sqlglot.tokens
1from __future__ import annotations 2 3import os 4import typing as t 5from enum import auto 6 7from sqlglot.errors import SqlglotError, TokenError 8from sqlglot.helper import AutoName 9from sqlglot.trie import TrieResult, in_trie, new_trie 10 11if t.TYPE_CHECKING: 12 from sqlglot.dialects.dialect import DialectType 13 14 15try: 16 from sqlglotrs import ( # type: ignore 17 Tokenizer as RsTokenizer, 18 TokenizerDialectSettings as RsTokenizerDialectSettings, 19 TokenizerSettings as RsTokenizerSettings, 20 TokenTypeSettings as RsTokenTypeSettings, 21 ) 22 23 USE_RS_TOKENIZER = os.environ.get("SQLGLOTRS_TOKENIZER", "1") == "1" 24except ImportError: 25 USE_RS_TOKENIZER = False 26 27 28class TokenType(AutoName): 29 L_PAREN = auto() 30 R_PAREN = auto() 31 L_BRACKET = auto() 32 R_BRACKET = auto() 33 L_BRACE = auto() 34 R_BRACE = auto() 35 COMMA = auto() 36 DOT = auto() 37 DASH = auto() 38 PLUS = auto() 39 COLON = auto() 40 DCOLON = auto() 41 DQMARK = auto() 42 SEMICOLON = auto() 43 STAR = auto() 44 BACKSLASH = auto() 45 SLASH = auto() 46 LT = auto() 47 LTE = auto() 48 GT = auto() 49 GTE = auto() 50 NOT = auto() 51 EQ = auto() 52 NEQ = auto() 53 NULLSAFE_EQ = auto() 54 COLON_EQ = auto() 55 AND = auto() 56 OR = auto() 57 AMP = auto() 58 DPIPE = auto() 59 PIPE = auto() 60 PIPE_SLASH = auto() 61 DPIPE_SLASH = auto() 62 CARET = auto() 63 TILDA = auto() 64 ARROW = auto() 65 DARROW = auto() 66 FARROW = auto() 67 HASH = auto() 68 HASH_ARROW = auto() 69 DHASH_ARROW = auto() 70 LR_ARROW = auto() 71 DAT = auto() 72 LT_AT = auto() 73 AT_GT = auto() 74 DOLLAR = auto() 75 PARAMETER = auto() 76 SESSION_PARAMETER = auto() 77 DAMP = auto() 78 XOR = auto() 79 DSTAR = auto() 80 81 BLOCK_START = auto() 82 BLOCK_END = auto() 83 84 SPACE = auto() 85 BREAK = auto() 86 87 STRING = auto() 88 NUMBER = auto() 89 IDENTIFIER = auto() 90 DATABASE = auto() 91 COLUMN = auto() 92 COLUMN_DEF = auto() 93 SCHEMA = auto() 94 TABLE = auto() 95 VAR = auto() 96 BIT_STRING = auto() 97 HEX_STRING = auto() 98 BYTE_STRING = auto() 99 NATIONAL_STRING = auto() 100 RAW_STRING = auto() 101 HEREDOC_STRING = auto() 102 UNICODE_STRING = auto() 103 104 # types 105 BIT = auto() 106 BOOLEAN = auto() 107 TINYINT = auto() 108 UTINYINT = auto() 109 SMALLINT = auto() 110 USMALLINT = auto() 111 MEDIUMINT = auto() 112 UMEDIUMINT = auto() 113 INT = auto() 114 UINT = auto() 115 BIGINT = auto() 116 UBIGINT = auto() 117 INT128 = auto() 118 UINT128 = auto() 119 INT256 = auto() 120 UINT256 = auto() 121 FLOAT = auto() 122 DOUBLE = auto() 123 DECIMAL = auto() 124 UDECIMAL = auto() 125 BIGDECIMAL = auto() 126 CHAR = auto() 127 NCHAR = auto() 128 VARCHAR = auto() 129 NVARCHAR = auto() 130 BPCHAR = auto() 131 TEXT = auto() 132 MEDIUMTEXT = auto() 133 LONGTEXT = auto() 134 MEDIUMBLOB = auto() 135 LONGBLOB = auto() 136 TINYBLOB = auto() 137 TINYTEXT = auto() 138 NAME = auto() 139 BINARY = auto() 140 VARBINARY = auto() 141 JSON = auto() 142 JSONB = auto() 143 TIME = auto() 144 TIMETZ = auto() 145 TIMESTAMP = auto() 146 TIMESTAMPTZ = auto() 147 TIMESTAMPLTZ = auto() 148 TIMESTAMPNTZ = auto() 149 TIMESTAMP_S = auto() 150 TIMESTAMP_MS = auto() 151 TIMESTAMP_NS = auto() 152 DATETIME = auto() 153 DATETIME64 = auto() 154 DATE = auto() 155 DATE32 = auto() 156 INT4RANGE = auto() 157 INT4MULTIRANGE = auto() 158 INT8RANGE = auto() 159 INT8MULTIRANGE = auto() 160 NUMRANGE = auto() 161 NUMMULTIRANGE = auto() 162 TSRANGE = auto() 163 TSMULTIRANGE = auto() 164 TSTZRANGE = auto() 165 TSTZMULTIRANGE = auto() 166 DATERANGE = auto() 167 DATEMULTIRANGE = auto() 168 UUID = auto() 169 GEOGRAPHY = auto() 170 NULLABLE = auto() 171 GEOMETRY = auto() 172 HLLSKETCH = auto() 173 HSTORE = auto() 174 SUPER = auto() 175 SERIAL = auto() 176 SMALLSERIAL = auto() 177 BIGSERIAL = auto() 178 XML = auto() 179 YEAR = auto() 180 UNIQUEIDENTIFIER = auto() 181 USERDEFINED = auto() 182 MONEY = auto() 183 SMALLMONEY = auto() 184 ROWVERSION = auto() 185 IMAGE = auto() 186 VARIANT = auto() 187 OBJECT = auto() 188 INET = auto() 189 IPADDRESS = auto() 190 IPPREFIX = auto() 191 IPV4 = auto() 192 IPV6 = auto() 193 ENUM = auto() 194 ENUM8 = auto() 195 ENUM16 = auto() 196 FIXEDSTRING = auto() 197 LOWCARDINALITY = auto() 198 NESTED = auto() 199 AGGREGATEFUNCTION = auto() 200 SIMPLEAGGREGATEFUNCTION = auto() 201 TDIGEST = auto() 202 UNKNOWN = auto() 203 204 # keywords 205 ALIAS = auto() 206 ALTER = auto() 207 ALWAYS = auto() 208 ALL = auto() 209 ANTI = auto() 210 ANY = auto() 211 APPLY = auto() 212 ARRAY = auto() 213 ASC = auto() 214 ASOF = auto() 215 AUTO_INCREMENT = auto() 216 BEGIN = auto() 217 BETWEEN = auto() 218 CACHE = auto() 219 CASE = auto() 220 CHARACTER_SET = auto() 221 CLUSTER_BY = auto() 222 COLLATE = auto() 223 COMMAND = auto() 224 COMMENT = auto() 225 COMMIT = auto() 226 CONNECT_BY = auto() 227 CONSTRAINT = auto() 228 COPY = auto() 229 CREATE = auto() 230 CROSS = auto() 231 CUBE = auto() 232 CURRENT_DATE = auto() 233 CURRENT_DATETIME = auto() 234 CURRENT_TIME = auto() 235 CURRENT_TIMESTAMP = auto() 236 CURRENT_USER = auto() 237 DECLARE = auto() 238 DEFAULT = auto() 239 DELETE = auto() 240 DESC = auto() 241 DESCRIBE = auto() 242 DICTIONARY = auto() 243 DISTINCT = auto() 244 DISTRIBUTE_BY = auto() 245 DIV = auto() 246 DROP = auto() 247 ELSE = auto() 248 END = auto() 249 ESCAPE = auto() 250 EXCEPT = auto() 251 EXECUTE = auto() 252 EXISTS = auto() 253 FALSE = auto() 254 FETCH = auto() 255 FILTER = auto() 256 FINAL = auto() 257 FIRST = auto() 258 FOR = auto() 259 FORCE = auto() 260 FOREIGN_KEY = auto() 261 FORMAT = auto() 262 FROM = auto() 263 FULL = auto() 264 FUNCTION = auto() 265 GLOB = auto() 266 GLOBAL = auto() 267 GROUP_BY = auto() 268 GROUPING_SETS = auto() 269 HAVING = auto() 270 HINT = auto() 271 IGNORE = auto() 272 ILIKE = auto() 273 ILIKE_ANY = auto() 274 IN = auto() 275 INDEX = auto() 276 INNER = auto() 277 INSERT = auto() 278 INTERSECT = auto() 279 INTERVAL = auto() 280 INTO = auto() 281 INTRODUCER = auto() 282 IRLIKE = auto() 283 IS = auto() 284 ISNULL = auto() 285 JOIN = auto() 286 JOIN_MARKER = auto() 287 KEEP = auto() 288 KILL = auto() 289 LANGUAGE = auto() 290 LATERAL = auto() 291 LEFT = auto() 292 LIKE = auto() 293 LIKE_ANY = auto() 294 LIMIT = auto() 295 LOAD = auto() 296 LOCK = auto() 297 MAP = auto() 298 MATCH_CONDITION = auto() 299 MATCH_RECOGNIZE = auto() 300 MEMBER_OF = auto() 301 MERGE = auto() 302 MOD = auto() 303 MODEL = auto() 304 NATURAL = auto() 305 NEXT = auto() 306 NOTNULL = auto() 307 NULL = auto() 308 OBJECT_IDENTIFIER = auto() 309 OFFSET = auto() 310 ON = auto() 311 ONLY = auto() 312 OPERATOR = auto() 313 ORDER_BY = auto() 314 ORDER_SIBLINGS_BY = auto() 315 ORDERED = auto() 316 ORDINALITY = auto() 317 OUTER = auto() 318 OVER = auto() 319 OVERLAPS = auto() 320 OVERWRITE = auto() 321 PARTITION = auto() 322 PARTITION_BY = auto() 323 PERCENT = auto() 324 PIVOT = auto() 325 PLACEHOLDER = auto() 326 POSITIONAL = auto() 327 PRAGMA = auto() 328 PREWHERE = auto() 329 PRIMARY_KEY = auto() 330 PROCEDURE = auto() 331 PROPERTIES = auto() 332 PSEUDO_TYPE = auto() 333 QUALIFY = auto() 334 QUOTE = auto() 335 RANGE = auto() 336 RECURSIVE = auto() 337 REFRESH = auto() 338 REPLACE = auto() 339 RETURNING = auto() 340 REFERENCES = auto() 341 RIGHT = auto() 342 RLIKE = auto() 343 ROLLBACK = auto() 344 ROLLUP = auto() 345 ROW = auto() 346 ROWS = auto() 347 SELECT = auto() 348 SEMI = auto() 349 SEPARATOR = auto() 350 SEQUENCE = auto() 351 SERDE_PROPERTIES = auto() 352 SET = auto() 353 SETTINGS = auto() 354 SHOW = auto() 355 SIMILAR_TO = auto() 356 SOME = auto() 357 SORT_BY = auto() 358 START_WITH = auto() 359 STORAGE_INTEGRATION = auto() 360 STRUCT = auto() 361 TABLE_SAMPLE = auto() 362 TAG = auto() 363 TEMPORARY = auto() 364 TOP = auto() 365 THEN = auto() 366 TRUE = auto() 367 TRUNCATE = auto() 368 UNCACHE = auto() 369 UNION = auto() 370 UNNEST = auto() 371 UNPIVOT = auto() 372 UPDATE = auto() 373 USE = auto() 374 USING = auto() 375 VALUES = auto() 376 VIEW = auto() 377 VOLATILE = auto() 378 WHEN = auto() 379 WHERE = auto() 380 WINDOW = auto() 381 WITH = auto() 382 UNIQUE = auto() 383 VERSION_SNAPSHOT = auto() 384 TIMESTAMP_SNAPSHOT = auto() 385 OPTION = auto() 386 387 388_ALL_TOKEN_TYPES = list(TokenType) 389_TOKEN_TYPE_TO_INDEX = {token_type: i for i, token_type in enumerate(_ALL_TOKEN_TYPES)} 390 391 392class Token: 393 __slots__ = ("token_type", "text", "line", "col", "start", "end", "comments") 394 395 @classmethod 396 def number(cls, number: int) -> Token: 397 """Returns a NUMBER token with `number` as its text.""" 398 return cls(TokenType.NUMBER, str(number)) 399 400 @classmethod 401 def string(cls, string: str) -> Token: 402 """Returns a STRING token with `string` as its text.""" 403 return cls(TokenType.STRING, string) 404 405 @classmethod 406 def identifier(cls, identifier: str) -> Token: 407 """Returns an IDENTIFIER token with `identifier` as its text.""" 408 return cls(TokenType.IDENTIFIER, identifier) 409 410 @classmethod 411 def var(cls, var: str) -> Token: 412 """Returns an VAR token with `var` as its text.""" 413 return cls(TokenType.VAR, var) 414 415 def __init__( 416 self, 417 token_type: TokenType, 418 text: str, 419 line: int = 1, 420 col: int = 1, 421 start: int = 0, 422 end: int = 0, 423 comments: t.Optional[t.List[str]] = None, 424 ) -> None: 425 """Token initializer. 426 427 Args: 428 token_type: The TokenType Enum. 429 text: The text of the token. 430 line: The line that the token ends on. 431 col: The column that the token ends on. 432 start: The start index of the token. 433 end: The ending index of the token. 434 comments: The comments to attach to the token. 435 """ 436 self.token_type = token_type 437 self.text = text 438 self.line = line 439 self.col = col 440 self.start = start 441 self.end = end 442 self.comments = [] if comments is None else comments 443 444 def __repr__(self) -> str: 445 attributes = ", ".join(f"{k}: {getattr(self, k)}" for k in self.__slots__) 446 return f"<Token {attributes}>" 447 448 449class _Tokenizer(type): 450 def __new__(cls, clsname, bases, attrs): 451 klass = super().__new__(cls, clsname, bases, attrs) 452 453 def _convert_quotes(arr: t.List[str | t.Tuple[str, str]]) -> t.Dict[str, str]: 454 return dict( 455 (item, item) if isinstance(item, str) else (item[0], item[1]) for item in arr 456 ) 457 458 def _quotes_to_format( 459 token_type: TokenType, arr: t.List[str | t.Tuple[str, str]] 460 ) -> t.Dict[str, t.Tuple[str, TokenType]]: 461 return {k: (v, token_type) for k, v in _convert_quotes(arr).items()} 462 463 klass._QUOTES = _convert_quotes(klass.QUOTES) 464 klass._IDENTIFIERS = _convert_quotes(klass.IDENTIFIERS) 465 466 klass._FORMAT_STRINGS = { 467 **{ 468 p + s: (e, TokenType.NATIONAL_STRING) 469 for s, e in klass._QUOTES.items() 470 for p in ("n", "N") 471 }, 472 **_quotes_to_format(TokenType.BIT_STRING, klass.BIT_STRINGS), 473 **_quotes_to_format(TokenType.BYTE_STRING, klass.BYTE_STRINGS), 474 **_quotes_to_format(TokenType.HEX_STRING, klass.HEX_STRINGS), 475 **_quotes_to_format(TokenType.RAW_STRING, klass.RAW_STRINGS), 476 **_quotes_to_format(TokenType.HEREDOC_STRING, klass.HEREDOC_STRINGS), 477 **_quotes_to_format(TokenType.UNICODE_STRING, klass.UNICODE_STRINGS), 478 } 479 480 klass._STRING_ESCAPES = set(klass.STRING_ESCAPES) 481 klass._IDENTIFIER_ESCAPES = set(klass.IDENTIFIER_ESCAPES) 482 klass._COMMENTS = { 483 **dict( 484 (comment, None) if isinstance(comment, str) else (comment[0], comment[1]) 485 for comment in klass.COMMENTS 486 ), 487 "{#": "#}", # Ensure Jinja comments are tokenized correctly in all dialects 488 } 489 490 klass._KEYWORD_TRIE = new_trie( 491 key.upper() 492 for key in ( 493 *klass.KEYWORDS, 494 *klass._COMMENTS, 495 *klass._QUOTES, 496 *klass._FORMAT_STRINGS, 497 ) 498 if " " in key or any(single in key for single in klass.SINGLE_TOKENS) 499 ) 500 501 if USE_RS_TOKENIZER: 502 settings = RsTokenizerSettings( 503 white_space={k: _TOKEN_TYPE_TO_INDEX[v] for k, v in klass.WHITE_SPACE.items()}, 504 single_tokens={k: _TOKEN_TYPE_TO_INDEX[v] for k, v in klass.SINGLE_TOKENS.items()}, 505 keywords={k: _TOKEN_TYPE_TO_INDEX[v] for k, v in klass.KEYWORDS.items()}, 506 numeric_literals=klass.NUMERIC_LITERALS, 507 identifiers=klass._IDENTIFIERS, 508 identifier_escapes=klass._IDENTIFIER_ESCAPES, 509 string_escapes=klass._STRING_ESCAPES, 510 quotes=klass._QUOTES, 511 format_strings={ 512 k: (v1, _TOKEN_TYPE_TO_INDEX[v2]) 513 for k, (v1, v2) in klass._FORMAT_STRINGS.items() 514 }, 515 has_bit_strings=bool(klass.BIT_STRINGS), 516 has_hex_strings=bool(klass.HEX_STRINGS), 517 comments=klass._COMMENTS, 518 var_single_tokens=klass.VAR_SINGLE_TOKENS, 519 commands={_TOKEN_TYPE_TO_INDEX[v] for v in klass.COMMANDS}, 520 command_prefix_tokens={ 521 _TOKEN_TYPE_TO_INDEX[v] for v in klass.COMMAND_PREFIX_TOKENS 522 }, 523 heredoc_tag_is_identifier=klass.HEREDOC_TAG_IS_IDENTIFIER, 524 ) 525 token_types = RsTokenTypeSettings( 526 bit_string=_TOKEN_TYPE_TO_INDEX[TokenType.BIT_STRING], 527 break_=_TOKEN_TYPE_TO_INDEX[TokenType.BREAK], 528 dcolon=_TOKEN_TYPE_TO_INDEX[TokenType.DCOLON], 