range(68480, 68527 + 1),
"Old Turkic": range(68608, 68687 + 1),
"Old Hungarian": range(68736, 68863 + 1),
"Rumi Numeral Symbols": range(69216, 69247 + 1),
"Brahmi": range(69632, 69759 + 1),
"Kaithi": range(69760, 69839 + 1),
"Sora Sompeng": range(69840, 69887 + 1),
"Chakma": range(69888, 69967 + 1),
"Mahajani": range(69968, 70015 + 1),
"Sharada": range(70016, 70111 + 1),
"Sinhala Archaic Numbers": range(70112, 70143 + 1),
"Khojki": range(70144, 70223 + 1),
"Multani": range(70272, 70319 + 1),
"Khudawadi": range(70320, 70399 + 1),
"Grantha": range(70400, 70527 + 1),
"Newa": range(70656, 70783 + 1),
"Tirhuta": range(70784, 70879 + 1),
"Siddham": range(71040, 71167 + 1),
"Modi": range(71168, 71263 + 1),
"Mongolian Supplement": range(71264, 71295 + 1),
"Takri": range(71296, 71375 + 1),
"Ahom": range(71424, 71487 + 1),
"Warang Citi": range(71840, 71935 + 1),
"Zanabazar Square": range(72192, 72271 + 1),
"Soyombo": range(72272, 72367 + 1),
"Pau Cin Hau": range(72384, 72447 + 1),
"Bhaiksuki": range(72704, 72815 + 1),
"Marchen": range(72816, 72895 + 1),
"Masaram Gondi": range(72960, 73055 + 1),
"Cuneiform": range(73728, 74751 + 1),
"Cuneiform Numbers and Punctuation": range(74752, 74879 + 1),
"Early Dynastic Cuneiform": range(74880, 75087 + 1),
"Egyptian Hieroglyphs": range(77824, 78895 + 1),
"Anatolian Hieroglyphs": range(82944, 83583 + 1),
"Bamum Supplement": range(92160, 92735 + 1),
"Mro": range(92736, 92783 + 1),
"Bassa Vah": range(92880, 92927 + 1),
"Pahawh Hmong": range(92928, 93071 + 1),
"Miao": range(93952, 94111 + 1),
"Ideographic Symbols and Punctuation": range(94176, 94207 + 1),
"Tangut": range(94208, 100351 + 1),
"Tangut Components": range(100352, 101119 + 1),
"Kana Supplement": range(110592, 110847 + 1),
"Kana Extended-A": range(110848, 110895 + 1),
"Nushu": range(110960, 111359 + 1),
"Duployan": range(113664, 113823 + 1),
"Shorthand Format Controls": range(113824, 113839 + 1),
"Byzantine Musical Symbols": range(118784, 119039 + 1),
"Musical Symbols": range(119040, 119295 + 1),
"Ancient Greek Musical Notation": range(119296, 119375 + 1),
"Tai Xuan Jing Symbols": range(119552, 119647 + 1),
"Counting Rod Numerals": range(119648, 119679 + 1),
"Mathematical Alphanumeric Symbols": range(119808, 120831 + 1),
"Sutton SignWriting": range(120832, 121519 + 1),
"Glagolitic Supplement": range(122880, 122927 + 1),
"Mende Kikakui": range(124928, 125151 + 1),
"Adlam": range(125184, 125279 + 1),
"Arabic Mathematical Alphabetic Symbols": range(126464, 126719 + 1),
"Mahjong Tiles": range(126976, 127023 + 1),
"Domino Tiles": range(127024, 127135 + 1),
"Playing Cards": range(127136, 127231 + 1),
"Enclosed Alphanumeric Supplement": range(127232, 127487 + 1),
"Enclosed Ideographic Supplement": range(127488, 127743 + 1),
"Miscellaneous Symbols and Pictographs": range(127744, 128511 + 1),
"Emoticons range(Emoji)": range(128512, 128591 + 1),
"Ornamental Dingbats": range(128592, 128639 + 1),
"Transport and Map Symbols": range(128640, 128767 + 1),
"Alchemical Symbols": range(128768, 128895 + 1),
"Geometric Shapes Extended": range(128896, 129023 + 1),
"Supplemental Arrows-C": range(129024, 129279 + 1),
"Supplemental Symbols and Pictographs": range(129280, 129535 + 1),
"CJK Unified Ideographs Extension B": range(131072, 173791 + 1),
"CJK Unified Ideographs Extension C": range(173824, 177983 + 1),
"CJK Unified Ideographs Extension D": range(177984, 178207 + 1),
"CJK Unified Ideographs Extension E": range(178208, 183983 + 1),
"CJK Unified Ideographs Extension F": range(183984, 191471 + 1),
"CJK Compatibility Ideographs Supplement": range(194560, 195103 + 1),
"Tags": range(917504, 917631 + 1),
"Variation Selectors Supplement": range(917760, 917999 + 1),
}
UNICODE_SECONDARY_RANGE_KEYWORD: List[str] = [
"Supplement",
"Extended",
"Extensions",
"Modifier",
"Marks",
"Punctuation",
"Symbols",
"Forms",
"Operators",
"Miscellaneous",
"Drawing",
"Block",
"Shapes",
"Supplemental",
"Tags",
]
RE_POSSIBLE_ENCODING_INDICATION = re_compile(
r"(?:(?:encoding)|(?:charset)|(?:coding))(?:[\:= ]{1,10})(?:[\"\']?)([a-zA-Z0-9\-_]+)(?:[\"\']?)",
IGNORECASE,
)
IANA_SUPPORTED: List[str] = sorted(
filter(
lambda x: x.endswith("_codec") is False
and x not in {"rot_13", "tactis", "mbcs"},
list(set(aliases.values())),
)
)
IANA_SUPPORTED_COUNT: int = len(IANA_SUPPORTED)
