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 CARET_AT = 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 WAREHOUSE = auto() 97 STREAMLIT = auto() 98 VAR = auto() 99 BIT_STRING = auto() 100 HEX_STRING = auto() 101 BYTE_STRING = auto() 102 NATIONAL_STRING = auto() 103 RAW_STRING = auto() 104 HEREDOC_STRING = auto() 105 UNICODE_STRING = auto() 106 107 # types 108 BIT = auto() 109 BOOLEAN = auto() 110 TINYINT = auto() 111 UTINYINT = auto() 112 SMALLINT = auto() 113 USMALLINT = auto() 114 MEDIUMINT = auto() 115 UMEDIUMINT = auto() 116 INT = auto() 117 UINT = auto() 118 BIGINT = auto() 119 UBIGINT = auto() 120 INT128 = auto() 121 UINT128 = auto() 122 INT256 = auto() 123 UINT256 = auto() 124 FLOAT = auto() 125 DOUBLE = auto() 126 DECIMAL = auto() 127 DECIMAL32 = auto() 128 DECIMAL64 = auto() 129 DECIMAL128 = auto() 130 DECIMAL256 = auto() 131 UDECIMAL = auto() 132 BIGDECIMAL = auto() 133 CHAR = auto() 134 NCHAR = auto() 135 VARCHAR = auto() 136 NVARCHAR = auto() 137 BPCHAR = auto() 138 TEXT = auto() 139 MEDIUMTEXT = auto() 140 LONGTEXT = auto() 141 MEDIUMBLOB = auto() 142 LONGBLOB = auto() 143 TINYBLOB = auto() 144 TINYTEXT = auto() 145 NAME = auto() 146 BINARY = auto() 147 VARBINARY = auto() 148 JSON = auto() 149 JSONB = auto() 150 TIME = auto() 151 TIMETZ = auto() 152 TIMESTAMP = auto() 153 TIMESTAMPTZ = auto() 154 TIMESTAMPLTZ = auto() 155 TIMESTAMPNTZ = auto() 156 TIMESTAMP_S = auto() 157 TIMESTAMP_MS = auto() 158 TIMESTAMP_NS = auto() 159 DATETIME = auto() 160 DATETIME2 = auto() 161 DATETIME64 = auto() 162 SMALLDATETIME = auto() 163 DATE = auto() 164 DATE32 = auto() 165 INT4RANGE = auto() 166 INT4MULTIRANGE = auto() 167 INT8RANGE = auto() 168 INT8MULTIRANGE = auto() 169 NUMRANGE = auto() 170 NUMMULTIRANGE = auto() 171 TSRANGE = auto() 172 TSMULTIRANGE = auto() 173 TSTZRANGE = auto() 174 TSTZMULTIRANGE = auto() 175 DATERANGE = auto() 176 DATEMULTIRANGE = auto() 177 UUID = auto() 178 GEOGRAPHY = auto() 179 NULLABLE = auto() 180 GEOMETRY = auto() 181 POINT = auto() 182 RING = auto() 183 LINESTRING = auto() 184 MULTILINESTRING = auto() 185 POLYGON = auto() 186 MULTIPOLYGON = auto() 187 HLLSKETCH = auto() 188 HSTORE = auto() 189 SUPER = auto() 190 SERIAL = auto() 191 SMALLSERIAL = auto() 192 BIGSERIAL = auto() 193 XML = auto() 194 YEAR = auto() 195 USERDEFINED = auto() 196 MONEY = auto() 197 SMALLMONEY = auto() 198 ROWVERSION = auto() 199 IMAGE = auto() 200 VARIANT = auto() 201 OBJECT = auto() 202 INET = auto() 203 IPADDRESS = auto() 204 IPPREFIX = auto() 205 IPV4 = auto() 206 IPV6 = auto() 207 ENUM = auto() 208 ENUM8 = auto() 209 ENUM16 = auto() 210 FIXEDSTRING = auto() 211 LOWCARDINALITY = auto() 212 NESTED = auto() 213 AGGREGATEFUNCTION = auto() 214 SIMPLEAGGREGATEFUNCTION = auto() 215 TDIGEST = auto() 216 UNKNOWN = auto() 217 VECTOR = auto() 218 DYNAMIC = auto() 219 220 # keywords 221 ALIAS = auto() 222 ALTER = auto() 223 ALWAYS = auto() 224 ALL = auto() 225 ANTI = auto() 226 ANY = auto() 227 APPLY = auto() 228 ARRAY = auto() 229 ASC = auto() 230 ASOF = auto() 231 ATTACH = auto() 232 AUTO_INCREMENT = auto() 233 BEGIN = auto() 234 BETWEEN = auto() 235 BULK_COLLECT_INTO = auto() 236 CACHE = auto() 237 CASE = auto() 238 CHARACTER_SET = auto() 239 CLUSTER_BY = auto() 240 COLLATE = auto() 241 COMMAND = auto() 242 COMMENT = auto() 243 COMMIT = auto() 244 CONNECT_BY = auto() 245 CONSTRAINT = auto() 246 COPY = auto() 247 CREATE = auto() 248 CROSS = auto() 249 CUBE = auto() 250 CURRENT_DATE = auto() 251 CURRENT_DATETIME = auto() 252 CURRENT_SCHEMA = auto() 253 CURRENT_TIME = auto() 254 CURRENT_TIMESTAMP = auto() 255 CURRENT_USER = auto() 256 DECLARE = auto() 257 DEFAULT = auto() 258 DELETE = auto() 259 DESC = auto() 260 DESCRIBE = auto() 261 DETACH = auto() 262 DICTIONARY = auto() 263 DISTINCT = auto() 264 DISTRIBUTE_BY = auto() 265 DIV = auto() 266 DROP = auto() 267 ELSE = auto() 268 END = auto() 269 ESCAPE = auto() 270 EXCEPT = auto() 271 EXECUTE = auto() 272 EXISTS = auto() 273 