529 heredoc_string=_TOKEN_TYPE_TO_INDEX[TokenType.HEREDOC_STRING], 530 raw_string=_TOKEN_TYPE_TO_INDEX[TokenType.RAW_STRING], 531 hex_string=_TOKEN_TYPE_TO_INDEX[TokenType.HEX_STRING], 532 identifier=_TOKEN_TYPE_TO_INDEX[TokenType.IDENTIFIER], 533 number=_TOKEN_TYPE_TO_INDEX[TokenType.NUMBER], 534 parameter=_TOKEN_TYPE_TO_INDEX[TokenType.PARAMETER], 535 semicolon=_TOKEN_TYPE_TO_INDEX[TokenType.SEMICOLON], 536 string=_TOKEN_TYPE_TO_INDEX[TokenType.STRING], 537 var=_TOKEN_TYPE_TO_INDEX[TokenType.VAR], 538 heredoc_string_alternative=_TOKEN_TYPE_TO_INDEX[klass.HEREDOC_STRING_ALTERNATIVE], 539 ) 540 klass._RS_TOKENIZER = RsTokenizer(settings, token_types) 541 else: 542 klass._RS_TOKENIZER = None 543 544 return klass 545 546 547class Tokenizer(metaclass=_Tokenizer): 548 SINGLE_TOKENS = { 549 "(": TokenType.L_PAREN, 550 ")": TokenType.R_PAREN, 551 "[": TokenType.L_BRACKET, 552 "]": TokenType.R_BRACKET, 553 "{": TokenType.L_BRACE, 554 "}": TokenType.R_BRACE, 555 "&": TokenType.AMP, 556 "^": TokenType.CARET, 557 ":": TokenType.COLON, 558 ",": TokenType.COMMA, 559 ".": TokenType.DOT, 560 "-": TokenType.DASH, 561 "=": TokenType.EQ, 562 ">": TokenType.GT, 563 "<": TokenType.LT, 564 "%": TokenType.MOD, 565 "!": TokenType.NOT, 566 "|": TokenType.PIPE, 567 "+": TokenType.PLUS, 568 ";": TokenType.SEMICOLON, 569 "/": TokenType.SLASH, 570 "\\": TokenType.BACKSLASH, 571 "*": TokenType.STAR, 572 "~": TokenType.TILDA, 573 "?": TokenType.PLACEHOLDER, 574 "@": TokenType.PARAMETER, 575 "#": TokenType.HASH, 576 # Used for breaking a var like x'y' but nothing else the token type doesn't matter 577 "'": TokenType.UNKNOWN, 578 "`": TokenType.UNKNOWN, 579 '"': TokenType.UNKNOWN, 580 } 581 582 BIT_STRINGS: t.List[str | t.Tuple[str, str]] = [] 583 BYTE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 584 HEX_STRINGS: t.List[str | t.Tuple[str, str]] = [] 585 RAW_STRINGS: t.List[str | t.Tuple[str, str]] = [] 586 HEREDOC_STRINGS: t.List[str | t.Tuple[str, str]] = [] 587 UNICODE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 588 IDENTIFIERS: t.List[str | t.Tuple[str, str]] = ['"'] 589 IDENTIFIER_ESCAPES = ['"'] 590 QUOTES: t.List[t.Tuple[str, str] | str] = ["'"] 591 STRING_ESCAPES = ["'"] 592 VAR_SINGLE_TOKENS: t.Set[str] = set() 593 594 # Whether the heredoc tags follow the same lexical rules as unquoted identifiers 595 HEREDOC_TAG_IS_IDENTIFIER = False 596 597 # Token that we'll generate as a fallback if the heredoc prefix doesn't correspond to a heredoc 598 HEREDOC_STRING_ALTERNATIVE = TokenType.VAR 599 600 # Autofilled 601 _COMMENTS: t.Dict[str, str] = {} 602 _FORMAT_STRINGS: t.Dict[str, t.Tuple[str, TokenType]] = {} 603 _IDENTIFIERS: t.Dict[str, str] = {} 604 _IDENTIFIER_ESCAPES: t.Set[str] = set() 605 _QUOTES: t.Dict[str, str] = {} 606 _STRING_ESCAPES: t.Set[str] = set() 607 _KEYWORD_TRIE: t.Dict = {} 608 _RS_TOKENIZER: t.Optional[t.Any] = None 609 610 KEYWORDS: t.Dict[str, TokenType] = { 611 **{f"{{%{postfix}": TokenType.BLOCK_START for postfix in ("", "+", "-")}, 612 **{f"{prefix}%}}": TokenType.BLOCK_END for prefix in ("", "+", "-")}, 613 **{f"{{{{{postfix}": TokenType.BLOCK_START for postfix in ("+", "-")}, 614 **{f"{prefix}}}}}": TokenType.BLOCK_END for prefix in ("+", "-")}, 615 "/*+": TokenType.HINT, 616 "==": TokenType.EQ, 617 "::": TokenType.DCOLON, 618 "||": TokenType.DPIPE, 619 ">=": TokenType.GTE, 620 "<=": TokenType.LTE, 621 "<>": TokenType.NEQ, 622 "!=": TokenType.NEQ, 623 ":=": TokenType.COLON_EQ, 624 "<=>": TokenType.NULLSAFE_EQ, 625 "->": TokenType.ARROW, 626 "->>": TokenType.DARROW, 627 "=>": TokenType.FARROW, 628 "#>": TokenType.HASH_ARROW, 629 "#>>": TokenType.DHASH_ARROW, 630 "<->": TokenType.LR_ARROW, 631 "&&": TokenType.DAMP, 632 "??": TokenType.DQMARK, 633 "ALL": TokenType.ALL, 634 "ALWAYS": TokenType.ALWAYS, 635 "AND": TokenType.AND, 636 "ANTI": TokenType.ANTI, 637 "ANY": TokenType.ANY, 638 "ASC": TokenType.ASC, 639 "AS": TokenType.ALIAS, 640 "ASOF": TokenType.ASOF, 641 "AUTOINCREMENT": TokenType.AUTO_INCREMENT, 642 "AUTO_INCREMENT": TokenType.AUTO_INCREMENT, 643 "BEGIN": TokenType.BEGIN, 644 "BETWEEN": TokenType.BETWEEN, 645 "CACHE": TokenType.CACHE, 646 "UNCACHE": TokenType.UNCACHE, 647 "CASE": TokenType.CASE, 648 "CHARACTER SET": TokenType.CHARACTER_SET, 649 "CLUSTER BY": TokenType.CLUSTER_BY, 650 "COLLATE": TokenType.COLLATE, 651 "COLUMN": TokenType.COLUMN, 652 "COMMIT": TokenType.COMMIT, 653 "CONNECT BY": TokenType.CONNECT_BY, 654 "CONSTRAINT": TokenType.CONSTRAINT, 655 "COPY": TokenType.COPY, 656 "CREATE": TokenType.CREATE, 657 "CROSS": TokenType.CROSS, 658 "CUBE": TokenType.CUBE, 659 "CURRENT_DATE": TokenType.CURRENT_DATE, 660 "CURRENT_TIME": TokenType.CURRENT_TIME, 661 "CURRENT_TIMESTAMP": TokenType.CURRENT_TIMESTAMP, 662 "CURRENT_USER": TokenType.CURRENT_USER, 663 "DATABASE": TokenType.DATABASE, 664 "DEFAULT": TokenType.DEFAULT, 665 "DELETE": TokenType.DELETE, 666 "DESC": TokenType.DESC, 667 "DESCRIBE": TokenType.DESCRIBE, 668 "DISTINCT": TokenType.DISTINCT, 669 "DISTRIBUTE BY": TokenType.DISTRIBUTE_BY, 670 "DIV": TokenType.DIV, 671 "DROP": TokenType.DROP, 672 "ELSE": TokenType.ELSE, 673 "END": TokenType.END, 674 "ENUM": TokenType.ENUM, 675 "ESCAPE": TokenType.ESCAPE, 676 "EXCEPT": TokenType.EXCEPT, 677 "EXECUTE": TokenType.EXECUTE, 678 "EXISTS": TokenType.EXISTS, 679 "FALSE": TokenType.FALSE, 680 "FETCH": TokenType.FETCH, 681 "FILTER": TokenType.FILTER, 682 "FIRST": TokenType.FIRST, 683 "FULL": TokenType.FULL, 684 "FUNCTION": TokenType.FUNCTION, 685 "FOR": TokenType.FOR, 686 "FOREIGN KEY": TokenType.FOREIGN_KEY, 687 "FORMAT": TokenType.FORMAT, 688 "FROM": TokenType.FROM, 689 "GEOGRAPHY": TokenType.GEOGRAPHY, 690 "GEOMETRY": TokenType.GEOMETRY, 691 "GLOB": TokenType.GLOB, 692 "GROUP BY": TokenType.GROUP_BY, 693 "GROUPING SETS": TokenType.GROUPING_SETS, 694 "HAVING": TokenType.HAVING, 695 "ILIKE": TokenType.ILIKE, 696 "IN": TokenType.IN, 697 "INDEX": TokenType.INDEX, 698 "INET": TokenType.INET, 699 "INNER": TokenType.INNER, 700 "INSERT": TokenType.INSERT, 701 "INTERVAL": TokenType.INTERVAL, 702 "INTERSECT": TokenType.INTERSECT, 703 "INTO": TokenType.INTO, 704 "IS": TokenType.IS, 705 "ISNULL": TokenType.ISNULL, 706 "JOIN": TokenType.JOIN, 707 "KEEP": TokenType.KEEP, 708 "KILL": TokenType.KILL, 709 "LATERAL": TokenType.LATERAL, 710 "LEFT": TokenType.LEFT, 711 "LIKE": TokenType.LIKE, 712 "LIMIT": TokenType.LIMIT, 713 "LOAD": TokenType.LOAD, 714 "LOCK": TokenType.LOCK, 715 "MERGE": TokenType.MERGE, 716 "NATURAL": TokenType.NATURAL, 717 "NEXT": TokenType.NEXT, 718 "NOT": TokenType.NOT, 719 "NOTNULL": TokenType.NOTNULL, 720 "NULL": TokenType.NULL, 721 "OBJECT": TokenType.OBJECT, 722 "OFFSET": TokenType.OFFSET, 723 "ON": TokenType.ON, 724 "OR": TokenType.OR, 725 "XOR": TokenType.XOR, 726 "ORDER BY": TokenType.ORDER_BY, 727 "ORDINALITY": TokenType.ORDINALITY, 728 "OUTER": TokenType.OUTER, 729 "OVER": TokenType.OVER, 730 "OVERLAPS": TokenType.OVERLAPS, 731 "OVERWRITE": TokenType.OVERWRITE, 732 "PARTITION": TokenType.PARTITION, 733 "PARTITION BY": TokenType.PARTITION_BY, 734 "PARTITIONED BY": TokenType.PARTITION_BY, 735 "PARTITIONED_BY": TokenType.PARTITION_BY, 736 "PERCENT": TokenType.PERCENT, 737 "PIVOT": TokenType.PIVOT, 738 "PRAGMA": TokenType.PRAGMA, 739 "PRIMARY KEY": TokenType.PRIMARY_KEY, 740 "PROCEDURE": TokenType.PROCEDURE, 741 "QUALIFY": TokenType.QUALIFY, 742 "RANGE": TokenType.RANGE, 743 "RECURSIVE": TokenType.RECURSIVE, 744 "REGEXP": TokenType.RLIKE, 745 "REPLACE": TokenType.REPLACE, 746 "RETURNING": TokenType.RETURNING, 747 "REFERENCES": TokenType.REFERENCES, 748 "RIGHT": TokenType.RIGHT, 749 "RLIKE": TokenType.RLIKE, 750 "ROLLBACK": TokenType.ROLLBACK, 751 "ROLLUP": TokenType.ROLLUP, 752 "ROW": TokenType.ROW, 753 "ROWS": TokenType.ROWS, 754 "SCHEMA": TokenType.SCHEMA, 755 "SELECT": TokenType.SELECT, 756 "SEMI": TokenType.SEMI, 757 "SET": TokenType.SET, 758 "SETTINGS": TokenType.SETTINGS, 759 "SHOW": TokenType.SHOW, 760 "SIMILAR TO": TokenType.SIMILAR_TO, 761 "SOME": TokenType.SOME, 762 "SORT BY": TokenType.SORT_BY, 763 "START WITH": TokenType.START_WITH, 764 "TABLE": TokenType.TABLE, 765 "TABLESAMPLE": TokenType.TABLE_SAMPLE, 766 "TEMP": TokenType.TEMPORARY, 767 "TEMPORARY": TokenType.TEMPORARY, 768 "THEN": TokenType.THEN, 769 "TRUE": TokenType.TRUE, 770 "TRUNCATE": TokenType.TRUNCATE, 771 "UNION": TokenType.UNION, 772 "UNKNOWN": TokenType.UNKNOWN, 773 "UNNEST": TokenType.UNNEST, 774 "UNPIVOT": TokenType.UNPIVOT, 775 "UPDATE": TokenType.UPDATE, 776 "USE": TokenType.USE, 777 "USING": TokenType.USING, 778 "UUID": TokenType.UUID, 779 "VALUES": TokenType.VALUES, 780 "VIEW": TokenType.VIEW, 781 "VOLATILE": TokenType.VOLATILE, 782 "WHEN": TokenType.WHEN, 783 "WHERE": TokenType.WHERE, 784 "WINDOW": TokenType.WINDOW, 785 "WITH": TokenType.WITH, 786 "APPLY": TokenType.APPLY, 787 "ARRAY": TokenType.ARRAY, 788 "BIT": TokenType.BIT, 789 "BOOL": TokenType.BOOLEAN, 790 "BOOLEAN": TokenType.BOOLEAN, 791 "BYTE": TokenType.TINYINT, 792 "MEDIUMINT": TokenType.MEDIUMINT, 793 "INT1": TokenType.TINYINT, 794 "TINYINT": TokenType.TINYINT, 795 "INT16": TokenType.SMALLINT, 796 "SHORT": TokenType.SMALLINT, 797 "SMALLINT": TokenType.SMALLINT, 798 "INT128": TokenType.INT128, 799 "HUGEINT": TokenType.INT128, 800 "INT2": TokenType.SMALLINT, 801 "INTEGER": TokenType.INT, 802 "INT": TokenType.INT, 803 "INT4": TokenType.INT, 804 "INT32": TokenType.INT, 805 "INT64": TokenType.BIGINT, 806 "LONG": TokenType.BIGINT, 807 "BIGINT": TokenType.BIGINT, 808 "INT8": TokenType.TINYINT, 809 "UINT": TokenType.UINT, 810 "DEC": TokenType.DECIMAL, 811 "DECIMAL": TokenType.DECIMAL, 812 "BIGDECIMAL": TokenType.BIGDECIMAL, 813 "BIGNUMERIC": TokenType.BIGDECIMAL, 814 "MAP": TokenType.MAP, 815 "NULLABLE": TokenType.NULLABLE, 816 "NUMBER": TokenType.DECIMAL, 817 "NUMERIC": TokenType.DECIMAL, 818 "FIXED": TokenType.DECIMAL, 819 "REAL": TokenType.FLOAT, 820 "FLOAT": TokenType.FLOAT, 821 "FLOAT4": TokenType.FLOAT, 822 "FLOAT8": TokenType.DOUBLE, 823 "DOUBLE": TokenType.DOUBLE, 824 "DOUBLE PRECISION": TokenType.DOUBLE, 825 "JSON": TokenType.JSON, 826 "CHAR": TokenType.CHAR, 827 "CHARACTER": TokenType.CHAR, 828 "NCHAR": TokenType.NCHAR, 829 "VARCHAR": TokenType.VARCHAR, 830 "VARCHAR2": TokenType.VARCHAR, 831 "NVARCHAR": TokenType.NVARCHAR, 832 "NVARCHAR2": TokenType.NVARCHAR, 833 "BPCHAR": TokenType.BPCHAR, 834 "STR": TokenType.TEXT, 835 "STRING": TokenType.TEXT, 836 "TEXT": TokenType.TEXT, 837 "LONGTEXT": TokenType.LONGTEXT, 838 "MEDIUMTEXT": TokenType.MEDIUMTEXT, 839 "TINYTEXT": TokenType.TINYTEXT, 840 "CLOB": TokenType.TEXT, 841 "LONGVARCHAR": TokenType.TEXT, 842 "BINARY": TokenType.BINARY, 843 "BLOB": TokenType.VARBINARY, 844 "LONGBLOB": TokenType.LONGBLOB, 845 "MEDIUMBLOB": TokenType.MEDIUMBLOB, 846 "TINYBLOB": TokenType.TINYBLOB, 847 "BYTEA": TokenType.VARBINARY, 848 "VARBINARY": TokenType.VARBINARY, 849 "TIME": TokenType.TIME, 850 "TIMETZ": TokenType.TIMETZ, 851 "TIMESTAMP": TokenType.TIMESTAMP, 852 "TIMESTAMPTZ": TokenType.TIMESTAMPTZ, 853 "TIMESTAMPLTZ": TokenType.TIMESTAMPLTZ, 854 "TIMESTAMP_LTZ": TokenType.TIMESTAMPLTZ, 855 "TIMESTAMPNTZ": TokenType.TIMESTAMPNTZ, 856 "TIMESTAMP_NTZ": TokenType.TIMESTAMPNTZ, 857 "DATE": TokenType.DATE, 858 "DATETIME": TokenType.DATETIME, 859 "INT4RANGE": TokenType.INT4RANGE, 860 "INT4MULTIRANGE": TokenType.INT4MULTIRANGE, 861 "INT8RANGE": TokenType.INT8RANGE, 862 "INT8MULTIRANGE": TokenType.INT8MULTIRANGE, 863 "NUMRANGE": TokenType.NUMRANGE, 864 "NUMMULTIRANGE": TokenType.NUMMULTIRANGE, 865 "TSRANGE": TokenType.TSRANGE, 866 "TSMULTIRANGE": TokenType.TSMULTIRANGE, 867 "TSTZRANGE": TokenType.TSTZRANGE, 868 "TSTZMULTIRANGE": TokenType.TSTZMULTIRANGE, 869 "DATERANGE": TokenType.DATERANGE, 870 "DATEMULTIRANGE": TokenType.DATEMULTIRANGE, 871 "UNIQUE": TokenType.UNIQUE, 872 "STRUCT": TokenType.STRUCT, 873 "SEQUENCE": TokenType.SEQUENCE, 874 "VARIANT": TokenType.VARIANT, 875 "ALTER": TokenType.ALTER, 876 "ANALYZE": TokenType.COMMAND, 877 "CALL": TokenType.COMMAND, 878 "COMMENT": TokenType.COMMENT, 879 "EXPLAIN": TokenType.COMMAND, 880 "GRANT": TokenType.COMMAND, 881 "OPTIMIZE": TokenType.COMMAND, 882 "PREPARE": TokenType.COMMAND, 883 "VACUUM": TokenType.COMMAND, 884 "USER-DEFINED": TokenType.USERDEFINED, 885 "FOR VERSION": TokenType.VERSION_SNAPSHOT, 886 "FOR TIMESTAMP": TokenType.TIMESTAMP_SNAPSHOT, 887 } 888 889 WHITE_SPACE: t.Dict[t.Optional[str], TokenType] = { 890 " ": TokenType.SPACE, 891 "\t": TokenType.SPACE, 892 "\n": TokenType.BREAK, 893 "\r": TokenType.BREAK, 894 } 895 896 COMMANDS = { 897 TokenType.COMMAND, 898 TokenType.EXECUTE, 899 TokenType.FETCH, 900 TokenType.SHOW, 901 } 902 903 COMMAND_PREFIX_TOKENS = {TokenType.SEMICOLON, TokenType.BEGIN} 904 905 # Handle numeric literals like in hive (3L = BIGINT) 906 NUMERIC_LITERALS: t.Dict[str, str] = {} 907 908 COMMENTS = ["--", ("/*", "*/")] 909 910 __slots__ = ( 911 "sql", 912 "size", 913 "tokens", 914 "dialect", 915 "_start", 916 "_current", 917 "_line", 918 "_col", 919 "_comments", 920 "_char", 921 "_end", 922 "_peek", 923 "_prev_token_line", 924 "_rs_dialect_settings", 925 ) 926 927 def __init__(self, dialect: DialectType = None) -> None: 928 from sqlglot.dialects import Dialect 929 930 self.dialect = Dialect.get_or_raise(dialect) 931 932 if USE_RS_TOKENIZER: 933 self._rs_dialect_settings = RsTokenizerDialectSettings( 934 unescaped_sequences=self.dialect.UNESCAPED_SEQUENCES, 935 identifiers_can_start_with_digit=self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT, 936 ) 937 938 self.reset() 939 940 def reset(self) -> None: 941 self.sql = "" 942 self.size = 0 943 self.tokens: t.List[Token] = [] 944 self._start = 0 945 self._current = 0 946 self._line = 1 947 self._col = 0 948 self._comments: t.List[str] = [] 949 950 self._char = "" 951 self._end = False 952 self._peek = "" 953 self._prev_token_line = -1 954 955 def tokenize(self, sql: str) -> t.List[Token]: 956 """Returns a list of tokens corresponding to the SQL string `sql`.""" 