# pre-computed code page that are similar using the function cp_similarity.
IANA_SUPPORTED_SIMILAR: Dict[str, List[str]] = {
"cp037": ["cp1026", "cp1140", "cp273", "cp500"],
"cp1026": ["cp037", "cp1140", "cp273", "cp500"],
"cp1125": ["cp866"],
"cp1140": ["cp037", "cp1026", "cp273", "cp500"],
"cp1250": ["iso8859_2"],
"cp1251": ["kz1048", "ptcp154"],
"cp1252": ["iso8859_15", "iso8859_9", "latin_1"],
"cp1253": ["iso8859_7"],
"cp1254": ["iso8859_15", "iso8859_9", "latin_1"],
"cp1257": ["iso8859_13"],
"cp273": ["cp037", "cp1026", "cp1140", "cp500"],
"cp437": ["cp850", "cp858", "cp860", "cp861", "cp862", "cp863", "cp865"],
"cp500": ["cp037", "cp1026", "cp1140", "cp273"],
"cp850": ["cp437", "cp857", "cp858", "cp865"],
"cp857": ["cp850", "cp858", "cp865"],
"cp858": ["cp437", "cp850", "cp857", "cp865"],
"cp860": ["cp437", "cp861", "cp862", "cp863", "cp865"],
"cp861": ["cp437", "cp860", "cp862", "cp863", "cp865"],
"cp862": ["cp437", "cp860", "cp861", "cp863", "cp865"],
"cp863": ["cp437", "cp860", "cp861", "cp862", "cp865"],
"cp865": ["cp437", "cp850", "cp857", "cp858", "cp860", "cp861", "cp862", "cp863"],
"cp866": ["cp1125"],
"iso8859_10": ["iso8859_14", "iso8859_15", "iso8859_4", "iso8859_9", "latin_1"],
"iso8859_11": ["tis_620"],
"iso8859_13": ["cp1257"],
"iso8859_14": [
"iso8859_10",
"iso8859_15",
"iso8859_16",
"iso8859_3",
"iso8859_9",
"latin_1",
],
"iso8859_15": [
"cp1252",
"cp1254",
"iso8859_10",
"iso8859_14",
"iso8859_16",
"iso8859_3",
"iso8859_9",
"latin_1",
],
"iso8859_16": [
"iso8859_14",
"iso8859_15",
"iso8859_2",
"iso8859_3",
"iso8859_9",
"latin_1",
],
"iso8859_2": ["cp1250", "iso8859_16", "iso8859_4"],
"iso8859_3": ["iso8859_14", "iso8859_15", "iso8859_16", "iso8859_9", "latin_1"],
"iso8859_4": ["iso8859_10", "iso8859_2", "iso8859_9", "latin_1"],
"iso8859_7": ["cp1253"],
"iso8859_9": [
"cp1252",
"cp1254",
"cp1258",
"iso8859_10",
"iso8859_14",
"iso8859_15",
"iso8859_16",
"iso8859_3",
"iso8859_4",
"latin_1",
],
"kz1048": ["cp1251", "ptcp154"],
"latin_1": [
"cp1252",
"cp1254",
"cp1258",
"iso8859_10",
"iso8859_14",
"iso8859_15",
"iso8859_16",
"iso8859_3",
"iso8859_4",
"iso8859_9",
],
"mac_iceland": ["mac_roman", "mac_turkish"],
"mac_roman": ["mac_iceland", "mac_turkish"],
"mac_turkish": ["mac_iceland", "mac_roman"],
"ptcp154": ["cp1251", "kz1048"],
"tis_620": ["iso8859_11"],
}
CHARDET_CORRESPONDENCE: Dict[str, str] = {
"iso2022_kr": "ISO-2022-KR",
"iso2022_jp": "ISO-2022-JP",
"euc_kr": "EUC-KR",
"tis_620": "TIS-620",
"utf_32": "UTF-32",
"euc_jp": "EUC-JP",
"koi8_r": "KOI8-R",
"iso8859_1": "ISO-8859-1",
"iso8859_2": "ISO-8859-2",
"iso8859_5": "ISO-8859-5",
"iso8859_6": "ISO-8859-6",
"iso8859_7": "ISO-8859-7",
"iso8859_8": "ISO-8859-8",
"utf_16": "UTF-16",
"cp855": "IBM855",
"mac_cyrillic": "MacCyrillic",
"gb2312": "GB2312",
"gb18030": "GB18030",
"cp932": "CP932",
"cp866": "IBM866",
"utf_8": "utf-8",
"utf_8_sig": "UTF-8-SIG",
"shift_jis": "SHIFT_JIS",
"big5": "Big5",
"cp1250": "windows-1250",
"cp1251": "windows-1251",
"cp1252": "Windows-1252",
"cp1253": "windows-1253",
"cp1255": "windows-1255",
"cp1256": "windows-1256",
"cp1254": "Windows-1254",
"cp949": "CP949",
}
COMMON_SAFE_ASCII_CHARACTERS: Set[str] = {
"<",
">",
"=",
":",
"/",
"&",
";",
"{",
"}",
"[",
"]",
",",
"|",
'"',
"-",
}
KO_NAMES: Set[str] = {"johab", "cp949", "euc_kr"}
ZH_NAMES: Set[str] = {"big5", "cp950", "big5hkscs", "hz"}
LANGUAGE_SUPPORTED_COUNT: int = len(FREQUENCIES)
# Logging LEVEL below DEBUG
TRACE: int = 5