FALSE = auto() 274 FETCH = auto() 275 FILTER = auto() 276 FINAL = auto() 277 FIRST = auto() 278 FOR = auto() 279 FORCE = auto() 280 FOREIGN_KEY = auto() 281 FORMAT = auto() 282 FROM = auto() 283 FULL = auto() 284 FUNCTION = auto() 285 GLOB = auto() 286 GLOBAL = auto() 287 GRANT = auto() 288 GROUP_BY = auto() 289 GROUPING_SETS = auto() 290 HAVING = auto() 291 HINT = auto() 292 IGNORE = auto() 293 ILIKE = auto() 294 ILIKE_ANY = auto() 295 IN = auto() 296 INDEX = auto() 297 INNER = auto() 298 INSERT = auto() 299 INTERSECT = auto() 300 INTERVAL = auto() 301 INTO = auto() 302 INTRODUCER = auto() 303 IRLIKE = auto() 304 IS = auto() 305 ISNULL = auto() 306 JOIN = auto() 307 JOIN_MARKER = auto() 308 KEEP = auto() 309 KEY = auto() 310 KILL = auto() 311 LANGUAGE = auto() 312 LATERAL = auto() 313 LEFT = auto() 314 LIKE = auto() 315 LIKE_ANY = auto() 316 LIMIT = auto() 317 LIST = auto() 318 LOAD = auto() 319 LOCK = auto() 320 MAP = auto() 321 MATCH_CONDITION = auto() 322 MATCH_RECOGNIZE = auto() 323 MEMBER_OF = auto() 324 MERGE = auto() 325 MOD = auto() 326 MODEL = auto() 327 NATURAL = auto() 328 NEXT = auto() 329 NOTNULL = auto() 330 NULL = auto() 331 OBJECT_IDENTIFIER = auto() 332 OFFSET = auto() 333 ON = auto() 334 ONLY = auto() 335 OPERATOR = auto() 336 ORDER_BY = auto() 337 ORDER_SIBLINGS_BY = auto() 338 ORDERED = auto() 339 ORDINALITY = auto() 340 OUTER = auto() 341 OVER = auto() 342 OVERLAPS = auto() 343 OVERWRITE = auto() 344 PARTITION = auto() 345 PARTITION_BY = auto() 346 PERCENT = auto() 347 PIVOT = auto() 348 PLACEHOLDER = auto() 349 POSITIONAL = auto() 350 PRAGMA = auto() 351 PREWHERE = auto() 352 PRIMARY_KEY = auto() 353 PROCEDURE = auto() 354 PROPERTIES = auto() 355 PSEUDO_TYPE = auto() 356 QUALIFY = auto() 357 QUOTE = auto() 358 RANGE = auto() 359 RECURSIVE = auto() 360 REFRESH = auto() 361 RENAME = auto() 362 REPLACE = auto() 363 RETURNING = auto() 364 REFERENCES = auto() 365 RIGHT = auto() 366 RLIKE = auto() 367 ROLLBACK = auto() 368 ROLLUP = auto() 369 ROW = auto() 370 ROWS = auto() 371 SELECT = auto() 372 SEMI = auto() 373 SEPARATOR = auto() 374 SEQUENCE = auto() 375 SERDE_PROPERTIES = auto() 376 SET = auto() 377 SETTINGS = auto() 378 SHOW = auto() 379 SIMILAR_TO = auto() 380 SOME = auto() 381 SORT_BY = auto() 382 START_WITH = auto() 383 STORAGE_INTEGRATION = auto() 384 STRAIGHT_JOIN = auto() 385 STRUCT = auto() 386 SUMMARIZE = auto() 387 TABLE_SAMPLE = auto() 388 TAG = auto() 389 TEMPORARY = auto() 390 TOP = auto() 391 THEN = auto() 392 TRUE = auto() 393 TRUNCATE = auto() 394 UNCACHE = auto() 395 UNION = auto() 396 UNNEST = auto() 397 UNPIVOT = auto() 398 UPDATE = auto() 399 USE = auto() 400 USING = auto() 401 VALUES = auto() 402 VIEW = auto() 403 VOLATILE = auto() 404 WHEN = auto() 405 WHERE = auto() 406 WINDOW = auto() 407 WITH = auto() 408 UNIQUE = auto() 409 VERSION_SNAPSHOT = auto() 410 TIMESTAMP_SNAPSHOT = auto() 411 OPTION = auto() 412 SINK = auto() 413 SOURCE = auto() 414 ANALYZE = auto() 415 NAMESPACE = auto() 416 EXPORT = auto() 417 418 419_ALL_TOKEN_TYPES = list(TokenType) 420_TOKEN_TYPE_TO_INDEX = {token_type: i for i, token_type in enumerate(_ALL_TOKEN_TYPES)} 421 422 423class Token: 424 __slots__ = ("token_type", "text", "line", "col", "start", "end", "comments") 425 426 @classmethod 427 def number(cls, number: int) -> Token: 428 """Returns a NUMBER token with `number` as its text.""" 429 return cls(TokenType.NUMBER, str(number)) 430 431 @classmethod 432 def string(cls, string: str) -> Token: 433 """Returns a STRING token with `string` as its text.""" 434 return cls(TokenType.STRING, string) 435 436 @classmethod 437 def identifier(cls, identifier: str) -> Token: 438 """Returns an IDENTIFIER token with `identifier` as its text.""" 439 return cls(TokenType.IDENTIFIER, identifier) 440 441 @classmethod 442 def var(cls, var: str) -> Token: 443 """Returns an VAR token with `var` as its text.""" 444 return cls(TokenType.VAR, var) 445 446 def __init__( 447 self, 448 token_type: TokenType, 449 text: str, 450 line: int = 1, 451 col: int = 1, 452 start: int = 0, 453 end: int = 0, 454 comments: t.Optional[t.List[str]] = None, 455 ) -> None: 456 """Token initializer. 