957 if USE_RS_TOKENIZER: 958 return self.tokenize_rs(sql) 959 960 self.reset() 961 self.sql = sql 962 self.size = len(sql) 963 964 try: 965 self._scan() 966 except Exception as e: 967 start = max(self._current - 50, 0) 968 end = min(self._current + 50, self.size - 1) 969 context = self.sql[start:end] 970 raise TokenError(f"Error tokenizing '{context}'") from e 971 972 return self.tokens 973 974 def _scan(self, until: t.Optional[t.Callable] = None) -> None: 975 while self.size and not self._end: 976 current = self._current 977 978 # Skip spaces here rather than iteratively calling advance() for performance reasons 979 while current < self.size: 980 char = self.sql[current] 981 982 if char.isspace() and (char == " " or char == "\t"): 983 current += 1 984 else: 985 break 986 987 offset = current - self._current if current > self._current else 1 988 989 self._start = current 990 self._advance(offset) 991 992 if not self._char.isspace(): 993 if self._char.isdigit(): 994 self._scan_number() 995 elif self._char in self._IDENTIFIERS: 996 self._scan_identifier(self._IDENTIFIERS[self._char]) 997 else: 998 self._scan_keywords() 999 1000 if until and until(): 1001 break 1002 1003 if self.tokens and self._comments: 1004 self.tokens[-1].comments.extend(self._comments) 1005 1006 def _chars(self, size: int) -> str: 1007 if size == 1: 1008 return self._char 1009 1010 start = self._current - 1 1011 end = start + size 1012 1013 return self.sql[start:end] if end <= self.size else "" 1014 1015 def _advance(self, i: int = 1, alnum: bool = False) -> None: 1016 if self.WHITE_SPACE.get(self._char) is TokenType.BREAK: 1017 # Ensures we don't count an extra line if we get a \r\n line break sequence 1018 if not (self._char == "\r" and self._peek == "\n"): 1019 self._col = 1 1020 self._line += 1 1021 else: 1022 self._col += i 1023 1024 self._current += i 1025 self._end = self._current >= self.size 1026 self._char = self.sql[self._current - 1] 1027 self._peek = "" if self._end else self.sql[self._current] 1028 1029 if alnum and self._char.isalnum(): 1030 # Here we use local variables instead of attributes for better performance 1031 _col = self._col 1032 _current = self._current 1033 _end = self._end 1034 _peek = self._peek 1035 1036 while _peek.isalnum(): 1037 _col += 1 1038 _current += 1 1039 _end = _current >= self.size 1040 _peek = "" if _end else self.sql[_current] 1041 1042 self._col = _col 1043 self._current = _current 1044 self._end = _end 1045 self._peek = _peek 1046 self._char = self.sql[_current - 1] 1047 1048 @property 1049 def _text(self) -> str: 1050 return self.sql[self._start : self._current] 1051 1052 def _add(self, token_type: TokenType, text: t.Optional[str] = None) -> None: 1053 self._prev_token_line = self._line 1054 1055 if self._comments and token_type == TokenType.SEMICOLON and self.tokens: 1056 self.tokens[-1].comments.extend(self._comments) 1057 self._comments = [] 1058 1059 self.tokens.append( 1060 Token( 1061 token_type, 1062 text=self._text if text is None else text, 1063 line=self._line, 1064 col=self._col, 1065 start=self._start, 1066 end=self._current - 1, 1067 comments=self._comments, 1068 ) 1069 ) 1070 self._comments = [] 1071 1072 # If we have either a semicolon or a begin token before the command's token, we'll parse 1073 # whatever follows the command's token as a string 1074 if ( 1075 token_type in self.COMMANDS 1076 and self._peek != ";" 1077 and (len(self.tokens) == 1 or self.tokens[-2].token_type in self.COMMAND_PREFIX_TOKENS) 1078 ): 1079 start = self._current 1080 tokens = len(self.tokens) 1081 self._scan(lambda: self._peek == ";") 1082 self.tokens = self.tokens[:tokens] 1083 text = self.sql[start : self._current].strip() 1084 if text: 1085 self._add(TokenType.STRING, text) 1086 1087 def _scan_keywords(self) -> None: 1088 size = 0 1089 word = None 1090 chars = self._text 1091 char = chars 1092 prev_space = False 1093 skip = False 1094 trie = self._KEYWORD_TRIE 1095 single_token = char in self.SINGLE_TOKENS 1096 1097 while chars: 1098 if skip: 1099 result = TrieResult.PREFIX 1100 else: 1101 result, trie = in_trie(trie, char.upper()) 1102 1103 if result == TrieResult.FAILED: 1104 break 1105 if result == TrieResult.EXISTS: 1106 word = chars 1107 1108 end = self._current + size 1109 size += 1 1110 1111 if end < self.size: 1112 char = self.sql[end] 1113 single_token = single_token or char in self.SINGLE_TOKENS 1114 is_space = char.isspace() 1115 1116 if not is_space or not prev_space: 1117 if is_space: 1118 char = " " 1119 chars += char 1120 prev_space = is_space 1121 skip = False 1122 else: 1123 skip = True 1124 else: 1125 char = "" 1126 break 1127 1128 if word: 1129 if self._scan_string(word): 1130 return 1131 if self._scan_comment(word): 1132 return 1133 if prev_space or single_token or not char: 1134 self._advance(size - 1) 1135 word = word.upper() 1136 self._add(self.KEYWORDS[word], text=word) 1137 return 1138 1139 if self._char in self.SINGLE_TOKENS: 1140 self._add(self.SINGLE_TOKENS[self._char], text=self._char) 1141 return 1142 1143 self._scan_var() 1144 1145 def _scan_comment(self, comment_start: str) -> bool: 1146 if comment_start not in self._COMMENTS: 1147 return False 1148 1149 comment_start_line = self._line 1150 comment_start_size = len(comment_start) 1151 comment_end = self._COMMENTS[comment_start] 1152 1153 if comment_end: 1154 # Skip the comment's start delimiter 1155 self._advance(comment_start_size) 1156 1157 comment_end_size = len(comment_end) 1158 while not self._end and self._chars(comment_end_size) != comment_end: 1159 self._advance(alnum=True) 1160 1161 self._comments.append(self._text[comment_start_size : -comment_end_size + 1]) 1162 self._advance(comment_end_size - 1) 1163 else: 1164 while not self._end and self.WHITE_SPACE.get(self._peek) is not TokenType.BREAK: 1165 self._advance(alnum=True) 1166 self._comments.append(self._text[comment_start_size:]) 1167 1168 # Leading comment is attached to the succeeding token, whilst trailing comment to the preceding. 1169 # Multiple consecutive comments are preserved by appending them to the current comments list. 1170 if comment_start_line == self._prev_token_line: 1171 self.tokens[-1].comments.extend(self._comments) 1172 self._comments = [] 1173 self._prev_token_line = self._line 1174 1175 return True 1176 1177 def _scan_number(self) -> None: 1178 if self._char == "0": 1179 peek = self._peek.upper() 1180 if peek == "B": 1181 return self._scan_bits() if self.BIT_STRINGS else self._add(TokenType.NUMBER) 1182 elif peek == "X": 1183 return self._scan_hex() if self.HEX_STRINGS else self._add(TokenType.NUMBER) 1184 1185 decimal = False 1186 scientific = 0 1187 1188 while True: 1189 if self._peek.isdigit(): 1190 self._advance() 1191 elif self._peek == "." and not decimal: 1192 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER: 1193 return self._add(TokenType.NUMBER) 1194 decimal = True 1195 self._advance() 1196 elif self._peek in ("-", "+") and scientific == 1: 1197 scientific += 1 1198 self._advance() 1199 elif self._peek.upper() == "E" and not scientific: 1200 scientific += 1 1201 self._advance() 1202 elif self._peek.isidentifier(): 1203 number_text = self._text 1204 literal = "" 1205 1206 while self._peek.strip() and self._peek not in self.SINGLE_TOKENS: 1207 literal += self._peek 1208 self._advance() 1209 1210 token_type = self.KEYWORDS.get(self.NUMERIC_LITERALS.get(literal.upper(), "")) 1211 1212 if token_type: 1213 self._add(TokenType.NUMBER, number_text) 1214 self._add(TokenType.DCOLON, "::") 1215 return self._add(token_type, literal) 1216 elif self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT: 1217 return self._add(TokenType.VAR) 1218 1219 self._advance(-len(literal)) 1220 return self._add(TokenType.NUMBER, number_text) 1221 else: 1222 return self._add(TokenType.NUMBER) 1223 1224 def _scan_bits(self) -> None: 1225 self._advance() 1226 value = self._extract_value() 1227 try: 1228 # If `value` can't be converted to a binary, fallback to tokenizing it as an identifier 1229 int(value, 2) 1230 self._add(TokenType.BIT_STRING, value[2:]) # Drop the 0b 1231 except ValueError: 1232 self._add(TokenType.IDENTIFIER) 1233 1234 def _scan_hex(self) -> None: 1235 self._advance() 1236 value = self._extract_value() 1237 try: 1238 # If `value` can't be converted to a hex, fallback to tokenizing it as an identifier 1239 int(value, 16) 1240 self._add(TokenType.HEX_STRING, value[2:]) # Drop the 0x 1241 except ValueError: 1242 self._add(TokenType.IDENTIFIER) 1243 1244 def _extract_value(self) -> str: 1245 while True: 1246 char = self._peek.strip() 1247 if char and char not in self.SINGLE_TOKENS: 1248 self._advance(alnum=True) 1249 else: 1250 break 1251 1252 return self._text 1253 1254 def _scan_string(self, start: str) -> bool: 1255 base = None 1256 token_type = TokenType.STRING 1257 1258 if start in self._QUOTES: 1259 end = self._QUOTES[start] 1260 elif start in self._FORMAT_STRINGS: 1261 end, token_type = self._FORMAT_STRINGS[start] 1262 1263 if token_type == TokenType.HEX_STRING: 1264 base = 16 1265 elif token_type == TokenType.BIT_STRING: 1266 base = 2 1267 elif token_type == TokenType.HEREDOC_STRING: 1268 if ( 1269 self.HEREDOC_TAG_IS_IDENTIFIER 1270 and not self._peek.isidentifier() 1271 and not self._peek == end 1272 ): 1273 if self.HEREDOC_STRING_ALTERNATIVE != token_type.VAR: 1274 self._add(self.HEREDOC_STRING_ALTERNATIVE) 1275 else: 1276 self._scan_var() 1277 1278 return True 1279 1280 self._advance() 1281 1282 if self._char == end: 1283 tag = "" 1284 else: 1285 tag = self._extract_string( 1286 end, 1287 unescape_sequences=False, 1288 raise_unmatched=not self.HEREDOC_TAG_IS_IDENTIFIER, 1289 ) 1290 1291 if self._end and tag and self.HEREDOC_TAG_IS_IDENTIFIER: 1292 self._advance(-len(tag)) 1293 self._add(self.HEREDOC_STRING_ALTERNATIVE) 1294 return True 1295 1296 end = f"{start}{tag}{end}" 1297 else: 1298 return False 1299 1300 self._advance(len(start)) 1301 text = self._extract_string(end, unescape_sequences=token_type != TokenType.RAW_STRING) 1302 1303 if base: 1304 try: 1305 int(text, base) 1306 except Exception: 1307 raise TokenError( 1308 f"Numeric string contains invalid characters from {self._line}:{self._start}" 1309 ) 1310 1311 self._add(token_type, text) 1312 return True 1313 1314 def _scan_identifier(self, identifier_end: str) -> None: 1315 self._advance() 1316 text = self._extract_string(identifier_end, escapes=self._IDENTIFIER_ESCAPES) 1317 self._add(TokenType.IDENTIFIER, text) 1318 1319 def _scan_var(self) -> None: 1320 while True: 1321 char = self._peek.strip() 1322 if char and (char in self.VAR_SINGLE_TOKENS or char not in self.SINGLE_TOKENS): 1323 self._advance(alnum=True) 1324 else: 1325 break 1326 1327 self._add( 1328 TokenType.VAR 1329 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER 1330 else self.KEYWORDS.get(self._text.upper(), TokenType.VAR) 1331 ) 1332 1333 def _extract_string( 1334 self, 1335 delimiter: str, 1336 escapes: t.Optional[t.Set[str]] = None, 1337 unescape_sequences: bool = True, 1338 raise_unmatched: bool = True, 1339 ) -> str: 1340 text = "" 1341 delim_size = len(delimiter) 1342 escapes = self._STRING_ESCAPES if escapes is None else escapes 1343 1344 while True: 1345 if ( 1346 unescape_sequences 1347 and self.dialect.UNESCAPED_SEQUENCES 1348 and self._peek 1349 and self._char in self.STRING_ESCAPES 1350 ): 1351 unescaped_sequence = self.dialect.UNESCAPED_SEQUENCES.get(self._char + self._peek) 1352 if unescaped_sequence: 1353 self._advance(2) 1354 text += unescaped_sequence 1355 continue 1356 if ( 1357 self._char in escapes 1358 and (self._peek == delimiter or self._peek in escapes) 1359 and (self._char not in self._QUOTES or self._char == self._peek) 1360 ): 1361 if self._peek == delimiter: 1362 text += self._peek 1363 else: 1364 text += self._char + self._peek 1365 1366 if self._current + 1 < self.size: 1367 self._advance(2) 1368 else: 1369 raise TokenError(f"Missing {delimiter} from {self._line}:{self._current}") 1370 else: 1371 if self._chars(delim_size) == delimiter: 1372 if delim_size > 1: 1373 self._advance(delim_size - 1) 1374 break 1375 1376 if self._end: 1377 if not raise_unmatched: 1378 return text + self._char 1379 1380 raise TokenError(f"Missing {delimiter} from {self._line}:{self._start}") 1381 1382 current = self._current - 1 1383 self._advance(alnum=True) 1384 text += self.sql[current : self._current - 1] 1385 1386 return text 1387 1388 def tokenize_rs(self, sql: str) -> t.List[Token]: 1389 if not self._RS_TOKENIZER: 1390 raise SqlglotError("Rust tokenizer is not available") 1391 1392 try: 1393 tokens = self._RS_TOKENIZER.tokenize(sql, self._rs_dialect_settings) 1394 for token in tokens: 1395 token.token_type = _ALL_TOKEN_TYPES[token.token_type_index] 1396 return tokens 1397 except Exception as e: 1398 raise TokenError(str(e))
29class TokenType(AutoName): 30 L_PAREN = auto() 31 R_PAREN = auto() 32 L_BRACKET = auto() 33 R_BRACKET = auto() 34 L_BRACE = auto() 35 R_BRACE = auto() 36 COMMA = auto() 37 DOT = auto() 38 DASH = auto() 39 PLUS = auto() 40 COLON = auto() 41 DCOLON = auto() 42 DQMARK = auto() 43 SEMICOLON = auto() 44 STAR = auto() 45 BACKSLASH = auto() 46 SLASH = auto() 47 LT = auto() 48 LTE = auto() 49 GT = auto() 50 GTE = auto() 51 NOT = auto() 52 EQ = auto() 53 NEQ = auto() 54 NULLSAFE_EQ = auto() 55 COLON_EQ = auto() 56 AND = auto() 57 OR = auto() 58 AMP = auto() 59 DPIPE = auto() 60 PIPE = auto() 61 PIPE_SLASH = auto() 62 DPIPE_SLASH = auto() 63 CARET = auto() 64 TILDA = auto() 65 ARROW = auto() 66 DARROW = auto() 67 FARROW = auto() 68 HASH = auto() 69 HASH_ARROW = auto() 70 DHASH_ARROW = auto() 71 LR_ARROW = auto() 72 DAT = auto() 73 LT_AT = auto() 74 AT_GT = auto() 75 DOLLAR = auto() 76 PARAMETER = auto() 77 SESSION_PARAMETER = auto() 78 DAMP = auto() 79 XOR = auto() 80 DSTAR = auto() 81 82 BLOCK_START = auto() 83 BLOCK_END = auto() 84 85 SPACE = auto() 86 BREAK = auto() 87 88 STRING = auto() 89 NUMBER = auto() 90 IDENTIFIER = auto() 91 DATABASE = auto() 92 COLUMN = auto() 93 COLUMN_DEF = auto() 94 SCHEMA = auto() 95 TABLE = auto() 96 VAR = auto() 97 BIT_STRING = auto() 98 HEX_STRING = auto() 99 BYTE_STRING = auto() 100 NATIONAL_STRING = auto() 101 RAW_STRING = auto() 102 HEREDOC_STRING = auto() 103 UNICODE_STRING = auto() 104 105 # types 106 BIT = auto() 107 BOOLEAN = auto() 108 TINYINT = auto() 109 UTINYINT = auto() 110 SMALLINT = auto() 111 USMALLINT = auto() 112 MEDIUMINT = auto() 113 UMEDIUMINT = auto() 114 INT = auto() 115 UINT = auto() 116 BIGINT = auto() 117 UBIGINT = auto() 118 INT128 = auto() 119 UINT128 = auto() 120 INT256 = auto() 121 UINT256 = auto() 122 FLOAT = auto() 123 DOUBLE = auto() 124 DECIMAL = auto() 125 UDECIMAL = auto() 126 BIGDECIMAL = auto() 127 CHAR = auto() 128 NCHAR = auto() 129 VARCHAR = auto() 130 NVARCHAR = auto() 131 BPCHAR = auto() 132 TEXT = auto() 133 MEDIUMTEXT = auto() 134 LONGTEXT = auto() 135 MEDIUMBLOB = auto() 136 LONGBLOB = auto() 137 TINYBLOB = auto() 138 TINYTEXT = auto() 139 NAME = auto() 140 BINARY = auto() 141 VARBINARY = auto() 142 JSON = auto() 143 JSONB = auto() 144 TIME = auto() 145 TIMETZ = auto() 146 TIMESTAMP = auto() 147 TIMESTAMPTZ = auto() 148 TIMESTAMPLTZ = auto() 149 TIMESTAMPNTZ = auto() 150 TIMESTAMP_S = auto() 151 TIMESTAMP_MS = auto() 152 TIMESTAMP_NS = auto() 153 DATETIME = auto() 154 DATETIME64 = auto() 155 DATE = auto() 156 DATE32 = auto() 157 INT4RANGE = auto() 158 INT4MULTIRANGE = auto() 159 INT8RANGE = auto() 160 INT8MULTIRANGE = auto() 161 NUMRANGE = auto() 162 NUMMULTIRANGE = auto() 163 TSRANGE = auto() 164 TSMULTIRANGE = auto() 165 TSTZRANGE = auto() 166 TSTZMULTIRANGE = auto() 167 DATERANGE = auto() 168 DATEMULTIRANGE = auto() 169 UUID = auto() 170 GEOGRAPHY = auto() 171 NULLABLE = auto() 172 GEOMETRY = auto() 173 HLLSKETCH = auto() 174 HSTORE = auto() 175 SUPER = auto() 176 SERIAL = auto() 177 SMALLSERIAL = auto() 178 BIGSERIAL = auto() 179 XML = auto() 180 YEAR = auto() 181 UNIQUEIDENTIFIER = auto() 182 USERDEFINED = auto() 183 MONEY = auto() 184 SMALLMONEY = auto() 185 ROWVERSION = auto() 186 IMAGE = auto() 187 VARIANT = auto() 188 OBJECT = auto() 189 INET = auto() 190 IPADDRESS = auto() 191 IPPREFIX = auto() 192 IPV4 = auto() 193 IPV6 = auto() 194 ENUM = auto() 195 ENUM8 = auto() 196 ENUM16 = auto() 197 FIXEDSTRING = auto() 198 LOWCARDINALITY = auto() 199 NESTED = auto() 200 AGGREGATEFUNCTION = auto() 201 SIMPLEAGGREGATEFUNCTION = auto() 202 TDIGEST = auto() 203 UNKNOWN = auto() 204 205 # keywords 206 ALIAS = auto() 207 ALTER = auto() 208 ALWAYS = auto() 209 ALL = auto() 210 ANTI = auto() 211 ANY = auto() 212 APPLY = auto() 213 ARRAY = auto() 214 ASC = auto() 215 ASOF = auto() 216 AUTO_INCREMENT = auto() 217 BEGIN = auto() 218 BETWEEN = auto() 219 CACHE = auto() 220 CASE = auto() 221 CHARACTER_SET = auto() 222 CLUSTER_BY = auto() 223 COLLATE = auto() 224 COMMAND = auto() 225 COMMENT = auto() 226 COMMIT = auto() 227 CONNECT_BY = auto() 228 CONSTRAINT = auto() 229 COPY = auto() 230 CREATE = auto() 231 CROSS = auto() 232 CUBE = auto() 233 CURRENT_DATE = auto() 234 CURRENT_DATETIME = auto() 235 CURRENT_TIME = auto() 236 CURRENT_TIMESTAMP = auto() 237 CURRENT_USER = auto() 238 DECLARE = auto() 239 DEFAULT = auto() 240 DELETE = auto() 241 DESC = auto() 242 DESCRIBE = auto() 243 DICTIONARY = auto() 244 DISTINCT = auto() 245 DISTRIBUTE_BY = auto() 246 DIV = auto() 247 DROP = auto() 248 ELSE = auto() 249 END = auto() 250 ESCAPE = auto() 251 EXCEPT = auto() 252 EXECUTE = auto() 253 EXISTS = auto() 254 FALSE = auto() 255 FETCH = auto() 256 FILTER = auto() 257 FINAL = auto() 258 FIRST = auto() 259 FOR = auto() 260 FORCE = auto() 261 FOREIGN_KEY = auto() 262 FORMAT = auto() 263 FROM = auto() 264 FULL = auto() 265 FUNCTION = auto() 266 GLOB = auto() 267 GLOBAL = auto() 268 GROUP_BY = auto() 269 GROUPING_SETS = auto() 270 HAVING = auto() 271 HINT = auto() 272 IGNORE = auto() 273 ILIKE = auto() 274 ILIKE_ANY = auto() 275 IN = auto() 276 INDEX = auto() 277 INNER = auto() 278 INSERT = auto() 279 INTERSECT = auto() 280 INTERVAL = auto() 281 INTO = auto() 282 INTRODUCER = auto() 283 IRLIKE = auto() 284 IS = auto() 285 ISNULL = auto() 286 JOIN = auto() 287 JOIN_MARKER = auto() 288 KEEP = auto() 289 KILL = auto() 290 LANGUAGE = auto() 291 LATERAL = auto() 292 LEFT = auto() 293 LIKE = auto() 294 LIKE_ANY = auto() 295 LIMIT = auto() 296 LOAD = auto() 297 LOCK = auto() 298 MAP = auto() 299 MATCH_CONDITION = auto() 300 MATCH_RECOGNIZE = auto() 301 MEMBER_OF = auto() 302 MERGE = auto() 303 MOD = auto() 304 MODEL = auto() 305 NATURAL = auto() 306 NEXT = auto() 307 NOTNULL = auto() 308 NULL = auto() 309 OBJECT_IDENTIFIER = auto() 310 OFFSET = auto() 311 ON = auto() 312 ONLY = auto() 313 OPERATOR = auto() 314 ORDER_BY = auto() 315 ORDER_SIBLINGS_BY = auto() 316 ORDERED = auto() 317 ORDINALITY = auto() 318 OUTER = auto() 319 OVER = auto() 320 OVERLAPS = auto() 321 OVERWRITE = auto() 322 PARTITION = auto() 323 PARTITION_BY = auto() 324 PERCENT = auto() 325 PIVOT = auto() 326 PLACEHOLDER = auto() 327 POSITIONAL = auto() 328 PRAGMA = auto() 329 PREWHERE = auto() 330 PRIMARY_KEY = auto() 331 PROCEDURE = auto() 332 PROPERTIES = auto() 333 PSEUDO_TYPE = auto() 334 QUALIFY = auto() 335 QUOTE = auto() 336 RANGE = auto() 337 RECURSIVE = auto() 338 REFRESH = auto() 339 REPLACE = auto() 340 RETURNING = auto() 341 REFERENCES = auto() 342 RIGHT = auto() 343 RLIKE = auto() 344 ROLLBACK = auto() 345 ROLLUP = auto() 346 ROW = auto() 347 ROWS = auto() 348 SELECT = auto() 349 SEMI = auto() 350 SEPARATOR = auto() 351 SEQUENCE = auto() 352 SERDE_PROPERTIES = auto() 353 SET = auto() 354 SETTINGS = auto() 355 SHOW = auto() 356 SIMILAR_TO = auto() 357 SOME = auto() 358 SORT_BY = auto() 359 START_WITH = auto() 360 STORAGE_INTEGRATION = auto() 361 STRUCT = auto() 362 TABLE_SAMPLE = auto() 363 TAG = auto() 364 TEMPORARY = auto() 365 TOP = auto() 366 THEN = auto() 367 TRUE = auto() 368 TRUNCATE = auto() 369 UNCACHE = auto() 370 UNION = auto() 371 UNNEST = auto() 372 UNPIVOT = auto() 373 UPDATE = auto() 374 USE = auto() 375 USING = auto() 376 VALUES = auto() 377 VIEW = auto() 378 VOLATILE = auto() 379 WHEN = auto() 380 WHERE = auto() 381 WINDOW = auto() 382 WITH = auto() 383 UNIQUE = auto() 384 VERSION_SNAPSHOT = auto() 385 TIMESTAMP_SNAPSHOT = auto() 386 OPTION = auto()
An enumeration.
L_PAREN =
<TokenType.L_PAREN: 'L_PAREN'>
R_PAREN =
<TokenType.R_PAREN: 'R_PAREN'>
L_BRACKET =
<TokenType.L_BRACKET: 'L_BRACKET'>
R_BRACKET =
<TokenType.R_BRACKET: 'R_BRACKET'>
L_BRACE =
<TokenType.L_BRACE: 'L_BRACE'>
R_BRACE =
<TokenType.R_BRACE: 'R_BRACE'>
COMMA =
<TokenType.COMMA: 'COMMA'>
DOT =
<TokenType.DOT: 'DOT'>
DASH =
<TokenType.DASH: 'DASH'>
PLUS =
<TokenType.PLUS: 'PLUS'>
COLON =
<TokenType.COLON: 'COLON'>
DCOLON =
<TokenType.DCOLON: 'DCOLON'>
DQMARK =
<TokenType.DQMARK: 'DQMARK'>
SEMICOLON =
<TokenType.SEMICOLON: 'SEMICOLON'>
STAR =
<TokenType.STAR: 'STAR'>
BACKSLASH =
<TokenType.BACKSLASH: 'BACKSLASH'>
SLASH =
<TokenType.SLASH: 'SLASH'>
LT =
<TokenType.LT: 'LT'>
LTE =
<TokenType.LTE: 'LTE'>
GT =
<TokenType.GT: 'GT'>
GTE =
<TokenType.GTE: 'GTE'>
NOT =
<TokenType.NOT: 'NOT'>
EQ =
<TokenType.EQ: 'EQ'>
NEQ =
<TokenType.NEQ: 'NEQ'>
NULLSAFE_EQ =
<TokenType.NULLSAFE_EQ: 'NULLSAFE_EQ'>
COLON_EQ =
<TokenType.COLON_EQ: 'COLON_EQ'>
AND =
<TokenType.AND: 'AND'>
OR =
<TokenType.OR: 'OR'>
AMP =
<TokenType.AMP: 'AMP'>
DPIPE =
<TokenType.DPIPE: 'DPIPE'>
PIPE =
<TokenType.PIPE: 'PIPE'>
PIPE_SLASH =
<TokenType.PIPE_SLASH: 'PIPE_SLASH'>
DPIPE_SLASH =
<TokenType.DPIPE_SLASH: 'DPIPE_SLASH'>
CARET =
<TokenType.CARET: 'CARET'>
TILDA =
<TokenType.TILDA: 'TILDA'>
ARROW =
<TokenType.ARROW: 'ARROW'>
DARROW =
<TokenType.DARROW: 'DARROW'>
FARROW =
<TokenType.FARROW: 'FARROW'>
HASH =
<TokenType.HASH: 'HASH'>
HASH_ARROW =
<TokenType.HASH_ARROW: 'HASH_ARROW'>
DHASH_ARROW =
<TokenType.DHASH_ARROW: 'DHASH_ARROW'>
LR_ARROW =
<TokenType.LR_ARROW: 'LR_ARROW'>
DAT =
<TokenType.DAT: 'DAT'>
LT_AT =
<TokenType.LT_AT: 'LT_AT'>
AT_GT =
<TokenType.AT_GT: 'AT_GT'>
DOLLAR =
<TokenType.DOLLAR: 'DOLLAR'>
PARAMETER =
<TokenType.PARAMETER: 'PARAMETER'>
SESSION_PARAMETER =
<TokenType.SESSION_PARAMETER: 'SESSION_PARAMETER'>
DAMP =
<TokenType.DAMP: 'DAMP'>
XOR =
<TokenType.XOR: 'XOR'>
DSTAR =
<TokenType.DSTAR: 'DSTAR'>
BLOCK_START =
<TokenType.BLOCK_START: 'BLOCK_START'>
BLOCK_END =
<TokenType.BLOCK_END: 'BLOCK_END'>
SPACE =
<TokenType.SPACE: 'SPACE'>
BREAK =
<TokenType.BREAK: 'BREAK'>
STRING =
<TokenType.STRING: 'STRING'>
NUMBER =
<TokenType.NUMBER: 'NUMBER'>
IDENTIFIER =
<TokenType.IDENTIFIER: 'IDENTIFIER'>
DATABASE =
<TokenType.DATABASE: 'DATABASE'>
COLUMN =
<TokenType.COLUMN: 'COLUMN'>
COLUMN_DEF =
<TokenType.COLUMN_DEF: 'COLUMN_DEF'>
SCHEMA =
<TokenType.SCHEMA: 'SCHEMA'>
TABLE =
<TokenType.TABLE: 'TABLE'>
VAR =
<TokenType.VAR: 'VAR'>
BIT_STRING =
<TokenType.BIT_STRING: 'BIT_STRING'>
HEX_STRING =
<TokenType.HEX_STRING: 'HEX_STRING'>
BYTE_STRING =
<TokenType.BYTE_STRING: 'BYTE_STRING'>
NATIONAL_STRING =
<TokenType.NATIONAL_STRING: 'NATIONAL_STRING'>
RAW_STRING =
<TokenType.RAW_STRING: 'RAW_STRING'>
HEREDOC_STRING =
<TokenType.HEREDOC_STRING: 'HEREDOC_STRING'>
UNICODE_STRING =
<TokenType.UNICODE_STRING: 'UNICODE_STRING'>
BIT =
<TokenType.BIT: 'BIT'>
BOOLEAN =
<TokenType.BOOLEAN: 'BOOLEAN'>
TINYINT =
<TokenType.TINYINT: 'TINYINT'>
UTINYINT =
<TokenType.UTINYINT: 'UTINYINT'>
SMALLINT =
<TokenType.SMALLINT: 'SMALLINT'>
USMALLINT =
<TokenType.USMALLINT: 'USMALLINT'>
MEDIUMINT =
<TokenType.MEDIUMINT: 'MEDIUMINT'>
UMEDIUMINT =
<TokenType.UMEDIUMINT: 'UMEDIUMINT'>
INT =
<TokenType.INT: 'INT'>
UINT =
<TokenType.UINT: 'UINT'>
BIGINT =
<TokenType.BIGINT: 'BIGINT'>
UBIGINT =
<TokenType.UBIGINT: 'UBIGINT'>
INT128 =
<TokenType.INT128: 'INT128'>
UINT128 =
<TokenType.UINT128: 'UINT128'>
INT256 =
<TokenType.INT256: 'INT256'>
UINT256 =
<TokenType.UINT256: 'UINT256'>
FLOAT =
<TokenType.FLOAT: 'FLOAT'>
DOUBLE =
<TokenType.DOUBLE: 'DOUBLE'>
DECIMAL =
<TokenType.DECIMAL: 'DECIMAL'>
UDECIMAL =
<TokenType.UDECIMAL: 'UDECIMAL'>
BIGDECIMAL =
<TokenType.BIGDECIMAL: 'BIGDECIMAL'>
CHAR =
<TokenType.CHAR: 'CHAR'>
NCHAR =
<TokenType.NCHAR: 'NCHAR'>
VARCHAR =
<TokenType.VARCHAR: 'VARCHAR'>
NVARCHAR =
<TokenType.NVARCHAR: 'NVARCHAR'>
BPCHAR =
<TokenType.BPCHAR: 'BPCHAR'>
TEXT =
<TokenType.TEXT: 'TEXT'>
MEDIUMTEXT =
<TokenType.MEDIUMTEXT: 'MEDIUMTEXT'>
LONGTEXT =
<TokenType.LONGTEXT: 'LONGTEXT'>
MEDIUMBLOB =
<TokenType.MEDIUMBLOB: 'MEDIUMBLOB'>
LONGBLOB =
<TokenType.LONGBLOB: 'LONGBLOB'>
TINYBLOB =
<TokenType.TINYBLOB: 'TINYBLOB'>
TINYTEXT =
<TokenType.TINYTEXT: 'TINYTEXT'>
NAME =
<TokenType.NAME: 'NAME'>
BINARY =
<TokenType.BINARY: 'BINARY'>
VARBINARY =
<TokenType.VARBINARY: 'VARBINARY'>
JSON =
<TokenType.JSON: 'JSON'>
JSONB =
<TokenType.JSONB: 'JSONB'>
TIME =
<TokenType.TIME: 'TIME'>
TIMETZ =
<TokenType.TIMETZ: 'TIMETZ'>
TIMESTAMP =
<TokenType.TIMESTAMP: 'TIMESTAMP'>
TIMESTAMPTZ =
<TokenType.TIMESTAMPTZ: 'TIMESTAMPTZ'>
TIMESTAMPLTZ =
<TokenType.TIMESTAMPLTZ: 'TIMESTAMPLTZ'>
TIMESTAMPNTZ =
<TokenType.TIMESTAMPNTZ: 'TIMESTAMPNTZ'>
TIMESTAMP_S =
<TokenType.TIMESTAMP_S: 'TIMESTAMP_S'>
TIMESTAMP_MS =
<TokenType.TIMESTAMP_MS: 'TIMESTAMP_MS'>
TIMESTAMP_NS =
<TokenType.TIMESTAMP_NS: 'TIMESTAMP_NS'>
DATETIME =
<TokenType.DATETIME: 'DATETIME'>
DATETIME64 =