457 458 Args: 459 token_type: The TokenType Enum. 460 text: The text of the token. 461 line: The line that the token ends on. 462 col: The column that the token ends on. 463 start: The start index of the token. 464 end: The ending index of the token. 465 comments: The comments to attach to the token. 466 """ 467 self.token_type = token_type 468 self.text = text 469 self.line = line 470 self.col = col 471 self.start = start 472 self.end = end 473 self.comments = [] if comments is None else comments 474 475 def __repr__(self) -> str: 476 attributes = ", ".join(f"{k}: {getattr(self, k)}" for k in self.__slots__) 477 return f"<Token {attributes}>" 478 479 480class _Tokenizer(type): 481 def __new__(cls, clsname, bases, attrs): 482 klass = super().__new__(cls, clsname, bases, attrs) 483 484 def _convert_quotes(arr: t.List[str | t.Tuple[str, str]]) -> t.Dict[str, str]: 485 return dict( 486 (item, item) if isinstance(item, str) else (item[0], item[1]) for item in arr 487 ) 488 489 def _quotes_to_format( 490 token_type: TokenType, arr: t.List[str | t.Tuple[str, str]] 491 ) -> t.Dict[str, t.Tuple[str, TokenType]]: 492 return {k: (v, token_type) for k, v in _convert_quotes(arr).items()} 493 494 klass._QUOTES = _convert_quotes(klass.QUOTES) 495 klass._IDENTIFIERS = _convert_quotes(klass.IDENTIFIERS) 496 497 klass._FORMAT_STRINGS = { 498 **{ 499 p + s: (e, TokenType.NATIONAL_STRING) 500 for s, e in klass._QUOTES.items() 501 for p in ("n", "N") 502 }, 503 **_quotes_to_format(TokenType.BIT_STRING, klass.BIT_STRINGS), 504 **_quotes_to_format(TokenType.BYTE_STRING, klass.BYTE_STRINGS), 505 **_quotes_to_format(TokenType.HEX_STRING, klass.HEX_STRINGS), 506 **_quotes_to_format(TokenType.RAW_STRING, klass.RAW_STRINGS), 507 **_quotes_to_format(TokenType.HEREDOC_STRING, klass.HEREDOC_STRINGS), 508 **_quotes_to_format(TokenType.UNICODE_STRING, klass.UNICODE_STRINGS), 509 } 510 511 klass._STRING_ESCAPES = set(klass.STRING_ESCAPES) 512 klass._IDENTIFIER_ESCAPES = set(klass.IDENTIFIER_ESCAPES) 513 klass._COMMENTS = { 514 **dict( 515 (comment, None) if isinstance(comment, str) else (comment[0], comment[1]) 516 for comment in klass.COMMENTS 517 ), 518 "{#": "#}", # Ensure Jinja comments are tokenized correctly in all dialects 519 } 520 if klass.HINT_START in klass.KEYWORDS: 521 klass._COMMENTS[klass.HINT_START] = "*/" 522 523 klass._KEYWORD_TRIE = new_trie( 524 key.upper() 525 for key in ( 526 *klass.KEYWORDS, 527 *klass._COMMENTS, 528 *klass._QUOTES, 529 *klass._FORMAT_STRINGS, 530 ) 531 if " " in key or any(single in key for single in klass.SINGLE_TOKENS) 532 ) 533 534 if USE_RS_TOKENIZER: 535 settings = RsTokenizerSettings( 536 white_space={k: _TOKEN_TYPE_TO_INDEX[v] for k, v in klass.WHITE_SPACE.items()}, 537 single_tokens={k: _TOKEN_TYPE_TO_INDEX[v] for k, v in klass.SINGLE_TOKENS.items()}, 538 keywords={k: _TOKEN_TYPE_TO_INDEX[v] for k, v in klass.KEYWORDS.items()}, 539 numeric_literals=klass.NUMERIC_LITERALS, 540 identifiers=klass._IDENTIFIERS, 541 identifier_escapes=klass._IDENTIFIER_ESCAPES, 542 string_escapes=klass._STRING_ESCAPES, 543 quotes=klass._QUOTES, 544 format_strings={ 545 k: (v1, _TOKEN_TYPE_TO_INDEX[v2]) 546 for k, (v1, v2) in klass._FORMAT_STRINGS.items() 547 }, 548 has_bit_strings=bool(klass.BIT_STRINGS), 549 has_hex_strings=bool(klass.HEX_STRINGS), 550 comments=klass._COMMENTS, 551 var_single_tokens=klass.VAR_SINGLE_TOKENS, 552 commands={_TOKEN_TYPE_TO_INDEX[v] for v in klass.COMMANDS}, 553 command_prefix_tokens={ 554 _TOKEN_TYPE_TO_INDEX[v] for v in klass.COMMAND_PREFIX_TOKENS 555 }, 556 heredoc_tag_is_identifier=klass.HEREDOC_TAG_IS_IDENTIFIER, 557 string_escapes_allowed_in_raw_strings=klass.STRING_ESCAPES_ALLOWED_IN_RAW_STRINGS, 558 nested_comments=klass.NESTED_COMMENTS, 559 hint_start=klass.HINT_START, 