<TokenType.DATETIME64: 'DATETIME64'>
DATE =
<TokenType.DATE: 'DATE'>
DATE32 =
<TokenType.DATE32: 'DATE32'>
INT4RANGE =
<TokenType.INT4RANGE: 'INT4RANGE'>
INT4MULTIRANGE =
<TokenType.INT4MULTIRANGE: 'INT4MULTIRANGE'>
INT8RANGE =
<TokenType.INT8RANGE: 'INT8RANGE'>
INT8MULTIRANGE =
<TokenType.INT8MULTIRANGE: 'INT8MULTIRANGE'>
NUMRANGE =
<TokenType.NUMRANGE: 'NUMRANGE'>
NUMMULTIRANGE =
<TokenType.NUMMULTIRANGE: 'NUMMULTIRANGE'>
TSRANGE =
<TokenType.TSRANGE: 'TSRANGE'>
TSMULTIRANGE =
<TokenType.TSMULTIRANGE: 'TSMULTIRANGE'>
TSTZRANGE =
<TokenType.TSTZRANGE: 'TSTZRANGE'>
TSTZMULTIRANGE =
<TokenType.TSTZMULTIRANGE: 'TSTZMULTIRANGE'>
DATERANGE =
<TokenType.DATERANGE: 'DATERANGE'>
DATEMULTIRANGE =
<TokenType.DATEMULTIRANGE: 'DATEMULTIRANGE'>
UUID =
<TokenType.UUID: 'UUID'>
GEOGRAPHY =
<TokenType.GEOGRAPHY: 'GEOGRAPHY'>
NULLABLE =
<TokenType.NULLABLE: 'NULLABLE'>
GEOMETRY =
<TokenType.GEOMETRY: 'GEOMETRY'>
HLLSKETCH =
<TokenType.HLLSKETCH: 'HLLSKETCH'>
HSTORE =
<TokenType.HSTORE: 'HSTORE'>
SUPER =
<TokenType.SUPER: 'SUPER'>
SERIAL =
<TokenType.SERIAL: 'SERIAL'>
SMALLSERIAL =
<TokenType.SMALLSERIAL: 'SMALLSERIAL'>
BIGSERIAL =
<TokenType.BIGSERIAL: 'BIGSERIAL'>
XML =
<TokenType.XML: 'XML'>
YEAR =
<TokenType.YEAR: 'YEAR'>
UNIQUEIDENTIFIER =
<TokenType.UNIQUEIDENTIFIER: 'UNIQUEIDENTIFIER'>
USERDEFINED =
<TokenType.USERDEFINED: 'USERDEFINED'>
MONEY =
<TokenType.MONEY: 'MONEY'>
SMALLMONEY =
<TokenType.SMALLMONEY: 'SMALLMONEY'>
ROWVERSION =
<TokenType.ROWVERSION: 'ROWVERSION'>
IMAGE =
<TokenType.IMAGE: 'IMAGE'>
VARIANT =
<TokenType.VARIANT: 'VARIANT'>
OBJECT =
<TokenType.OBJECT: 'OBJECT'>
INET =
<TokenType.INET: 'INET'>
IPADDRESS =
<TokenType.IPADDRESS: 'IPADDRESS'>
IPPREFIX =
<TokenType.IPPREFIX: 'IPPREFIX'>
IPV4 =
<TokenType.IPV4: 'IPV4'>
IPV6 =
<TokenType.IPV6: 'IPV6'>
ENUM =
<TokenType.ENUM: 'ENUM'>
ENUM8 =
<TokenType.ENUM8: 'ENUM8'>
ENUM16 =
<TokenType.ENUM16: 'ENUM16'>
FIXEDSTRING =
<TokenType.FIXEDSTRING: 'FIXEDSTRING'>
LOWCARDINALITY =
<TokenType.LOWCARDINALITY: 'LOWCARDINALITY'>
NESTED =
<TokenType.NESTED: 'NESTED'>
AGGREGATEFUNCTION =
<TokenType.AGGREGATEFUNCTION: 'AGGREGATEFUNCTION'>
SIMPLEAGGREGATEFUNCTION =
<TokenType.SIMPLEAGGREGATEFUNCTION: 'SIMPLEAGGREGATEFUNCTION'>
TDIGEST =
<TokenType.TDIGEST: 'TDIGEST'>
UNKNOWN =
<TokenType.UNKNOWN: 'UNKNOWN'>
ALIAS =
<TokenType.ALIAS: 'ALIAS'>
ALTER =
<TokenType.ALTER: 'ALTER'>
ALWAYS =
<TokenType.ALWAYS: 'ALWAYS'>
ALL =
<TokenType.ALL: 'ALL'>
ANTI =
<TokenType.ANTI: 'ANTI'>
ANY =
<TokenType.ANY: 'ANY'>
APPLY =
<TokenType.APPLY: 'APPLY'>
ARRAY =
<TokenType.ARRAY: 'ARRAY'>
ASC =
<TokenType.ASC: 'ASC'>
ASOF =
<TokenType.ASOF: 'ASOF'>
AUTO_INCREMENT =
<TokenType.AUTO_INCREMENT: 'AUTO_INCREMENT'>
BEGIN =
<TokenType.BEGIN: 'BEGIN'>
BETWEEN =
<TokenType.BETWEEN: 'BETWEEN'>
CACHE =
<TokenType.CACHE: 'CACHE'>
CASE =
<TokenType.CASE: 'CASE'>
CHARACTER_SET =
<TokenType.CHARACTER_SET: 'CHARACTER_SET'>
CLUSTER_BY =
<TokenType.CLUSTER_BY: 'CLUSTER_BY'>
COLLATE =
<TokenType.COLLATE: 'COLLATE'>
COMMAND =
<TokenType.COMMAND: 'COMMAND'>
COMMENT =
<TokenType.COMMENT: 'COMMENT'>
COMMIT =
<TokenType.COMMIT: 'COMMIT'>
CONNECT_BY =
<TokenType.CONNECT_BY: 'CONNECT_BY'>
CONSTRAINT =
<TokenType.CONSTRAINT: 'CONSTRAINT'>
COPY =
<TokenType.COPY: 'COPY'>
CREATE =
<TokenType.CREATE: 'CREATE'>
CROSS =
<TokenType.CROSS: 'CROSS'>
CUBE =
<TokenType.CUBE: 'CUBE'>
CURRENT_DATE =
<TokenType.CURRENT_DATE: 'CURRENT_DATE'>
CURRENT_DATETIME =
<TokenType.CURRENT_DATETIME: 'CURRENT_DATETIME'>
CURRENT_TIME =
<TokenType.CURRENT_TIME: 'CURRENT_TIME'>
CURRENT_TIMESTAMP =
<TokenType.CURRENT_TIMESTAMP: 'CURRENT_TIMESTAMP'>
CURRENT_USER =
<TokenType.CURRENT_USER: 'CURRENT_USER'>
DECLARE =
<TokenType.DECLARE: 'DECLARE'>
DEFAULT =
<TokenType.DEFAULT: 'DEFAULT'>
DELETE =
<TokenType.DELETE: 'DELETE'>
DESC =
<TokenType.DESC: 'DESC'>
DESCRIBE =
<TokenType.DESCRIBE: 'DESCRIBE'>
DICTIONARY =
<TokenType.DICTIONARY: 'DICTIONARY'>
DISTINCT =
<TokenType.DISTINCT: 'DISTINCT'>
DISTRIBUTE_BY =
<TokenType.DISTRIBUTE_BY: 'DISTRIBUTE_BY'>
DIV =
<TokenType.DIV: 'DIV'>
DROP =
<TokenType.DROP: 'DROP'>
ELSE =
<TokenType.ELSE: 'ELSE'>
END =
<TokenType.END: 'END'>
ESCAPE =
<TokenType.ESCAPE: 'ESCAPE'>
EXCEPT =
<TokenType.EXCEPT: 'EXCEPT'>
EXECUTE =
<TokenType.EXECUTE: 'EXECUTE'>
EXISTS =
<TokenType.EXISTS: 'EXISTS'>
FALSE =
<TokenType.FALSE: 'FALSE'>
FETCH =
<TokenType.FETCH: 'FETCH'>
FILTER =
<TokenType.FILTER: 'FILTER'>
FINAL =
<TokenType.FINAL: 'FINAL'>
FIRST =
<TokenType.FIRST: 'FIRST'>
FOR =
<TokenType.FOR: 'FOR'>
FORCE =
<TokenType.FORCE: 'FORCE'>
FOREIGN_KEY =
<TokenType.FOREIGN_KEY: 'FOREIGN_KEY'>
FORMAT =
<TokenType.FORMAT: 'FORMAT'>
FROM =
<TokenType.FROM: 'FROM'>
FULL =
<TokenType.FULL: 'FULL'>
FUNCTION =
<TokenType.FUNCTION: 'FUNCTION'>
GLOB =
<TokenType.GLOB: 'GLOB'>
GLOBAL =
<TokenType.GLOBAL: 'GLOBAL'>
GROUP_BY =
<TokenType.GROUP_BY: 'GROUP_BY'>
GROUPING_SETS =
<TokenType.GROUPING_SETS: 'GROUPING_SETS'>
HAVING =
<TokenType.HAVING: 'HAVING'>
HINT =
<TokenType.HINT: 'HINT'>
IGNORE =
<TokenType.IGNORE: 'IGNORE'>
ILIKE =
<TokenType.ILIKE: 'ILIKE'>
ILIKE_ANY =
<TokenType.ILIKE_ANY: 'ILIKE_ANY'>
IN =
<TokenType.IN: 'IN'>
INDEX =
<TokenType.INDEX: 'INDEX'>
INNER =
<TokenType.INNER: 'INNER'>
INSERT =
<TokenType.INSERT: 'INSERT'>
INTERSECT =
<TokenType.INTERSECT: 'INTERSECT'>
INTERVAL =
<TokenType.INTERVAL: 'INTERVAL'>
INTO =
<TokenType.INTO: 'INTO'>
INTRODUCER =
<TokenType.INTRODUCER: 'INTRODUCER'>
IRLIKE =
<TokenType.IRLIKE: 'IRLIKE'>
IS =
<TokenType.IS: 'IS'>
ISNULL =
<TokenType.ISNULL: 'ISNULL'>
JOIN =
<TokenType.JOIN: 'JOIN'>
JOIN_MARKER =
<TokenType.JOIN_MARKER: 'JOIN_MARKER'>
KEEP =
<TokenType.KEEP: 'KEEP'>
KILL =
<TokenType.KILL: 'KILL'>
LANGUAGE =
<TokenType.LANGUAGE: 'LANGUAGE'>
LATERAL =
<TokenType.LATERAL: 'LATERAL'>
LEFT =
<TokenType.LEFT: 'LEFT'>
LIKE =
<TokenType.LIKE: 'LIKE'>
LIKE_ANY =
<TokenType.LIKE_ANY: 'LIKE_ANY'>
LIMIT =
<TokenType.LIMIT: 'LIMIT'>
LOAD =
<TokenType.LOAD: 'LOAD'>
LOCK =
<TokenType.LOCK: 'LOCK'>
MAP =
<TokenType.MAP: 'MAP'>
MATCH_CONDITION =
<TokenType.MATCH_CONDITION: 'MATCH_CONDITION'>
MATCH_RECOGNIZE =
<TokenType.MATCH_RECOGNIZE: 'MATCH_RECOGNIZE'>
MEMBER_OF =
<TokenType.MEMBER_OF: 'MEMBER_OF'>
MERGE =
<TokenType.MERGE: 'MERGE'>
MOD =
<TokenType.MOD: 'MOD'>
MODEL =
<TokenType.MODEL: 'MODEL'>
NATURAL =
<TokenType.NATURAL: 'NATURAL'>
NEXT =
<TokenType.NEXT: 'NEXT'>
NOTNULL =
<TokenType.NOTNULL: 'NOTNULL'>
NULL =
<TokenType.NULL: 'NULL'>
OBJECT_IDENTIFIER =
<TokenType.OBJECT_IDENTIFIER: 'OBJECT_IDENTIFIER'>
OFFSET =
<TokenType.OFFSET: 'OFFSET'>
ON =
<TokenType.ON: 'ON'>
ONLY =
<TokenType.ONLY: 'ONLY'>
OPERATOR =
<TokenType.OPERATOR: 'OPERATOR'>
ORDER_BY =
<TokenType.ORDER_BY: 'ORDER_BY'>
ORDER_SIBLINGS_BY =
<TokenType.ORDER_SIBLINGS_BY: 'ORDER_SIBLINGS_BY'>
ORDERED =
<TokenType.ORDERED: 'ORDERED'>
ORDINALITY =
<TokenType.ORDINALITY: 'ORDINALITY'>
OUTER =
<TokenType.OUTER: 'OUTER'>
OVER =
<TokenType.OVER: 'OVER'>
OVERLAPS =
<TokenType.OVERLAPS: 'OVERLAPS'>
OVERWRITE =
<TokenType.OVERWRITE: 'OVERWRITE'>
PARTITION =
<TokenType.PARTITION: 'PARTITION'>
PARTITION_BY =
<TokenType.PARTITION_BY: 'PARTITION_BY'>
PERCENT =
<TokenType.PERCENT: 'PERCENT'>
PIVOT =
<TokenType.PIVOT: 'PIVOT'>
PLACEHOLDER =
<TokenType.PLACEHOLDER: 'PLACEHOLDER'>
POSITIONAL =
<TokenType.POSITIONAL: 'POSITIONAL'>
PRAGMA =
<TokenType.PRAGMA: 'PRAGMA'>
PREWHERE =
<TokenType.PREWHERE: 'PREWHERE'>
PRIMARY_KEY =
<TokenType.PRIMARY_KEY: 'PRIMARY_KEY'>
PROCEDURE =
<TokenType.PROCEDURE: 'PROCEDURE'>
PROPERTIES =
<TokenType.PROPERTIES: 'PROPERTIES'>
PSEUDO_TYPE =
<TokenType.PSEUDO_TYPE: 'PSEUDO_TYPE'>
QUALIFY =
<TokenType.QUALIFY: 'QUALIFY'>
QUOTE =
<TokenType.QUOTE: 'QUOTE'>
RANGE =
<TokenType.RANGE: 'RANGE'>
RECURSIVE =
<TokenType.RECURSIVE: 'RECURSIVE'>
REFRESH =
<TokenType.REFRESH: 'REFRESH'>
REPLACE =
<TokenType.REPLACE: 'REPLACE'>
RETURNING =
<TokenType.RETURNING: 'RETURNING'>
REFERENCES =
<TokenType.REFERENCES: 'REFERENCES'>
RIGHT =
<TokenType.RIGHT: 'RIGHT'>
RLIKE =
<TokenType.RLIKE: 'RLIKE'>
ROLLBACK =
<TokenType.ROLLBACK: 'ROLLBACK'>
ROLLUP =
<TokenType.ROLLUP: 'ROLLUP'>
ROW =
<TokenType.ROW: 'ROW'>
ROWS =
<TokenType.ROWS: 'ROWS'>
SELECT =
<TokenType.SELECT: 'SELECT'>
SEMI =
<TokenType.SEMI: 'SEMI'>
SEPARATOR =
<TokenType.SEPARATOR: 'SEPARATOR'>
SEQUENCE =
<TokenType.SEQUENCE: 'SEQUENCE'>
SERDE_PROPERTIES =
<TokenType.SERDE_PROPERTIES: 'SERDE_PROPERTIES'>
SET =
<TokenType.SET: 'SET'>
SETTINGS =
<TokenType.SETTINGS: 'SETTINGS'>
SHOW =
<TokenType.SHOW: 'SHOW'>
SIMILAR_TO =
<TokenType.SIMILAR_TO: 'SIMILAR_TO'>
SOME =
<TokenType.SOME: 'SOME'>
SORT_BY =
<TokenType.SORT_BY: 'SORT_BY'>
START_WITH =
<TokenType.START_WITH: 'START_WITH'>
STORAGE_INTEGRATION =
<TokenType.STORAGE_INTEGRATION: 'STORAGE_INTEGRATION'>
STRUCT =
<TokenType.STRUCT: 'STRUCT'>
TABLE_SAMPLE =
<TokenType.TABLE_SAMPLE: 'TABLE_SAMPLE'>
TAG =
<TokenType.TAG: 'TAG'>
TEMPORARY =
<TokenType.TEMPORARY: 'TEMPORARY'>
TOP =
<TokenType.TOP: 'TOP'>
THEN =
<TokenType.THEN: 'THEN'>
TRUE =
<TokenType.TRUE: 'TRUE'>
TRUNCATE =
<TokenType.TRUNCATE: 'TRUNCATE'>
UNCACHE =
<TokenType.UNCACHE: 'UNCACHE'>
UNION =
<TokenType.UNION: 'UNION'>
UNNEST =
<TokenType.UNNEST: 'UNNEST'>
UNPIVOT =
<TokenType.UNPIVOT: 'UNPIVOT'>
UPDATE =
<TokenType.UPDATE: 'UPDATE'>
USE =
<TokenType.USE: 'USE'>
USING =
<TokenType.USING: 'USING'>
VALUES =
<TokenType.VALUES: 'VALUES'>
VIEW =
<TokenType.VIEW: 'VIEW'>
VOLATILE =
<TokenType.VOLATILE: 'VOLATILE'>
WHEN =
<TokenType.WHEN: 'WHEN'>
WHERE =
<TokenType.WHERE: 'WHERE'>
WINDOW =
<TokenType.WINDOW: 'WINDOW'>
WITH =
<TokenType.WITH: 'WITH'>
UNIQUE =
<TokenType.UNIQUE: 'UNIQUE'>
VERSION_SNAPSHOT =
<TokenType.VERSION_SNAPSHOT: 'VERSION_SNAPSHOT'>
TIMESTAMP_SNAPSHOT =
<TokenType.TIMESTAMP_SNAPSHOT: 'TIMESTAMP_SNAPSHOT'>
OPTION =
<TokenType.OPTION: 'OPTION'>
Inherited Members
- enum.Enum
- name
- value
class
Token:
393class Token: 394 __slots__ = ("token_type", "text", "line", "col", "start", "end", "comments") 395 396 @classmethod 397 def number(cls, number: int) -> Token: 398 """Returns a NUMBER token with `number` as its text.""" 399 return cls(TokenType.NUMBER, str(number)) 400 401 @classmethod 402 def string(cls, string: str) -> Token: 403 """Returns a STRING token with `string` as its text.""" 404 return cls(TokenType.STRING, string) 405 406 @classmethod 407 def identifier(cls, identifier: str) -> Token: 408 """Returns an IDENTIFIER token with `identifier` as its text.""" 409 return cls(TokenType.IDENTIFIER, identifier) 410 411 @classmethod 412 def var(cls, var: str) -> Token: 413 """Returns an VAR token with `var` as its text.""" 414 return cls(TokenType.VAR, var) 415 416 def __init__( 417 self, 418 token_type: TokenType, 419 text: str, 420 line: int = 1, 421 col: int = 1, 422 start: int = 0, 423 end: int = 0, 424 comments: t.Optional[t.List[str]] = None, 425 ) -> None: 426 """Token initializer. 427 428 Args: 429 token_type: The TokenType Enum. 430 text: The text of the token. 431 line: The line that the token ends on. 432 col: The column that the token ends on. 433 start: The start index of the token. 434 end: The ending index of the token. 435 comments: The comments to attach to the token. 436 """ 437 self.token_type = token_type 438 self.text = text 439 self.line = line 440 self.col = col 441 self.start = start 442 self.end = end 443 self.comments = [] if comments is None else comments 444 445 def __repr__(self) -> str: 446 attributes = ", ".join(f"{k}: {getattr(self, k)}" for k in self.__slots__) 447 return f"<Token {attributes}>"
Token( token_type: TokenType, text: str, line: int = 1, col: int = 1, start: int = 0, end: int = 0, comments: Optional[List[str]] = None)
416 def __init__( 417 self, 418 token_type: TokenType, 419 text: str, 420 line: int = 1, 421 col: int = 1, 422 start: int = 0, 423 end: int = 0, 424 comments: t.Optional[t.List[str]] = None, 425 ) -> None: 426 """Token initializer. 427 428 Args: 429 token_type: The TokenType Enum. 430 text: The text of the token. 431 line: The line that the token ends on. 432 col: The column that the token ends on. 433 start: The start index of the token. 434 end: The ending index of the token. 435 comments: The comments to attach to the token. 436 """ 437 self.token_type = token_type 438 self.text = text 439 self.line = line 440 self.col = col 441 self.start = start 442 self.end = end 443 self.comments = [] if comments is None else comments
Token initializer.
Arguments:
- token_type: The TokenType Enum.
- text: The text of the token.
- line: The line that the token ends on.
- col: The column that the token ends on.
- start: The start index of the token.
- end: The ending index of the token.
- comments: The comments to attach to the token.
396 @classmethod 397 def number(cls, number: int) -> Token: 398 """Returns a NUMBER token with `number` as its text.""" 399 return cls(TokenType.NUMBER, str(number))
Returns a NUMBER token with number
as its text.
401 @classmethod 402 def string(cls, string: str) -> Token: 403 """Returns a STRING token with `string` as its text.""" 404 return cls(TokenType.STRING, string)
Returns a STRING token with string
as its text.