560 tokens_preceding_hint={ 561 _TOKEN_TYPE_TO_INDEX[v] for v in klass.TOKENS_PRECEDING_HINT 562 }, 563 ) 564 token_types = RsTokenTypeSettings( 565 bit_string=_TOKEN_TYPE_TO_INDEX[TokenType.BIT_STRING], 566 break_=_TOKEN_TYPE_TO_INDEX[TokenType.BREAK], 567 dcolon=_TOKEN_TYPE_TO_INDEX[TokenType.DCOLON], 568 heredoc_string=_TOKEN_TYPE_TO_INDEX[TokenType.HEREDOC_STRING], 569 raw_string=_TOKEN_TYPE_TO_INDEX[TokenType.RAW_STRING], 570 hex_string=_TOKEN_TYPE_TO_INDEX[TokenType.HEX_STRING], 571 identifier=_TOKEN_TYPE_TO_INDEX[TokenType.IDENTIFIER], 572 number=_TOKEN_TYPE_TO_INDEX[TokenType.NUMBER], 573 parameter=_TOKEN_TYPE_TO_INDEX[TokenType.PARAMETER], 574 semicolon=_TOKEN_TYPE_TO_INDEX[TokenType.SEMICOLON], 575 string=_TOKEN_TYPE_TO_INDEX[TokenType.STRING], 576 var=_TOKEN_TYPE_TO_INDEX[TokenType.VAR], 577 heredoc_string_alternative=_TOKEN_TYPE_TO_INDEX[klass.HEREDOC_STRING_ALTERNATIVE], 578 hint=_TOKEN_TYPE_TO_INDEX[TokenType.HINT], 579 ) 580 klass._RS_TOKENIZER = RsTokenizer(settings, token_types) 581 else: 582 klass._RS_TOKENIZER = None 583 584 return klass 585 586 587class Tokenizer(metaclass=_Tokenizer): 588 SINGLE_TOKENS = { 589 "(": TokenType.L_PAREN, 590 ")": TokenType.R_PAREN, 591 "[": TokenType.L_BRACKET, 592 "]": TokenType.R_BRACKET, 593 "{": TokenType.L_BRACE, 594 "}": TokenType.R_BRACE, 595 "&": TokenType.AMP, 596 "^": TokenType.CARET, 597 ":": TokenType.COLON, 598 ",": TokenType.COMMA, 599 ".": TokenType.DOT, 600 "-": TokenType.DASH, 601 "=": TokenType.EQ, 602 ">": TokenType.GT, 603 "<": TokenType.LT, 604 "%": TokenType.MOD, 605 "!": TokenType.NOT, 606 "|": TokenType.PIPE, 607 "+": TokenType.PLUS, 608 ";": TokenType.SEMICOLON, 609 "/": TokenType.SLASH, 610 "\\": TokenType.BACKSLASH, 611 "*": TokenType.STAR, 612 "~": TokenType.TILDA, 613 "?": TokenType.PLACEHOLDER, 614 "@": TokenType.PARAMETER, 615 "#": TokenType.HASH, 616 # Used for breaking a var like x'y' but nothing else the token type doesn't matter 617 "'": TokenType.UNKNOWN, 618 "`": TokenType.UNKNOWN, 619 '"': TokenType.UNKNOWN, 620 } 621 622 BIT_STRINGS: t.List[str | t.Tuple[str, str]] = [] 623 BYTE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 624 HEX_STRINGS: t.List[str | t.Tuple[str, str]] = [] 625 RAW_STRINGS: t.List[str | t.Tuple[str, str]] = [] 626 HEREDOC_STRINGS: t.List[str | t.Tuple[str, str]] = [] 627 UNICODE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 628 IDENTIFIERS: t.List[str | t.Tuple[str, str]] = ['"'] 629 QUOTES: t.List[t.Tuple[str, str] | str] = ["'"] 630 STRING_ESCAPES = ["'"] 631 VAR_SINGLE_TOKENS: t.Set[str] = set() 632 633 # The strings in this list can always be used as escapes, regardless of the surrounding 634 # identifier delimiters. By default, the closing delimiter is assumed to also act as an 635 # identifier escape, e.g. if we use double-quotes, then they also act as escapes: "x""" 636 IDENTIFIER_ESCAPES: t.List[str] = [] 637 638 # Whether the heredoc tags follow the same lexical rules as unquoted identifiers 639 HEREDOC_TAG_IS_IDENTIFIER = False 640 641 # Token that we'll generate as a fallback if the heredoc prefix doesn't correspond to a heredoc 642 HEREDOC_STRING_ALTERNATIVE = TokenType.VAR 643 644 # Whether string escape characters function as such when placed within raw strings 645 STRING_ESCAPES_ALLOWED_IN_RAW_STRINGS = True 646 647 NESTED_COMMENTS = True 648 649 HINT_START = "/*+" 650 651 TOKENS_PRECEDING_HINT = {TokenType.SELECT, TokenType.INSERT, TokenType.UPDATE, TokenType.DELETE} 652 653 # Autofilled 654 _COMMENTS: t.Dict[str, str] = {} 655 _FORMAT_STRINGS: t.Dict[str, t.Tuple[str, TokenType]] = {} 656 _IDENTIFIERS: t.Dict[str, str] = {} 657 _IDENTIFIER_ESCAPES: t.Set[str] = set() 658 _QUOTES: t.Dict[str, str] = {} 659 _STRING_ESCAPES: t.Set[str] = set() 660 _KEYWORD_TRIE: t.Dict = {} 661 _RS_TOKENIZER: t.Optional[t.Any] = None 662 663 KEYWORDS: t.Dict[str, TokenType] = { 664 **{f"{{%{postfix}": TokenType.BLOCK_START for postfix in ("", "+", "-")}, 665 **{f"{prefix}%}}": TokenType.BLOCK_END for prefix in ("", "+", "-")}, 666 **{f"{{{{{postfix}": TokenType.BLOCK_START for postfix in ("+", "-")}, 667 **{f"{prefix}}}}}": TokenType.BLOCK_END for prefix in ("+", "-")}, 668 HINT_START: TokenType.HINT, 669 "==": TokenType.EQ, 670 "::": TokenType.DCOLON, 671 "||": TokenType.DPIPE, 672 ">=": TokenType.GTE, 673 "<=": TokenType.LTE, 674 "<>": TokenType.NEQ, 675 "!