406 @classmethod 407 def identifier(cls, identifier: str) -> Token: 408 """Returns an IDENTIFIER token with `identifier` as its text.""" 409 return cls(TokenType.IDENTIFIER, identifier)
Returns an IDENTIFIER token with identifier
as its text.
class
Tokenizer:
548class Tokenizer(metaclass=_Tokenizer): 549 SINGLE_TOKENS = { 550 "(": TokenType.L_PAREN, 551 ")": TokenType.R_PAREN, 552 "[": TokenType.L_BRACKET, 553 "]": TokenType.R_BRACKET, 554 "{": TokenType.L_BRACE, 555 "}": TokenType.R_BRACE, 556 "&": TokenType.AMP, 557 "^": TokenType.CARET, 558 ":": TokenType.COLON, 559 ",": TokenType.COMMA, 560 ".": TokenType.DOT, 561 "-": TokenType.DASH, 562 "=": TokenType.EQ, 563 ">": TokenType.GT, 564 "<": TokenType.LT, 565 "%": TokenType.MOD, 566 "!": TokenType.NOT, 567 "|": TokenType.PIPE, 568 "+": TokenType.PLUS, 569 ";": TokenType.SEMICOLON, 570 "/": TokenType.SLASH, 571 "\\": TokenType.BACKSLASH, 572 "*": TokenType.STAR, 573 "~": TokenType.TILDA, 574 "?": TokenType.PLACEHOLDER, 575 "@": TokenType.PARAMETER, 576 "#": TokenType.HASH, 577 # Used for breaking a var like x'y' but nothing else the token type doesn't matter 578 "'": TokenType.UNKNOWN, 579 "`": TokenType.UNKNOWN, 580 '"': TokenType.UNKNOWN, 581 } 582 583 BIT_STRINGS: t.List[str | t.Tuple[str, str]] = [] 584 BYTE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 585 HEX_STRINGS: t.List[str | t.Tuple[str, str]] = [] 586 RAW_STRINGS: t.List[str | t.Tuple[str, str]] = [] 587 HEREDOC_STRINGS: t.List[str | t.Tuple[str, str]] = [] 588 UNICODE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 589 IDENTIFIERS: t.List[str | t.Tuple[str, str]] = ['"'] 590 IDENTIFIER_ESCAPES = ['"'] 591 QUOTES: t.List[t.Tuple[str, str] | str] = ["'"] 592 STRING_ESCAPES = ["'"] 593 VAR_SINGLE_TOKENS: t.Set[str] = set() 594 595 # Whether the heredoc tags follow the same lexical rules as unquoted identifiers 596 HEREDOC_TAG_IS_IDENTIFIER = False 597 598 # Token that we'll generate as a fallback if the heredoc prefix doesn't correspond to a heredoc 599 HEREDOC_STRING_ALTERNATIVE = TokenType.VAR 600 601 # Autofilled 602 _COMMENTS: t.Dict[str, str] = {} 603 _FORMAT_STRINGS: t.Dict[str, t.Tuple[str, TokenType]] = {} 604 _IDENTIFIERS: t.Dict[str, str] = {} 605 _IDENTIFIER_ESCAPES: t.Set[str] = set() 606 _QUOTES: t.Dict[str, str] = {} 607 _STRING_ESCAPES: t.Set[str] = set() 608 _KEYWORD_TRIE: t.Dict = {} 609 _RS_TOKENIZER: t.Optional[t.Any] = None 610 611 KEYWORDS: t.Dict[str, TokenType] = { 612 **{f"{{%{postfix}": TokenType.BLOCK_START for postfix in ("", "+", "-")}, 613 **{f"{prefix}%}}": TokenType.BLOCK_END for prefix in ("", "+", "-")}, 614 **{f"{{{{{postfix}": TokenType.BLOCK_START for postfix in ("+", "-")}, 615 **{f"{prefix}}}}}": TokenType.BLOCK_END for prefix in ("+", "-")}, 616 "/*+": TokenType.HINT, 617 "==": TokenType.EQ, 618 "::": TokenType.DCOLON, 619 "||": TokenType.DPIPE, 620 ">=": TokenType.GTE, 621 "<=": TokenType.LTE, 622 "<>": TokenType.NEQ, 623 "!=": TokenType.NEQ, 624 ":=": TokenType.COLON_EQ, 625 "<=>": TokenType.NULLSAFE_EQ, 626 "->": TokenType.ARROW, 627 "->>": TokenType.DARROW, 628 "=>": TokenType.FARROW, 629 "#>": TokenType.HASH_ARROW, 630 "#>>": TokenType.DHASH_ARROW, 631 "<->": TokenType.LR_ARROW, 632 "&&": TokenType.DAMP, 633 "??": TokenType.DQMARK, 634 "ALL": TokenType.ALL, 635 "ALWAYS": TokenType.ALWAYS, 636 "AND": TokenType.AND, 637 "ANTI": TokenType.ANTI, 638 "ANY": TokenType.ANY, 639 "ASC": TokenType.ASC, 640 "AS": TokenType.ALIAS, 641 "ASOF": TokenType.ASOF, 642 "AUTOINCREMENT": TokenType.AUTO_INCREMENT, 643 "AUTO_INCREMENT": TokenType.AUTO_INCREMENT, 644 "BEGIN": TokenType.BEGIN, 645 "BETWEEN": TokenType.BETWEEN, 646 "CACHE": TokenType.CACHE, 647 "UNCACHE": TokenType.UNCACHE, 648 "CASE": TokenType.CASE, 649 "CHARACTER SET": TokenType.CHARACTER_SET, 650 "CLUSTER BY": TokenType.CLUSTER_BY, 651 "COLLATE": TokenType.COLLATE, 652 "COLUMN": TokenType.COLUMN, 653 "COMMIT": TokenType.COMMIT, 654 "CONNECT BY": TokenType.CONNECT_BY, 655 "CONSTRAINT": TokenType.CONSTRAINT, 656 "COPY": TokenType.COPY, 657 "CREATE": TokenType.CREATE, 658 "CROSS": TokenType.CROSS, 659 "CUBE": TokenType.CUBE, 660 "CURRENT_DATE": TokenType.CURRENT_DATE, 661 "CURRENT_TIME": TokenType.CURRENT_TIME, 662 "CURRENT_TIMESTAMP": TokenType.CURRENT_TIMESTAMP, 663 "CURRENT_USER": TokenType.CURRENT_USER, 664 "DATABASE": TokenType.DATABASE, 665 "DEFAULT": TokenType.DEFAULT, 666 "DELETE": TokenType.DELETE, 667 "DESC": TokenType.DESC, 668 "DESCRIBE": TokenType.DESCRIBE, 669 "DISTINCT": TokenType.DISTINCT, 670 "DISTRIBUTE BY": TokenType.DISTRIBUTE_BY, 671 "DIV": TokenType.DIV, 672 "DROP": TokenType.DROP, 673 "ELSE": TokenType.ELSE, 674 "END": TokenType.END, 675 "ENUM": TokenType.ENUM, 676 "ESCAPE": TokenType.ESCAPE, 677 "EXCEPT": TokenType.EXCEPT, 678 "EXECUTE": TokenType.EXECUTE, 679 "EXISTS": TokenType.EXISTS, 680 "FALSE": TokenType.FALSE, 681 "FETCH": TokenType.FETCH, 682 "FILTER": TokenType.FILTER, 683 "FIRST": TokenType.FIRST, 684 "FULL": TokenType.FULL, 685 "FUNCTION": TokenType.FUNCTION, 686 "FOR": TokenType.FOR, 687 "FOREIGN KEY": TokenType.FOREIGN_KEY, 688 "FORMAT": TokenType.FORMAT, 689 "FROM": TokenType.FROM, 690 "GEOGRAPHY": TokenType.GEOGRAPHY, 691 "GEOMETRY": TokenType.GEOMETRY, 692 "GLOB": TokenType.GLOB, 693 "GROUP BY": TokenType.GROUP_BY, 694 "GROUPING SETS": TokenType.GROUPING_SETS, 695 "HAVING": TokenType.HAVING, 696 "ILIKE": TokenType.ILIKE, 697 "IN": TokenType.IN, 698 "INDEX": TokenType.INDEX, 699 "INET": TokenType.INET, 700 "INNER": TokenType.INNER, 701 "INSERT": TokenType.INSERT, 702 "INTERVAL": TokenType.INTERVAL, 703 "INTERSECT": TokenType.INTERSECT, 704 "INTO": TokenType.INTO, 705 "IS": TokenType.IS, 706 "ISNULL": TokenType.ISNULL, 707 "JOIN": TokenType.JOIN, 708 "KEEP": TokenType.KEEP, 709 "KILL": TokenType.KILL, 710 "LATERAL": TokenType.LATERAL, 711 "LEFT": TokenType.LEFT, 712 "LIKE": TokenType.LIKE, 713 "LIMIT": TokenType.LIMIT, 714 "LOAD": TokenType.LOAD, 715 "LOCK": TokenType.LOCK, 716 "MERGE": TokenType.MERGE, 717 "NATURAL": TokenType.NATURAL, 718 "NEXT": TokenType.NEXT, 719 "NOT": TokenType.NOT, 720 "NOTNULL": TokenType.NOTNULL, 721 "NULL": TokenType.NULL, 722 "OBJECT": TokenType.OBJECT, 723 "OFFSET": TokenType.OFFSET, 724 "ON": TokenType.ON, 725 "OR": TokenType.OR, 726 "XOR": TokenType.XOR, 727 "ORDER BY": TokenType.ORDER_BY, 728 "ORDINALITY": TokenType.ORDINALITY, 729 "OUTER": TokenType.OUTER, 730 "OVER": TokenType.OVER, 731 "OVERLAPS": TokenType.OVERLAPS, 732 "OVERWRITE": TokenType.OVERWRITE, 733 "PARTITION": TokenType.PARTITION, 734 "PARTITION BY": TokenType.PARTITION_BY, 735 "PARTITIONED BY": TokenType.PARTITION_BY, 736 "PARTITIONED_BY": TokenType.PARTITION_BY, 737 "PERCENT": TokenType.PERCENT, 738 "PIVOT": TokenType.PIVOT, 739 "PRAGMA": TokenType.PRAGMA, 740 "PRIMARY KEY": TokenType.PRIMARY_KEY, 741 "PROCEDURE": TokenType.PROCEDURE, 742 "QUALIFY": TokenType.QUALIFY, 743 "RANGE": TokenType.RANGE, 744 "RECURSIVE": TokenType.RECURSIVE, 745 "REGEXP": TokenType.RLIKE, 746 "REPLACE": TokenType.REPLACE, 747 "RETURNING": TokenType.RETURNING, 748 "REFERENCES": TokenType.REFERENCES, 749 "RIGHT": TokenType.RIGHT, 750 "RLIKE": TokenType.RLIKE, 751 "ROLLBACK": TokenType.ROLLBACK, 752 "ROLLUP": TokenType.ROLLUP, 753 "ROW": TokenType.ROW, 754 "ROWS": TokenType.ROWS, 755 "SCHEMA": TokenType.SCHEMA, 756 "SELECT": TokenType.SELECT, 757 "SEMI": TokenType.SEMI, 758 "SET": TokenType.SET, 759 "SETTINGS": TokenType.SETTINGS, 760 "SHOW": TokenType.SHOW, 761 "SIMILAR TO": TokenType.SIMILAR_TO, 762 "SOME": TokenType.SOME, 763 "SORT BY": TokenType.SORT_BY, 764 "START WITH": TokenType.START_WITH, 765 "TABLE": TokenType.TABLE, 766 "TABLESAMPLE": TokenType.TABLE_SAMPLE, 767 "TEMP": TokenType.TEMPORARY, 768 "TEMPORARY": TokenType.TEMPORARY, 769 "THEN": TokenType.THEN, 770 "TRUE": TokenType.TRUE, 771 "TRUNCATE": TokenType.TRUNCATE, 772 "UNION": TokenType.UNION, 773 "UNKNOWN": TokenType.UNKNOWN, 774 "UNNEST": TokenType.UNNEST, 775 "UNPIVOT": TokenType.UNPIVOT, 776 "UPDATE": TokenType.UPDATE, 777 "USE": TokenType.USE, 778 "USING": TokenType.USING, 779 "UUID": TokenType.UUID, 780 "VALUES": TokenType.VALUES, 781 "VIEW": TokenType.VIEW, 782 "VOLATILE": TokenType.VOLATILE, 783 "WHEN": TokenType.WHEN, 784 "WHERE": TokenType.WHERE, 785 "WINDOW": TokenType.WINDOW, 786 "WITH": TokenType.WITH, 787 "APPLY": TokenType.APPLY, 788 "ARRAY": TokenType.ARRAY, 789 "BIT": TokenType.BIT, 790 "BOOL": TokenType.BOOLEAN, 791 "BOOLEAN": TokenType.BOOLEAN, 792 "BYTE": TokenType.TINYINT, 793 "MEDIUMINT": TokenType.MEDIUMINT, 794 "INT1": TokenType.TINYINT, 795 "TINYINT": TokenType.TINYINT, 796 "INT16": TokenType.SMALLINT, 797 "SHORT": TokenType.SMALLINT, 798 "SMALLINT": TokenType.SMALLINT, 799 "INT128": TokenType.INT128, 800 "HUGEINT": TokenType.INT128, 801 "INT2": TokenType.SMALLINT, 802 "INTEGER": TokenType.INT, 803 "INT": TokenType.INT, 804 "INT4": TokenType.INT, 805 "INT32": TokenType.INT, 806 "INT64": TokenType.BIGINT, 807 "LONG": TokenType.BIGINT, 808 "BIGINT": TokenType.BIGINT, 809 "INT8": TokenType.TINYINT, 810 "UINT": TokenType.UINT, 811 "DEC": TokenType.DECIMAL, 812 "DECIMAL": TokenType.DECIMAL, 813 "BIGDECIMAL": TokenType.BIGDECIMAL, 814 "BIGNUMERIC": TokenType.BIGDECIMAL, 815 "MAP": TokenType.MAP, 816 "NULLABLE": TokenType.NULLABLE, 817 "NUMBER": TokenType.DECIMAL, 818 "NUMERIC": TokenType.DECIMAL, 819 "FIXED": TokenType.DECIMAL, 820 "REAL": TokenType.FLOAT, 821 "FLOAT": TokenType.FLOAT, 822 "FLOAT4": TokenType.FLOAT, 823 "FLOAT8": TokenType.DOUBLE, 824 "DOUBLE": TokenType.DOUBLE, 825 "DOUBLE PRECISION": TokenType.DOUBLE, 826 "JSON": TokenType.JSON, 827 "CHAR": TokenType.CHAR, 828 "CHARACTER": TokenType.CHAR, 829 "NCHAR": TokenType.NCHAR, 830 "VARCHAR": TokenType.VARCHAR, 831 "VARCHAR2": TokenType.VARCHAR, 832 "NVARCHAR": TokenType.NVARCHAR, 833 "NVARCHAR2": TokenType.NVARCHAR, 834 "BPCHAR": TokenType.BPCHAR, 835 "STR": TokenType.TEXT, 836 "STRING": TokenType.TEXT, 837 "TEXT": TokenType.TEXT, 838 "LONGTEXT": TokenType.LONGTEXT, 839 "MEDIUMTEXT": TokenType.MEDIUMTEXT, 840 "TINYTEXT": TokenType.TINYTEXT, 841 "CLOB": TokenType.TEXT, 842 "LONGVARCHAR": TokenType.TEXT, 843 "BINARY": TokenType.BINARY, 844 "BLOB": TokenType.VARBINARY, 845 "LONGBLOB": TokenType.LONGBLOB, 846 "MEDIUMBLOB": TokenType.MEDIUMBLOB, 847 "TINYBLOB": TokenType.TINYBLOB, 848 "BYTEA": TokenType.VARBINARY, 849 "VARBINARY": TokenType.VARBINARY, 850 "TIME": TokenType.TIME, 851 "TIMETZ": TokenType.TIMETZ, 852 "TIMESTAMP": TokenType.TIMESTAMP, 853 "TIMESTAMPTZ": TokenType.TIMESTAMPTZ, 854 "TIMESTAMPLTZ": TokenType.TIMESTAMPLTZ, 855 "TIMESTAMP_LTZ": TokenType.TIMESTAMPLTZ, 856 "TIMESTAMPNTZ": TokenType.TIMESTAMPNTZ, 857 "TIMESTAMP_NTZ": TokenType.TIMESTAMPNTZ, 858 "DATE": TokenType.DATE, 859 "DATETIME": TokenType.DATETIME, 860 "INT4RANGE": TokenType.INT4RANGE, 861 "INT4MULTIRANGE": TokenType.INT4MULTIRANGE, 862 "INT8RANGE": TokenType.INT8RANGE, 863 "INT8MULTIRANGE": TokenType.INT8MULTIRANGE, 864 "NUMRANGE": TokenType.NUMRANGE, 865 "NUMMULTIRANGE": TokenType.NUMMULTIRANGE, 866 "TSRANGE": TokenType.TSRANGE, 867 "TSMULTIRANGE": TokenType.TSMULTIRANGE, 868 "TSTZRANGE": TokenType.TSTZRANGE, 869 "TSTZMULTIRANGE": TokenType.TSTZMULTIRANGE, 870 "DATERANGE": TokenType.DATERANGE, 871 "DATEMULTIRANGE": TokenType.DATEMULTIRANGE, 872 "UNIQUE": TokenType.UNIQUE, 873 "STRUCT": TokenType.STRUCT, 874 "SEQUENCE": TokenType.SEQUENCE, 875 "VARIANT": TokenType.VARIANT, 876 "ALTER": TokenType.ALTER, 877 "ANALYZE": TokenType.COMMAND, 878 "CALL": TokenType.COMMAND, 879 "COMMENT": TokenType.COMMENT, 880 "EXPLAIN": TokenType.COMMAND, 881 "GRANT": TokenType.COMMAND, 882 "OPTIMIZE": TokenType.COMMAND, 883 "PREPARE": TokenType.COMMAND, 884 "VACUUM": TokenType.COMMAND, 885 "USER-DEFINED": TokenType.USERDEFINED, 886 "FOR VERSION": TokenType.VERSION_SNAPSHOT, 887 "FOR TIMESTAMP": TokenType.TIMESTAMP_SNAPSHOT, 888 } 889 890 WHITE_SPACE: t.Dict[t.Optional[str], TokenType] = { 891 " ": TokenType.SPACE, 892 "\t": TokenType.SPACE, 893 "\n": TokenType.BREAK, 894 "\r": TokenType.BREAK, 895 } 896 897 COMMANDS = { 898 TokenType.COMMAND, 899 TokenType.EXECUTE, 900 TokenType.FETCH, 901 TokenType.SHOW, 902 } 903 904 COMMAND_PREFIX_TOKENS = {TokenType.SEMICOLON, TokenType.BEGIN} 905 906 # Handle numeric literals like in hive (3L = BIGINT) 907 NUMERIC_LITERALS: t.Dict[str, str] = {} 908 909 COMMENTS = ["--", ("/*", "*/")] 910 911 __slots__ = ( 912 "sql", 913 "size", 914 "tokens", 915 "dialect", 916 "_start", 917 "_current", 918 "_line", 919 "_col", 920 "_comments", 921 "_char", 922 "_end", 923 "_peek", 924 "_prev_token_line", 925 "_rs_dialect_settings", 926 ) 927 928 def __init__(self, dialect: DialectType = None) -> None: 929 from sqlglot.dialects import Dialect 930 931 self.dialect = Dialect.get_or_raise(dialect) 932 933 if USE_RS_TOKENIZER: 934 self._rs_dialect_settings = RsTokenizerDialectSettings( 935 unescaped_sequences=self.dialect.UNESCAPED_SEQUENCES, 936 identifiers_can_start_with_digit=self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT, 937 ) 938 939 self.reset() 940 941 def reset(self) -> None: 942 self.sql = "" 943 self.size = 0 944 self.tokens: t.List[Token] = [] 945 self._start = 0 946 self._current = 0 947 self._line = 1 948 self._col = 0 949 self._comments: t.List[str] = [] 950 951 self._char = "" 952 self._end = False 953 self._peek = "" 954 self._prev_token_line = -1 955 956 def tokenize(self, sql: str) -> t.List[Token]: 957 """Returns a list of tokens corresponding to the SQL string `sql`.""" 958 if USE_RS_TOKENIZER: 959 return self.tokenize_rs(sql) 960 961 self.reset() 962 self.sql = sql 963 self.size = len(sql) 964 965 try: 966 self._scan() 967 except Exception as e: 968 start = max(self._current - 50, 0) 969 end = min(self._current + 50, self.size - 1) 970 context = self.sql[start:end] 971 raise TokenError(f"Error tokenizing '{context}'") from e 972 973 return self.tokens 974 975 def _scan(self, until: t.Optional[t.Callable] = None) -> None: 976 while self.size and not self._end: 977 current = self._current 978 979 # Skip spaces here rather than iteratively calling advance() for performance reasons 980 while current < self.size: 981 char = self.sql[current] 982 983 if char.isspace() and (char == " " or char == "\t"): 984 current += 1 985 else: 986 break 987 988 offset = current - self._current if current > self._current else 1 989 990 self._start = current 991 self._advance(offset) 992 993 if not self._char.isspace(): 994 if self._char.isdigit(): 995 self._scan_number() 996 elif self._char in self._IDENTIFIERS: 997 self._scan_identifier(self._IDENTIFIERS[self._char]) 998 else: 999 self._scan_keywords() 1000 1001 if until and until(): 1002 break 1003 1004 if self.tokens and self._comments: 1005 self.tokens[-1].comments.extend(self._comments) 1006 1007 def _chars(self, size: int) -> str: 1008 if size == 1: 1009 return self._char 1010 1011 start = self._current - 1 1012 end = start + size 1013 1014 return self.sql[start:end] if end <= self.size else "" 1015 1016 def _advance(self, i: int = 1, alnum: bool = False) -> None: 1017 if self.WHITE_SPACE.get(self._char) is TokenType.BREAK: 1018 # Ensures we don't count an extra line if we get a \r\n line break sequence 1019 if not (self._char == "\r" and self._peek == "\n"): 1020 self._col = 1 1021 self._line += 1 1022 else: 1023 self._col += i 1024 1025 self._current += i 1026 self._end = self._current >= self.size 1027 self._char = self.sql[self._current - 1] 1028 self._peek = "" if self._end else self.sql[self._current] 1029 1030 if alnum and self._char.isalnum(): 1031 # Here we use local variables instead of attributes for better performance 1032 _col = self._col 1033 _current = self._current 1034 _end = self._end 1035 _peek = self._peek 1036 1037 while _peek.isalnum(): 1038 _col += 1 1039 _current += 1 1040 _end = _current >= self.size 1041 _peek = "" if _end else self.sql[_current] 1042 1043 self._col = _col 1044 self._current = _current 1045 self._end = _end 1046 self._peek = _peek 1047 self._char = self.sql[_current - 1] 1048 1049 @property 1050 def _text(self) -> str: 1051 return self.sql[self._start : self._current] 1052 1053 def _add(self, token_type: TokenType, text: t.Optional[str] = None) -> None: 1054 self._prev_token_line = self._line 1055 1056 if self._comments and token_type == TokenType.SEMICOLON and self.tokens: 1057 self.tokens[-1].comments.extend(self._comments) 1058 self._comments = [] 1059 1060 self.tokens.append( 1061 Token( 1062 token_type, 1063 text=self._text if text is None else text, 1064 line=self._line, 1065 col=self._col, 1066 start=self._start, 1067 end=self._current - 1, 1068 comments=self._comments, 1069 ) 1070 ) 1071 self._comments = [] 1072 1073 # If we have either a semicolon or a begin token before the command's token, we'll parse 1074 # whatever follows the command's token as a string 1075 if ( 1076 token_type in self.COMMANDS 1077 and self._peek != ";" 1078 and (len(self.tokens) == 1 or self.tokens[-2].token_type in self.COMMAND_PREFIX_TOKENS) 1079 ): 1080 start = self._current 1081 tokens = len(self.tokens) 1082 self._scan(lambda: self._peek == ";") 1083 self.tokens = self.tokens[:tokens] 1084 text = self.sql[start : self._current].strip() 1085 if text: 1086 self._add(TokenType.STRING, text) 1087 1088 def _scan_keywords(self) -> None: 1089 size = 0 1090 word = None 1091 chars = self._text 1092 char = chars 1093 prev_space = False 1094 skip = False 1095 trie = self._KEYWORD_TRIE 1096 single_token = char in self.SINGLE_TOKENS 1097 1098 while chars: 1099 if skip: 1100 result = TrieResult.PREFIX 1101 else: 1102 result, trie = in_trie(trie, char.upper()) 1103 1104 if result == TrieResult.FAILED: 1105 break 1106 if result == TrieResult.EXISTS: 1107 word = chars 1108 1109 end = self._current + size 1110 size += 1 1111 1112 if end < self.size: 1113 char = self.sql[end] 1114 single_token = single_token or char in self.SINGLE_TOKENS 1115 is_space = char.isspace() 1116 1117 if not is_space or not prev_space: 1118 if is_space: 1119 char = " " 1120 chars += char 1121 prev_space = is_space 1122 skip = False 1123 else: 1124 skip = True 1125 else: 1126 char = "" 1127 break 1128 1129 if word: 1130 if self._scan_string(word): 1131 return 1132 if self._scan_comment(word): 1133 return 1134 if prev_space or single_token or not char: 1135 self._advance(size - 1) 1136 word = word.upper() 1137 self._add(self.KEYWORDS[word], text=word) 1138 return 1139 1140 if self._char in self.SINGLE_TOKENS: 1141 self._add(self.SINGLE_TOKENS[self._char], text=self._char) 1142 return 1143 1144 self._scan_var() 1145 1146 def _scan_comment(self, comment_start: str) -> bool: 1147 if comment_start not in self._COMMENTS: 1148 return False 1149 1150 comment_start_line = self._line 1151 comment_start_size = len(comment_start) 1152 comment_end = self._COMMENTS[comment_start] 1153 1154 if comment_end: 1155 # Skip the comment's start delimiter 1156 self._advance(comment_start_size) 1157 1158 comment_end_size = len(comment_end) 1159 while not self._end and self._chars(comment_end_size) != comment_end: 1160 self._advance(alnum=True) 1161 1162 self._comments.append(self._text[comment_start_size : -comment_end_size + 1]) 1163 self._advance(comment_end_size - 1) 1164 else: 1165 while not self._end and self.WHITE_SPACE.get(self._peek) is not TokenType.BREAK: 1166 self._advance(alnum=True) 1167 self._comments.append(self._text[comment_start_size:]) 1168 1169 # Leading comment is attached to the succeeding token, whilst trailing comment to the preceding. 1170 # Multiple consecutive comments are preserved by appending them to the current comments list. 1171 if comment_start_line == self._prev_token_line: 1172 self.tokens[-1].comments.extend(self._comments) 1173 self._comments = [] 1174 self._prev_token_line = self._line 1175 1176 return True 1177 1178 def _scan_number(self) -> None: 1179 if self._char == "0": 1180 peek = self._peek.upper() 1181 if peek == "B": 1182 return self._scan_bits() if self.BIT_STRINGS else self._add(TokenType.NUMBER) 1183 elif peek == "X": 1184 return self._scan_hex() if self.HEX_STRINGS else self._add(TokenType.NUMBER) 1185 1186 decimal = False 1187 scientific = 0 1188 1189 while True: 1190 if self._peek.isdigit(): 1191 self._advance() 1192 elif self._peek == "." and not decimal: 1193 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER: 1194 return self._add(TokenType.NUMBER) 1195 decimal = True 1196 self._advance() 1197 elif self._peek in ("-", "+") and scientific == 1: 1198 scientific += 1 1199 self._advance() 1200 elif self._peek.upper() == "E" and not scientific: 1201 scientific += 1 1202 self._advance() 1203 elif self._peek.isidentifier(): 1204 number_text = self._text 1205 literal = "" 1206 1207 while self._peek.strip() and self._peek not in self.SINGLE_TOKENS: 1208 literal += self._peek 1209 self._advance() 1210 1211 token_type = self.KEYWORDS.get(self.NUMERIC_LITERALS.get(literal.upper(), "")) 1212 1213 if token_type: 1214 self._add(TokenType.NUMBER, number_text) 1215 self._add(TokenType.DCOLON, "::") 1216 return self._add(token_type, literal) 1217 elif self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT: 1218 return self._add(TokenType.VAR) 1219 1220 self._advance(-len(literal)) 1221 return self._add(TokenType.NUMBER, number_text) 1222 else: 1223 return self._add(TokenType.NUMBER) 1224 1225 def _scan_bits(self) -> None: 1226 self._advance() 1227 value = self._extract_value() 1228 try: 1229 # If `value` can't be converted to a binary, fallback to tokenizing it as an identifier 1230 int(value, 2) 1231 self._add(TokenType.BIT_STRING, value[2:]) # Drop the 0b 1232 except ValueError: 1233 self._add(TokenType.IDENTIFIER) 1234 1235 def _scan_hex(self) -> None: 1236 self._advance() 1237 value = self._extract_value() 1238 try: 1239 # If `value` can't be converted to a hex, fallback to tokenizing it as an identifier 1240 int(value, 16) 1241 self._add(TokenType.HEX_STRING, value[2:]) # Drop the 0x 1242 except ValueError: 1243 self._add(TokenType.IDENTIFIER) 1244 1245 def _extract_value(self) -> str: 1246 while True: 1247 char = self._peek.strip() 1248 if char and char not in self.SINGLE_TOKENS: 1249 self._advance(alnum=True) 1250 else: 1251 break 1252 1253 return self._text 1254 1255 def _scan_string(self, start: str) -> bool: 1256 base = None 1257 token_type = TokenType.STRING 1258 1259 if start in self._QUOTES: 1260 end = self._QUOTES[start] 1261 elif start in self._FORMAT_STRINGS: 1262 end, token_type = self._FORMAT_STRINGS[start] 1263 1264 if token_type == TokenType.HEX_STRING: 1265 base = 16 1266 elif token_type == TokenType.BIT_STRING: 1267 base = 2 1268 elif token_type == TokenType.HEREDOC_STRING: 1269 if ( 1270 self.HEREDOC_TAG_IS_IDENTIFIER 1271 and not self._peek.isidentifier() 1272 and not self._peek == end 1273 ): 1274 if self.HEREDOC_STRING_ALTERNATIVE != token_type.VAR: 1275 self._add(self.HEREDOC_STRING_ALTERNATIVE) 1276 else: 1277 self._scan_var() 1278 1279 return True 1280 1281 self._advance() 1282 1283 if self._char == end: 1284 tag = "" 1285 else: 1286 tag = self._extract_string( 1287 end, 1288 unescape_sequences=False, 1289 raise_unmatched=not self.HEREDOC_TAG_IS_IDENTIFIER, 1290 ) 1291 1292 if self._end and tag and self.HEREDOC_TAG_IS_IDENTIFIER: 1293 self._advance(-len(tag)) 1294 self._add(self.HEREDOC_STRING_ALTERNATIVE) 1295 return True 1296 1297 end = f"{start}{tag}{end}" 1298 else: 1299 return False 1300 1301 self._advance(len(start)) 1302 text = self._extract_string(end, unescape_sequences=token_type != TokenType.RAW_STRING) 1303 1304 if base: 1305 try: 1306 int(text, base) 1307 except Exception: 1308 raise TokenError( 1309 f"Numeric string contains invalid characters from {self._line}:{self._start}" 1310 ) 1311 1312 self._add(token_type, text) 1313 return True 1314 1315 def _scan_identifier(self, identifier_end: str) -> None: 1316 self._advance() 1317 text = self._extract_string(identifier_end, escapes=self._IDENTIFIER_ESCAPES) 1318 self._add(TokenType.IDENTIFIER, text) 1319 1320 def _scan_var(self) -> None: 1321 while True: 1322 char = self._peek.strip() 1323 if char and (char in self.VAR_SINGLE_TOKENS or char not in self.SINGLE_TOKENS): 1324 self._advance(alnum=True) 1325 else: 1326 break 1327 1328 self._add( 1329 TokenType.VAR 1330 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER 1331 else self.KEYWORDS.get(self._text.upper(), TokenType.VAR) 1332 ) 1333 1334 def _extract_string( 1335 self, 1336 delimiter: str, 1337 escapes: t.Optional[t.Set[str]] = None, 1338 unescape_sequences: bool = True, 1339 raise_unmatched: bool = True, 1340 ) -> str: 1341 text = "" 1342 delim_size = len(delimiter) 1343 escapes = self._STRING_ESCAPES if escapes is None else escapes 1344 1345 while True: 1346 if ( 1347 unescape_sequences 1348 and self.dialect.UNESCAPED_SEQUENCES 1349 and self._peek 1350 and self._char in self.STRING_ESCAPES 1351 ): 1352 unescaped_sequence = self.dialect.UNESCAPED_SEQUENCES.get(self._char + self._peek) 1353 if unescaped_sequence: 1354 self._advance(2) 1355 text += unescaped_sequence 1356 continue 1357 if ( 1358 self._char in escapes 1359 and (self._peek == delimiter or self._peek in escapes) 1360 and (self._char not in self._QUOTES or self._char == self._peek) 1361 ): 1362 if self._peek == delimiter: 1363 text += self._peek 1364 else: 1365 text += self._char + self._peek 1366 1367 if self._current + 1 < self.size: 1368 self._advance(2) 1369 else: 1370 raise TokenError(f"Missing {delimiter} from {self._line}:{self._current}") 1371 else: 1372 if self._chars(delim_size) == delimiter: 1373 if delim_size > 1: 1374 self._advance(delim_size - 1) 1375 break 1376 1377 if self._end: 1378 if not raise_unmatched: 1379 return text + self._char 1380 1381 raise TokenError(f"Missing {delimiter} from {self._line}:{self._start}") 1382 1383 current = self._current - 1 1384 self._advance(alnum=True) 1385 text += self.sql[current : self._current - 1] 1386 1387 return text 1388 1389 def tokenize_rs(self, sql: str) -> t.List[Token]: 1390 if not self._RS_TOKENIZER: 1391 raise SqlglotError("Rust tokenizer is not available") 1392 1393 try: 1394 tokens = self._RS_TOKENIZER.tokenize(sql, self._rs_dialect_settings) 1395 for token in tokens: 1396 token.token_type = _ALL_TOKEN_TYPES[token.token_type_index] 1397 return tokens 1398 except Exception as e: 1399 raise TokenError(str(e))
Tokenizer( dialect: Union[str, sqlglot.dialects.dialect.Dialect, Type[sqlglot.dialects.dialect.Dialect], NoneType] = None)
928 def __init__(self, dialect: DialectType = None) -> None: 929 from sqlglot.dialects import Dialect 930 931 self.dialect = Dialect.get_or_raise(dialect) 932 933 if USE_RS_TOKENIZER: 934 self._rs_dialect_settings = RsTokenizerDialectSettings( 935 unescaped_sequences=self.dialect.UNESCAPED_SEQUENCES, 936 identifiers_can_start_with_digit=self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT, 937 ) 938 939 self.reset()
SINGLE_TOKENS =
{'(': <TokenType.L_PAREN: 'L_PAREN'>, ')': <TokenType.R_PAREN: 'R_PAREN'>, '[': <TokenType.L_BRACKET: 'L_BRACKET'>, ']': <TokenType.R_BRACKET: 'R_BRACKET'>, '{': <TokenType.L_BRACE: 'L_BRACE'>, '}': <TokenType.R_BRACE: 'R_BRACE'>, '&': <TokenType.AMP: 'AMP'>, '^': <TokenType.CARET: 'CARET'>, ':': <TokenType.COLON: 'COLON'>, ',': <TokenType.COMMA: 'COMMA'>, '.': <TokenType.DOT: 'DOT'>, '-': <TokenType.DASH: 'DASH'>, '=': <TokenType.EQ: 'EQ'>, '>': <TokenType.GT: 'GT'>, '<': <TokenType.LT: 'LT'>, '%': <TokenType.MOD: 'MOD'>, '!': <TokenType.NOT: 'NOT'>, '|': <TokenType.PIPE: 'PIPE'>, '+': <TokenType.PLUS: 'PLUS'>, ';': <TokenType.SEMICOLON: 'SEMICOLON'>, '/': <TokenType.SLASH: 'SLASH'>, '\\': <TokenType.BACKSLASH: 'BACKSLASH'>, '*': <TokenType.STAR: 'STAR'>, '~': <TokenType.TILDA: 'TILDA'>, '?': <TokenType.PLACEHOLDER: 'PLACEHOLDER'>, '@': <TokenType.PARAMETER: 'PARAMETER'>, '#': <TokenType.HASH: 'HASH'>, "'": <TokenType.UNKNOWN: 'UNKNOWN'>, '`': <TokenType.UNKNOWN: 'UNKNOWN'>, '"': <TokenType.UNKNOWN: 'UNKNOWN'>}