=": TokenType.NEQ, 676 ":=": TokenType.COLON_EQ, 677 "<=>": TokenType.NULLSAFE_EQ, 678 "->": TokenType.ARROW, 679 "->>": TokenType.DARROW, 680 "=>": TokenType.FARROW, 681 "#>": TokenType.HASH_ARROW, 682 "#>>": TokenType.DHASH_ARROW, 683 "<->": TokenType.LR_ARROW, 684 "&&": TokenType.DAMP, 685 "??": TokenType.DQMARK, 686 "~~~": TokenType.GLOB, 687 "~~": TokenType.LIKE, 688 "~~*": TokenType.ILIKE, 689 "~*": TokenType.IRLIKE, 690 "ALL": TokenType.ALL, 691 "ALWAYS": TokenType.ALWAYS, 692 "AND": TokenType.AND, 693 "ANTI": TokenType.ANTI, 694 "ANY": TokenType.ANY, 695 "ASC": TokenType.ASC, 696 "AS": TokenType.ALIAS, 697 "ASOF": TokenType.ASOF, 698 "AUTOINCREMENT": TokenType.AUTO_INCREMENT, 699 "AUTO_INCREMENT": TokenType.AUTO_INCREMENT, 700 "BEGIN": TokenType.BEGIN, 701 "BETWEEN": TokenType.BETWEEN, 702 "CACHE": TokenType.CACHE, 703 "UNCACHE": TokenType.UNCACHE, 704 "CASE": TokenType.CASE, 705 "CHARACTER SET": TokenType.CHARACTER_SET, 706 "CLUSTER BY": TokenType.CLUSTER_BY, 707 "COLLATE": TokenType.COLLATE, 708 "COLUMN": TokenType.COLUMN, 709 "COMMIT": TokenType.COMMIT, 710 "CONNECT BY": TokenType.CONNECT_BY, 711 "CONSTRAINT": TokenType.CONSTRAINT, 712 "COPY": TokenType.COPY, 713 "CREATE": TokenType.CREATE, 714 "CROSS": TokenType.CROSS, 715 "CUBE": TokenType.CUBE, 716 "CURRENT_DATE": TokenType.CURRENT_DATE, 717 "CURRENT_SCHEMA": TokenType.CURRENT_SCHEMA, 718 "CURRENT_TIME": TokenType.CURRENT_TIME, 719 "CURRENT_TIMESTAMP": TokenType.CURRENT_TIMESTAMP, 720 "CURRENT_USER": TokenType.CURRENT_USER, 721 "DATABASE": TokenType.DATABASE, 722 "DEFAULT": TokenType.DEFAULT, 723 "DELETE": TokenType.DELETE, 724 "DESC": TokenType.DESC, 725 "DESCRIBE": TokenType.DESCRIBE, 726 "DISTINCT": TokenType.DISTINCT, 727 "DISTRIBUTE BY": TokenType.DISTRIBUTE_BY, 728 "DIV": TokenType.DIV, 729 "DROP": TokenType.DROP, 730 "ELSE": TokenType.ELSE, 731 "END": TokenType.END, 732 "ENUM": TokenType.ENUM, 733 "ESCAPE": TokenType.ESCAPE, 734 "EXCEPT": TokenType.EXCEPT, 735 "EXECUTE": TokenType.EXECUTE, 736 "EXISTS": TokenType.EXISTS, 737 "FALSE": TokenType.FALSE, 738 "FETCH": TokenType.FETCH, 739 "FILTER": TokenType.FILTER, 740 "FIRST": TokenType.FIRST, 741 "FULL": TokenType.FULL, 742 "FUNCTION": TokenType.FUNCTION, 743 "FOR": TokenType.FOR, 744 "FOREIGN KEY": TokenType.FOREIGN_KEY, 745 "FORMAT": TokenType.FORMAT, 746 "FROM": TokenType.FROM, 747 "GEOGRAPHY": TokenType.GEOGRAPHY, 748 "GEOMETRY": TokenType.GEOMETRY, 749 "GLOB": TokenType.GLOB, 750 "GROUP BY": TokenType.GROUP_BY, 751 "GROUPING SETS": TokenType.GROUPING_SETS, 752 "HAVING": TokenType.HAVING, 753 "ILIKE": TokenType.ILIKE, 754 "IN": TokenType.IN, 755 "INDEX": TokenType.INDEX, 756 "INET": TokenType.INET, 757 "INNER": TokenType.INNER, 758 "INSERT": TokenType.INSERT, 759 "INTERVAL": TokenType.INTERVAL, 760 "INTERSECT": TokenType.INTERSECT, 761 "INTO": TokenType.INTO, 762 "IS": TokenType.IS, 763 "ISNULL": TokenType.ISNULL, 764 "JOIN": TokenType.JOIN, 765 "KEEP": TokenType.KEEP, 766 "KILL": TokenType.KILL, 767 "LATERAL": TokenType.LATERAL, 768 "LEFT": TokenType.LEFT, 769 "LIKE": TokenType.LIKE, 770 "LIMIT": TokenType.LIMIT, 771 "LOAD": TokenType.LOAD, 772 "LOCK": TokenType.LOCK, 773 "MERGE": TokenType.MERGE, 774 "NAMESPACE": TokenType.NAMESPACE, 775 "NATURAL": TokenType.NATURAL, 776 "NEXT": TokenType.NEXT, 777 "NOT": TokenType.NOT, 778 "NOTNULL": TokenType.NOTNULL, 779 "NULL": TokenType.NULL, 780 "OBJECT": TokenType.OBJECT, 781 "OFFSET": TokenType.OFFSET, 782 "ON": TokenType.ON, 783 "OR": TokenType.OR, 784 "XOR": TokenType.XOR, 785 "ORDER BY": TokenType.ORDER_BY, 