HEREDOC_STRING_ALTERNATIVE =
<TokenType.VAR: 'VAR'>
KEYWORDS: Dict[str, TokenType] =
{'{%': <TokenType.BLOCK_START: 'BLOCK_START'>, '{%+': <TokenType.BLOCK_START: 'BLOCK_START'>, '{%-': <TokenType.BLOCK_START: 'BLOCK_START'>, '%}': <TokenType.BLOCK_END: 'BLOCK_END'>, '+%}': <TokenType.BLOCK_END: 'BLOCK_END'>, '-%}': <TokenType.BLOCK_END: 'BLOCK_END'>, '{{+': <TokenType.BLOCK_START: 'BLOCK_START'>, '{{-': <TokenType.BLOCK_START: 'BLOCK_START'>, '+}}': <TokenType.BLOCK_END: 'BLOCK_END'>, '-}}': <TokenType.BLOCK_END: 'BLOCK_END'>, '/*+': <TokenType.HINT: 'HINT'>, '==': <TokenType.EQ: 'EQ'>, '::': <TokenType.DCOLON: 'DCOLON'>, '||': <TokenType.DPIPE: 'DPIPE'>, '>=': <TokenType.GTE: 'GTE'>, '<=': <TokenType.LTE: 'LTE'>, '<>': <TokenType.NEQ: 'NEQ'>, '!=': <TokenType.NEQ: 'NEQ'>, ':=': <TokenType.COLON_EQ: 'COLON_EQ'>, '<=>': <TokenType.NULLSAFE_EQ: 'NULLSAFE_EQ'>, '->': <TokenType.ARROW: 'ARROW'>, '->>': <TokenType.DARROW: 'DARROW'>, '=>': <TokenType.FARROW: 'FARROW'>, '#>': <TokenType.HASH_ARROW: 'HASH_ARROW'>, '#>>': <TokenType.DHASH_ARROW: 'DHASH_ARROW'>, '<->': <TokenType.LR_ARROW: 'LR_ARROW'>, '&&': <TokenType.DAMP: 'DAMP'>, '??': <TokenType.DQMARK: 'DQMARK'>, 'ALL': <TokenType.ALL: 'ALL'>, 'ALWAYS': <TokenType.ALWAYS: 'ALWAYS'>, 'AND': <TokenType.AND: 'AND'>, 'ANTI': <TokenType.ANTI: 'ANTI'>, 'ANY': <TokenType.ANY: 'ANY'>, 'ASC': <TokenType.ASC: 'ASC'>, 'AS': <TokenType.ALIAS: 'ALIAS'>, 'ASOF': <TokenType.ASOF: 'ASOF'>, 'AUTOINCREMENT': <TokenType.AUTO_INCREMENT: 'AUTO_INCREMENT'>, 'AUTO_INCREMENT': <TokenType.AUTO_INCREMENT: 'AUTO_INCREMENT'>, 'BEGIN': <TokenType.BEGIN: 'BEGIN'>, 'BETWEEN': <TokenType.BETWEEN: 'BETWEEN'>, 'CACHE': <TokenType.CACHE: 'CACHE'>, 'UNCACHE': <TokenType.UNCACHE: 'UNCACHE'>, 'CASE': <TokenType.CASE: 'CASE'>, 'CHARACTER SET': <TokenType.CHARACTER_SET: 'CHARACTER_SET'>, 'CLUSTER BY': <TokenType.CLUSTER_BY: 'CLUSTER_BY'>, 'COLLATE': <TokenType.COLLATE: 'COLLATE'>, 'COLUMN': <TokenType.COLUMN: 'COLUMN'>, 'COMMIT': <TokenType.COMMIT: 'COMMIT'>, 'CONNECT BY': <TokenType.CONNECT_BY: 'CONNECT_BY'>, 'CONSTRAINT': <TokenType.CONSTRAINT: 'CONSTRAINT'>, 'COPY': <TokenType.COPY: 'COPY'>, 'CREATE': <TokenType.CREATE: 'CREATE'>, 'CROSS': <TokenType.CROSS: 'CROSS'>, 'CUBE': <TokenType.CUBE: 'CUBE'>, 'CURRENT_DATE': <TokenType.CURRENT_DATE: 'CURRENT_DATE'>, 'CURRENT_TIME': <TokenType.CURRENT_TIME: 'CURRENT_TIME'>, 'CURRENT_TIMESTAMP': <TokenType.CURRENT_TIMESTAMP: 'CURRENT_TIMESTAMP'>, 'CURRENT_USER': <TokenType.CURRENT_USER: 'CURRENT_USER'>, 'DATABASE': <TokenType.DATABASE: 'DATABASE'>, 'DEFAULT': <TokenType.DEFAULT: 'DEFAULT'>, 'DELETE': <TokenType.DELETE: 'DELETE'>, 'DESC': <TokenType.DESC: 'DESC'>, 'DESCRIBE': <TokenType.DESCRIBE: 'DESCRIBE'>, 'DISTINCT': <TokenType.DISTINCT: 'DISTINCT'>, 'DISTRIBUTE BY': <TokenType.DISTRIBUTE_BY: 'DISTRIBUTE_BY'>, 'DIV': <TokenType.DIV: 'DIV'>, 'DROP': <TokenType.DROP: 'DROP'>, 'ELSE': <TokenType.ELSE: 'ELSE'>, 'END': <TokenType.END: 'END'>, 'ENUM': <TokenType.ENUM: 'ENUM'>, 'ESCAPE': <TokenType.ESCAPE: 'ESCAPE'>, 'EXCEPT': <TokenType.EXCEPT: 'EXCEPT'>, 'EXECUTE': <TokenType.EXECUTE: 'EXECUTE'>, 'EXISTS': <TokenType.EXISTS: 'EXISTS'>, 'FALSE': <TokenType.FALSE: 'FALSE'>, 'FETCH': <TokenType.FETCH: 'FETCH'>, 'FILTER': <TokenType.FILTER: 'FILTER'>, 'FIRST': <TokenType.FIRST: 'FIRST'>, 'FULL': <TokenType.FULL: 'FULL'>, 'FUNCTION': <TokenType.FUNCTION: 'FUNCTION'>, 'FOR': <TokenType.FOR: 'FOR'>, 'FOREIGN KEY': <TokenType.FOREIGN_KEY: 'FOREIGN_KEY'>, 'FORMAT': <TokenType.FORMAT: 'FORMAT'>, 'FROM': <TokenType.FROM: 'FROM'>, 'GEOGRAPHY': <TokenType.GEOGRAPHY: 'GEOGRAPHY'>, 'GEOMETRY': <TokenType.GEOMETRY: 'GEOMETRY'>, 'GLOB': <TokenType.GLOB: 'GLOB'>, 'GROUP BY': <TokenType.GROUP_BY: 'GROUP_BY'>, 'GROUPING SETS': <TokenType.GROUPING_SETS: 'GROUPING_SETS'>, 'HAVING': <TokenType.HAVING: 'HAVING'>, 'ILIKE': <TokenType.ILIKE: 'ILIKE'>, 'IN': <TokenType.IN: 'IN'>, 'INDEX': <TokenType.INDEX: 'INDEX'>, 'INET': <TokenType.INET: 'INET'>, 'INNER': <TokenType.INNER: 'INNER'>, 'INSERT': <TokenType.INSERT: 'INSERT'>, 'INTERVAL': <TokenType.INTERVAL: 'INTERVAL'>, 'INTERSECT': <TokenType.INTERSECT: 'INTERSECT'>, 'INTO': <TokenType.INTO: 'INTO'>, 'IS': <TokenType.IS: 'IS'>, 'ISNULL': <TokenType.ISNULL: 'ISNULL'>, 'JOIN': <TokenType.JOIN: 'JOIN'>, 'KEEP': <TokenType.KEEP: 'KEEP'>, 'KILL': <TokenType.KILL: 'KILL'>, 'LATERAL': <TokenType.LATERAL: 'LATERAL'>, 'LEFT': <TokenType.LEFT: 'LEFT'>, 'LIKE': <TokenType.LIKE: 'LIKE'>, 'LIMIT': <TokenType.LIMIT: 'LIMIT'>, 'LOAD': <TokenType.LOAD: 'LOAD'>, 'LOCK': <TokenType.LOCK: 'LOCK'>, 'MERGE': <TokenType.MERGE: 'MERGE'>, 'NATURAL': <TokenType.NATURAL: 'NATURAL'>, 'NEXT': <TokenType.NEXT: 'NEXT'>, 'NOT': <TokenType.NOT: 'NOT'>, 'NOTNULL': <TokenType.NOTNULL: 'NOTNULL'>, 'NULL': <TokenType.NULL: 'NULL'>, 'OBJECT': <TokenType.OBJECT: 'OBJECT'>, 'OFFSET': <TokenType.OFFSET: 'OFFSET'>, 'ON': <TokenType.ON: 'ON'>, 'OR': <TokenType.OR: 'OR'>, 'XOR': <TokenType.XOR: 'XOR'>, 'ORDER BY': <TokenType.ORDER_BY: 'ORDER_BY'>, 'ORDINALITY': <TokenType.ORDINALITY: 'ORDINALITY'>, 'OUTER': <TokenType.OUTER: 'OUTER'>, 'OVER': <TokenType.OVER: 'OVER'>, 'OVERLAPS': <TokenType.OVERLAPS: 'OVERLAPS'>, 'OVERWRITE': <TokenType.OVERWRITE: 'OVERWRITE'>, 'PARTITION': <TokenType.PARTITION: 'PARTITION'>, 'PARTITION BY': <TokenType.PARTITION_BY: 'PARTITION_BY'>, 'PARTITIONED BY': <TokenType.PARTITION_BY: 'PARTITION_BY'>, 'PARTITIONED_BY': <TokenType.PARTITION_BY: 'PARTITION_BY'>, 'PERCENT': <TokenType.PERCENT: 'PERCENT'>, 'PIVOT': <TokenType.PIVOT: 'PIVOT'>, 'PRAGMA': <TokenType.PRAGMA: 'PRAGMA'>, 'PRIMARY KEY': <TokenType.PRIMARY_KEY: 'PRIMARY_KEY'>, 'PROCEDURE': <TokenType.PROCEDURE: 'PROCEDURE'>, 'QUALIFY': <TokenType.QUALIFY: 'QUALIFY'>, 'RANGE': <TokenType.RANGE: 'RANGE'>, 'RECURSIVE': <TokenType.RECURSIVE: 'RECURSIVE'>, 'REGEXP': <TokenType.RLIKE: 'RLIKE'>, 'REPLACE': <TokenType.REPLACE: 'REPLACE'>, 'RETURNING': <TokenType.RETURNING: 'RETURNING'>, 'REFERENCES': <TokenType.REFERENCES: 'REFERENCES'>, 'RIGHT': <TokenType.RIGHT: 'RIGHT'>, 'RLIKE': <TokenType.RLIKE: 'RLIKE'>, 'ROLLBACK': <TokenType.ROLLBACK: 'ROLLBACK'>, 'ROLLUP': <TokenType.ROLLUP: 'ROLLUP'>, 'ROW': <TokenType.ROW: 'ROW'>, 'ROWS': <TokenType.ROWS: 'ROWS'>, 'SCHEMA': <TokenType.SCHEMA: 'SCHEMA'>, 'SELECT': <TokenType.SELECT: 'SELECT'>, 'SEMI': <TokenType.SEMI: 'SEMI'>, 'SET': <TokenType.SET: 'SET'>, 'SETTINGS': <TokenType.SETTINGS: 'SETTINGS'>, 'SHOW': <TokenType.SHOW: 'SHOW'>, 'SIMILAR TO': <TokenType.SIMILAR_TO: 'SIMILAR_TO'>, 'SOME': <TokenType.SOME: 'SOME'>, 'SORT BY': <TokenType.SORT_BY: 'SORT_BY'>, 'START WITH': <TokenType.START_WITH: 'START_WITH'>, 'TABLE': <TokenType.TABLE: 'TABLE'>, 'TABLESAMPLE': <TokenType.TABLE_SAMPLE: 'TABLE_SAMPLE'>, 'TEMP': <TokenType.TEMPORARY: 'TEMPORARY'>, 'TEMPORARY': <TokenType.TEMPORARY: 'TEMPORARY'>, 'THEN': <TokenType.THEN: 'THEN'>, 'TRUE': <TokenType.TRUE: 'TRUE'>, 'TRUNCATE': <TokenType.TRUNCATE: 'TRUNCATE'>, 'UNION': <TokenType.UNION: 'UNION'>, 'UNKNOWN': <TokenType.UNKNOWN: 'UNKNOWN'>, 'UNNEST': <TokenType.UNNEST: 'UNNEST'>, 'UNPIVOT': <TokenType.UNPIVOT: 'UNPIVOT'>, 'UPDATE': <TokenType.UPDATE: 'UPDATE'>, 'USE': <TokenType.USE: 'USE'>, 'USING': <TokenType.USING: 'USING'>, 'UUID': <TokenType.UUID: 'UUID'>, 'VALUES': <TokenType.VALUES: 'VALUES'>, 'VIEW': <TokenType.VIEW: 'VIEW'>, 'VOLATILE': <TokenType.VOLATILE: 'VOLATILE'>, 'WHEN': <TokenType.WHEN: 'WHEN'>, 'WHERE': <TokenType.WHERE: 'WHERE'>, 'WINDOW': <TokenType.WINDOW: 'WINDOW'>, 'WITH': <TokenType.WITH: 'WITH'>, 'APPLY': <TokenType.APPLY: 'APPLY'>, 'ARRAY': <TokenType.ARRAY: 'ARRAY'>, 'BIT': <TokenType.BIT: 'BIT'>, 'BOOL': <TokenType.BOOLEAN: 'BOOLEAN'>, 'BOOLEAN': <TokenType.BOOLEAN: 'BOOLEAN'>, 'BYTE': <TokenType.TINYINT: 'TINYINT'>, 'MEDIUMINT': <TokenType.MEDIUMINT: 'MEDIUMINT'>, 'INT1': <TokenType.TINYINT: 'TINYINT'>, 'TINYINT': <TokenType.TINYINT: 'TINYINT'>, 'INT16': <TokenType.SMALLINT: 'SMALLINT'>, 'SHORT': <TokenType.SMALLINT: 'SMALLINT'>, 'SMALLINT': <TokenType.SMALLINT: 'SMALLINT'>, 'INT128': <TokenType.INT128: 'INT128'>, 'HUGEINT': <TokenType.INT128: 'INT128'>, 'INT2': <TokenType.SMALLINT: 'SMALLINT'>, 'INTEGER': <TokenType.INT: 'INT'>, 'INT': <TokenType.INT: 'INT'>, 'INT4': <TokenType.INT: 'INT'>, 'INT32': <TokenType.INT: 'INT'>, 'INT64': <TokenType.BIGINT: 'BIGINT'>, 'LONG': <TokenType.BIGINT: 'BIGINT'>, 'BIGINT': <TokenType.BIGINT: 'BIGINT'>, 'INT8': <TokenType.TINYINT: 'TINYINT'>, 'UINT': <TokenType.UINT: 'UINT'>, 'DEC': <TokenType.DECIMAL: 'DECIMAL'>, 'DECIMAL': <TokenType.DECIMAL: 'DECIMAL'>, 'BIGDECIMAL': <TokenType.BIGDECIMAL: 'BIGDECIMAL'>, 'BIGNUMERIC': <TokenType.BIGDECIMAL: 'BIGDECIMAL'>, 'MAP': <TokenType.MAP: 'MAP'>, 'NULLABLE': <TokenType.NULLABLE: 'NULLABLE'>, 'NUMBER': <TokenType.DECIMAL: 'DECIMAL'>, 'NUMERIC': <TokenType.DECIMAL: 'DECIMAL'>, 'FIXED': <TokenType.DECIMAL: 'DECIMAL'>, 'REAL': <TokenType.FLOAT: 'FLOAT'>, 'FLOAT': <TokenType.FLOAT: 'FLOAT'>, 'FLOAT4': <TokenType.FLOAT: 'FLOAT'>, 'FLOAT8': <TokenType.DOUBLE: 'DOUBLE'>, 'DOUBLE': <TokenType.DOUBLE: 'DOUBLE'>, 'DOUBLE PRECISION': <TokenType.DOUBLE: 'DOUBLE'>, 'JSON': <TokenType.JSON: 'JSON'>, 'CHAR': <TokenType.CHAR: 'CHAR'>, 'CHARACTER': <TokenType.CHAR: 'CHAR'>, 'NCHAR': <TokenType.NCHAR: 'NCHAR'>, 'VARCHAR': <TokenType.VARCHAR: 'VARCHAR'>, 'VARCHAR2': <TokenType.VARCHAR: 'VARCHAR'>, 'NVARCHAR': <TokenType.NVARCHAR: 'NVARCHAR'>, 'NVARCHAR2': <TokenType.NVARCHAR: 'NVARCHAR'>, 'BPCHAR': <TokenType.BPCHAR: 'BPCHAR'>, 'STR': <TokenType.TEXT: 'TEXT'>, 'STRING': <TokenType.TEXT: 'TEXT'>, 'TEXT': <TokenType.TEXT: 'TEXT'>, 'LONGTEXT': <TokenType.LONGTEXT: 'LONGTEXT'>, 'MEDIUMTEXT': <TokenType.MEDIUMTEXT: 'MEDIUMTEXT'>, 'TINYTEXT': <TokenType.TINYTEXT: 'TINYTEXT'>, 'CLOB': <TokenType.TEXT: 'TEXT'>, 'LONGVARCHAR': <TokenType.TEXT: 'TEXT'>, 'BINARY': <TokenType.BINARY: 'BINARY'>, 'BLOB': <TokenType.VARBINARY: 'VARBINARY'>, 'LONGBLOB': <TokenType.LONGBLOB: 'LONGBLOB'>, 'MEDIUMBLOB': <TokenType.MEDIUMBLOB: 'MEDIUMBLOB'>, 'TINYBLOB': <TokenType.TINYBLOB: 'TINYBLOB'>, 'BYTEA': <TokenType.VARBINARY: 'VARBINARY'>, 'VARBINARY': <TokenType.VARBINARY: 'VARBINARY'>, 'TIME': <TokenType.TIME: 'TIME'>, 'TIMETZ': <TokenType.TIMETZ: 'TIMETZ'>, 'TIMESTAMP': <TokenType.TIMESTAMP: 'TIMESTAMP'>, 'TIMESTAMPTZ': <TokenType.TIMESTAMPTZ: 'TIMESTAMPTZ'>, 'TIMESTAMPLTZ': <TokenType.TIMESTAMPLTZ: 'TIMESTAMPLTZ'>, 'TIMESTAMP_LTZ': <TokenType.TIMESTAMPLTZ: 'TIMESTAMPLTZ'>, 'TIMESTAMPNTZ': <TokenType.TIMESTAMPNTZ: 'TIMESTAMPNTZ'>, 'TIMESTAMP_NTZ': <TokenType.TIMESTAMPNTZ: 'TIMESTAMPNTZ'>, 'DATE': <TokenType.DATE: 'DATE'>, 'DATETIME': <TokenType.DATETIME: 'DATETIME'>, 'INT4RANGE': <TokenType.INT4RANGE: 'INT4RANGE'>, 'INT4MULTIRANGE': <TokenType.INT4MULTIRANGE: 'INT4MULTIRANGE'>, 'INT8RANGE': <TokenType.INT8RANGE: 'INT8RANGE'>, 'INT8MULTIRANGE': <TokenType.INT8MULTIRANGE: 'INT8MULTIRANGE'>, 'NUMRANGE': <TokenType.NUMRANGE: 'NUMRANGE'>, 'NUMMULTIRANGE': <TokenType.NUMMULTIRANGE: 'NUMMULTIRANGE'>, 'TSRANGE': <TokenType.TSRANGE: 'TSRANGE'>, 'TSMULTIRANGE': <TokenType.TSMULTIRANGE: 'TSMULTIRANGE'>, 'TSTZRANGE': <TokenType.TSTZRANGE: 'TSTZRANGE'>, 'TSTZMULTIRANGE': <TokenType.TSTZMULTIRANGE: 'TSTZMULTIRANGE'>, 'DATERANGE': <TokenType.DATERANGE: 'DATERANGE'>, 'DATEMULTIRANGE': <TokenType.DATEMULTIRANGE: 'DATEMULTIRANGE'>, 'UNIQUE': <TokenType.UNIQUE: 'UNIQUE'>, 'STRUCT': <TokenType.STRUCT: 'STRUCT'>, 'SEQUENCE': <TokenType.SEQUENCE: 'SEQUENCE'>, 'VARIANT': <TokenType.VARIANT: 'VARIANT'>, 'ALTER': <TokenType.ALTER: 'ALTER'>, 'ANALYZE': <TokenType.COMMAND: 'COMMAND'>, 'CALL': <TokenType.COMMAND: 'COMMAND'>, 'COMMENT': <TokenType.COMMENT: 'COMMENT'>, 'EXPLAIN': <TokenType.COMMAND: 'COMMAND'>, 'GRANT': <TokenType.COMMAND: 'COMMAND'>, 'OPTIMIZE': <TokenType.COMMAND: 'COMMAND'>, 'PREPARE': <TokenType.COMMAND: 'COMMAND'>, 'VACUUM': <TokenType.COMMAND: 'COMMAND'>, 'USER-DEFINED': <TokenType.USERDEFINED: 'USERDEFINED'>, 'FOR VERSION': <TokenType.VERSION_SNAPSHOT: 'VERSION_SNAPSHOT'>, 'FOR TIMESTAMP': <TokenType.TIMESTAMP_SNAPSHOT: 'TIMESTAMP_SNAPSHOT'>}
WHITE_SPACE: Dict[Optional[str], TokenType] =
{' ': <TokenType.SPACE: 'SPACE'>, '\t': <TokenType.SPACE: 'SPACE'>, '\n': <TokenType.BREAK: 'BREAK'>, '\r': <TokenType.BREAK: 'BREAK'>}
COMMANDS =
{<TokenType.SHOW: 'SHOW'>, <TokenType.EXECUTE: 'EXECUTE'>, <TokenType.FETCH: 'FETCH'>, <TokenType.COMMAND: 'COMMAND'>}
COMMAND_PREFIX_TOKENS =
{<TokenType.BEGIN: 'BEGIN'>, <TokenType.SEMICOLON: 'SEMICOLON'>}
def
reset(self) -> None:
941 def reset(self) -> None: 942 self.sql = "" 943 self.size = 0 944 self.tokens: t.List[Token] = [] 945 self._start = 0 946 self._current = 0 947 self._line = 1 948 self._col = 0 949 self._comments: t.List[str] = [] 950 951 self._char = "" 952 self._end = False 953 self._peek = "" 954 self._prev_token_line = -1
956 def tokenize(self, sql: str) -> t.List[Token]: 957 """Returns a list of tokens corresponding to the SQL string `sql`.""" 958 if USE_RS_TOKENIZER: 959 return self.tokenize_rs(sql) 960 961 self.reset() 962 self.sql = sql 963 self.size = len(sql) 964 965 try: 966 self._scan() 967 except Exception as e: 968 start = max(self._current - 50, 0) 969 end = min(self._current + 50, self.size - 1) 970 context = self.sql[start:end] 971 raise TokenError(f"Error tokenizing '{context}'") from e 972 973 return self.tokens
Returns a list of tokens corresponding to the SQL string sql
.
1389 def tokenize_rs(self, sql: str) -> t.List[Token]: 1390 if not self._RS_TOKENIZER: 1391 raise SqlglotError("Rust tokenizer is not available") 1392 1393 try: 1394 tokens = self._RS_TOKENIZER.tokenize(sql, self._rs_dialect_settings) 1395 for token in tokens: 1396 token.token_type = _ALL_TOKEN_TYPES[token.token_type_index] 1397 return tokens 1398 except Exception as e: 1399 raise TokenError(str(e))