786 "ORDINALITY": TokenType.ORDINALITY, 787 "OUTER": TokenType.OUTER, 788 "OVER": TokenType.OVER, 789 "OVERLAPS": TokenType.OVERLAPS, 790 "OVERWRITE": TokenType.OVERWRITE, 791 "PARTITION": TokenType.PARTITION, 792 "PARTITION BY": TokenType.PARTITION_BY, 793 "PARTITIONED BY": TokenType.PARTITION_BY, 794 "PARTITIONED_BY": TokenType.PARTITION_BY, 795 "PERCENT": TokenType.PERCENT, 796 "PIVOT": TokenType.PIVOT, 797 "PRAGMA": TokenType.PRAGMA, 798 "PRIMARY KEY": TokenType.PRIMARY_KEY, 799 "PROCEDURE": TokenType.PROCEDURE, 800 "QUALIFY": TokenType.QUALIFY, 801 "RANGE": TokenType.RANGE, 802 "RECURSIVE": TokenType.RECURSIVE, 803 "REGEXP": TokenType.RLIKE, 804 "RENAME": TokenType.RENAME, 805 "REPLACE": TokenType.REPLACE, 806 "RETURNING": TokenType.RETURNING, 807 "REFERENCES": TokenType.REFERENCES, 808 "RIGHT": TokenType.RIGHT, 809 "RLIKE": TokenType.RLIKE, 810 "ROLLBACK": TokenType.ROLLBACK, 811 "ROLLUP": TokenType.ROLLUP, 812 "ROW": TokenType.ROW, 813 "ROWS": TokenType.ROWS, 814 "SCHEMA": TokenType.SCHEMA, 815 "SELECT": TokenType.SELECT, 816 "SEMI": TokenType.SEMI, 817 "SET": TokenType.SET, 818 "SETTINGS": TokenType.SETTINGS, 819 "SHOW": TokenType.SHOW, 820 "SIMILAR TO": TokenType.SIMILAR_TO, 821 "SOME": TokenType.SOME, 822 "SORT BY": TokenType.SORT_BY, 823 "START WITH": TokenType.START_WITH, 824 "STRAIGHT_JOIN": TokenType.STRAIGHT_JOIN, 825 "TABLE": TokenType.TABLE, 826 "TABLESAMPLE": TokenType.TABLE_SAMPLE, 827 "TEMP": TokenType.TEMPORARY, 828 "TEMPORARY": TokenType.TEMPORARY, 829 "THEN": TokenType.THEN, 830 "TRUE": TokenType.TRUE, 831 "TRUNCATE": TokenType.TRUNCATE, 832 "UNION": TokenType.UNION, 833 "UNKNOWN": TokenType.UNKNOWN, 834 "UNNEST": TokenType.UNNEST, 835 "UNPIVOT": TokenType.UNPIVOT, 836 "UPDATE": TokenType.UPDATE, 837 "USE": TokenType.USE, 838 "USING": TokenType.USING, 839 "UUID": TokenType.UUID, 840 "VALUES": TokenType.VALUES, 841 "VIEW": TokenType.VIEW, 842 "VOLATILE": TokenType.VOLATILE, 843 "WHEN": TokenType.WHEN, 844 "WHERE": TokenType.WHERE, 845 "WINDOW": TokenType.WINDOW, 846 "WITH": TokenType.WITH, 847 "APPLY": TokenType.APPLY, 848 "ARRAY": TokenType.ARRAY, 849 "BIT": TokenType.BIT, 850 "BOOL": TokenType.BOOLEAN, 851 "BOOLEAN": TokenType.BOOLEAN, 852 "BYTE": TokenType.TINYINT, 853 "MEDIUMINT": TokenType.MEDIUMINT, 854 "INT1": TokenType.TINYINT, 855 "TINYINT": TokenType.TINYINT, 856 "INT16": TokenType.SMALLINT, 857 "SHORT": TokenType.SMALLINT, 858 "SMALLINT": TokenType.SMALLINT, 859 "HUGEINT": TokenType.INT128, 860 "UHUGEINT": TokenType.UINT128, 861 "INT2": TokenType.SMALLINT, 862 "INTEGER": TokenType.INT, 863 "INT": TokenType.INT, 864 "INT4": TokenType.INT, 865 "INT32": TokenType.INT, 866 "INT64": TokenType.BIGINT, 867 "INT128": TokenType.INT128, 868 "INT256": TokenType.INT256, 869 "LONG": TokenType.BIGINT, 870 "BIGINT": TokenType.BIGINT, 871 "INT8": TokenType.TINYINT, 872 "UINT": TokenType.UINT, 873 "UINT128": TokenType.UINT128, 874 "UINT256": TokenType.UINT256, 875 "DEC": TokenType.DECIMAL, 876 "DECIMAL": TokenType.DECIMAL, 877 "DECIMAL32": TokenType.DECIMAL32, 878 "DECIMAL64": TokenType.DECIMAL64, 879 "DECIMAL128": TokenType.DECIMAL128, 880 "DECIMAL256": TokenType.DECIMAL256, 881 "BIGDECIMAL": TokenType.BIGDECIMAL, 882 "BIGNUMERIC": TokenType.BIGDECIMAL, 883 "LIST": TokenType.LIST, 884 "MAP": TokenType.MAP, 885 "NULLABLE": TokenType.NULLABLE, 886 "NUMBER": TokenType.DECIMAL, 887 "NUMERIC": TokenType.DECIMAL, 888 "FIXED": TokenType.DECIMAL, 889 "REAL": TokenType.FLOAT, 890 "FLOAT": TokenType.FLOAT, 891 "FLOAT4": TokenType.FLOAT, 892 "FLOAT8": TokenType.DOUBLE, 893 "DOUBLE": TokenType.DOUBLE, 894 "DOUBLE PRECISION": TokenType.DOUBLE, 895 "JSON": TokenType.JSON, 896 "JSONB": TokenType.JSONB, 897 "CHAR": TokenType.CHAR, 898 "CHARACTER": TokenType.CHAR, 899 "NCHAR": TokenType.NCHAR, 900 "VARCHAR": TokenType.VARCHAR, 901 "VARCHAR2": TokenType.VARCHAR, 902 "NVARCHAR": TokenType.NVARCHAR, 903 "NVARCHAR2": TokenType.NVARCHAR, 904 "BPCHAR": TokenType.BPCHAR, 905 "STR": TokenType.TEXT, 906 "STRING": TokenType.TEXT, 907 "TEXT": TokenType.TEXT, 908 "LONGTEXT": TokenType.LONGTEXT, 909 "MEDIUMTEXT": TokenType.MEDIUMTEXT, 910 "TINYTEXT": TokenType.TINYTEXT, 911 "CLOB": TokenType.TEXT, 912 "LONGVARCHAR": TokenType.TEXT, 913 "BINARY": TokenType.BINARY, 914 "BLOB": TokenType.VARBINARY, 915 "LONGBLOB": TokenType.LONGBLOB, 916 "MEDIUMBLOB": TokenType.MEDIUMBLOB, 917 "TINYBLOB": TokenType.TINYBLOB, 918 "BYTEA": TokenType.VARBINARY, 919 "VARBINARY": TokenType.VARBINARY, 920 "TIME": TokenType.TIME, 921 "TIMETZ": TokenType.TIMETZ, 922 "TIMESTAMP": TokenType.TIMESTAMP, 923 "TIMESTAMPTZ": TokenType.TIMESTAMPTZ, 924 "TIMESTAMPLTZ": TokenType.TIMESTAMPLTZ, 925 "TIMESTAMP_LTZ": TokenType.TIMESTAMPLTZ, 926 "TIMESTAMPNTZ": TokenType.TIMESTAMPNTZ, 927 "TIMESTAMP_NTZ": TokenType.TIMESTAMPNTZ, 928 "DATE": TokenType.DATE, 929 "DATETIME": TokenType.DATETIME, 930 "INT4RANGE": TokenType.INT4RANGE, 931 "INT4MULTIRANGE": TokenType.INT4MULTIRANGE, 932 "INT8RANGE": TokenType.INT8RANGE, 933 "INT8MULTIRANGE": TokenType.INT8MULTIRANGE, 934 "NUMRANGE": TokenType.NUMRANGE, 935 "NUMMULTIRANGE": TokenType.NUMMULTIRANGE, 936 "TSRANGE": TokenType.TSRANGE, 937 "TSMULTIRANGE": TokenType.TSMULTIRANGE, 938 "TSTZRANGE": TokenType.TSTZRANGE, 939 "TSTZMULTIRANGE": TokenType.TSTZMULTIRANGE, 940 "DATERANGE": TokenType.DATERANGE, 941 "DATEMULTIRANGE": TokenType.DATEMULTIRANGE, 942 "UNIQUE": TokenType.UNIQUE, 943 "VECTOR": TokenType.VECTOR, 944 "STRUCT": TokenType.STRUCT, 945 "SEQUENCE": TokenType.SEQUENCE, 946 "VARIANT": TokenType.VARIANT, 947 "ALTER": TokenType.ALTER, 948 "ANALYZE": TokenType.ANALYZE, 949 "CALL": TokenType.COMMAND, 950 "COMMENT": TokenType.COMMENT, 951 "EXPLAIN": TokenType.COMMAND, 952 "GRANT": TokenType.GRANT, 953 "OPTIMIZE": TokenType.COMMAND, 954 "PREPARE": TokenType.COMMAND, 955 "VACUUM": TokenType.COMMAND, 956 "USER-DEFINED": TokenType.USERDEFINED, 957 "FOR VERSION": TokenType.VERSION_SNAPSHOT, 958 "FOR TIMESTAMP": TokenType.TIMESTAMP_SNAPSHOT, 959 } 960 961 WHITE_SPACE: t.Dict[t.Optional[str], TokenType] = { 962 " ": TokenType.SPACE, 963 "\t": TokenType.SPACE, 964 "\n": TokenType.BREAK, 965 "\r": TokenType.BREAK, 966 } 967 968 COMMANDS = { 969 TokenType.COMMAND, 970 TokenType.EXECUTE, 971 TokenType.FETCH, 972 TokenType.SHOW, 973 TokenType.RENAME, 974 } 975 976 COMMAND_PREFIX_TOKENS = {TokenType.SEMICOLON, TokenType.BEGIN} 977 978 # Handle numeric literals like in hive (3L = BIGINT) 979 NUMERIC_LITERALS: t.Dict[str, str] = {} 980 981 COMMENTS = ["--", ("/*", "*/")] 982 983 __slots__ = ( 984 "sql", 985 "size", 986 "tokens", 987 "dialect", 988 "use_rs_tokenizer", 989 "_start", 990 "_current", 991 "_line", 992 "_col", 993 "_comments", 994 "_char", 995 "_end", 996 "_peek", 997 "_prev_token_line", 998 "_rs_dialect_settings", 999 ) 1000 1001 def __init__( 1002 self, dialect: DialectType = None, use_rs_tokenizer: t.Optional[bool] = None 1003 ) -> None: 1004 from sqlglot.dialects import Dialect 1005 1006 self.dialect = Dialect.get_or_raise(dialect) 1007 1008 # initialize `use_rs_tokenizer`, and allow it to be overwritten per Tokenizer instance 1009 self.use_rs_tokenizer = ( 1010 use_rs_tokenizer if use_rs_tokenizer is not None else USE_RS_TOKENIZER 1011 ) 1012 1013 if self.use_rs_tokenizer: 1014 self._rs_dialect_settings = RsTokenizerDialectSettings( 1015 unescaped_sequences=self.dialect.UNESCAPED_SEQUENCES, 1016 identifiers_can_start_with_digit=self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT, 1017 numbers_can_be_underscore_separated=self.dialect.NUMBERS_CAN_BE_UNDERSCORE_SEPARATED, 1018 ) 1019 1020 self.reset() 1021 1022 def reset(self) -> None: 1023 self.sql = "" 1024 self.size = 0 1025 self.tokens: t.List[Token] = [] 1026 self._start = 0 1027 self._current = 0 1028 self._line = 1 1029 self._col = 0 1030 self._comments: t.List[str] = [] 1031 1032 self._char = "" 1033 self._end = False 1034 self._peek = "" 1035 self._prev_token_line = -1 1036 1037 def tokenize(self, sql: str) -> t.List[Token]: 1038 """Returns a list of tokens corresponding to the SQL string `sql`.""" 1039 if self.use_rs_tokenizer: 1040 return self.tokenize_rs(sql) 1041 1042 self.reset() 1043 self.sql = sql 1044 self.size = len(sql) 1045 1046 try: 1047 self._scan() 1048 except Exception as e: 1049 start = max(self._current - 50, 0) 1050 end = min(self._current + 50, self.size - 1) 1051 context = self.sql[start:end] 1052 raise TokenError(f"Error tokenizing '{context}'") from e 1053 1054 return self.tokens 1055 1056 def _scan(self, until: t.Optional[t.Callable] = None) -> None: 1057 while self.size and not self._end: 1058 current = self._current 1059 1060 # Skip spaces here rather than iteratively calling advance() for performance reasons 1061 while current < self.size: 1062 char = self.sql[current] 1063 1064 if char.isspace() and (char == " " or char == "\t"): 1065 current += 1 1066 else: 1067 break 1068 1069 offset = current - self._current if current > self._current else 1 1070 1071 self._start = current 1072 self._advance(offset) 1073 1074 if not self._char.isspace(): 1075 if self._char.isdigit(): 1076 self._scan_number() 1077 elif self._char in self._IDENTIFIERS: 1078 self._scan_identifier(self._IDENTIFIERS[self._char]) 1079 else: 1080 self._scan_keywords() 1081 1082 if until and until(): 1083 break 1084 1085 if self.tokens and self._comments: 1086 self.tokens[-1].comments.extend(self._comments) 1087 1088 def _chars(self, size: int) -> str: 1089 if size == 1: 1090 return self._char 1091 1092 start = self._current - 1 1093 end = start + size 1094 1095 return self.sql[start:end] if end <= self.size else "" 1096 1097 def _advance(self, i: int = 1, alnum: bool = False) -> None: 1098 if self.WHITE_SPACE.get(self._char) is TokenType.BREAK: 1099 # Ensures we don't count an extra line if we get a \r\n line break sequence 1100 if not (self._char == "\r" and self._peek == "\n"): 1101 self._col = 1 1102 self._line += 1 1103 else: 1104 self._col += i 1105 1106 self._current += i 1107 self._end = self._current >= self.size 1108 self._char = self.sql[self._current - 1] 1109 self._peek = "" if self._end else self.sql[self._current] 1110 1111 if alnum and self._char.isalnum(): 1112 # Here we use local variables instead of attributes for better performance 1113 _col = self._col 1114 _current = self._current 1115 _end = self._end 1116 _peek = self._peek 1117 1118 while _peek.isalnum(): 1119 _col += 1 1120 _current += 1 1121 _end = _current >= self.size 1122 _peek = "" if _end else self.sql[_current] 1123 1124 self._col = _col 1125 self._current = _current 1126 self._end = _end 1127 self._peek = _peek 1128 self._char = self.sql[_current - 1] 1129 1130 @property 1131 def _text(self) -> str: 1132 return self.sql[self._start : self._current] 1133 1134 def _add(self, token_type: TokenType, text: t.Optional[str] = None) -> None: 1135 self._prev_token_line = self._line 1136 1137 if self._comments and token_type == TokenType.SEMICOLON and self.tokens: 1138 self.tokens[-1].comments.extend(self._comments) 1139 self._comments = [] 1140 1141 self.tokens.append( 1142 Token( 1143 token_type, 1144 text=self._text if text is None else text, 1145 line=self._line, 1146 col=self._col, 1147 start=self._start, 1148 end=self._current - 1, 1149 comments=self._comments, 1150 ) 1151 ) 1152 self._comments = [] 1153 1154 # If we have either a semicolon or a begin token before the command's token, we'll parse 1155 # whatever follows the command's token as a string 1156 if ( 1157 token_type in self.COMMANDS 1158 and self._peek != ";" 1159 and (len(self.tokens) == 1 or self.tokens[-2].token_type in self.COMMAND_PREFIX_TOKENS) 1160 ): 1161 start = self._current 1162 tokens = len(self.tokens) 1163 self._scan(lambda: self._peek == ";") 1164 self.tokens = self.tokens[:tokens] 1165 text = self.sql[start : self._current].strip() 1166 if text: 1167 self._add(TokenType.STRING, text) 1168 1169 def _scan_keywords(self) -> None: 1170 size = 0 1171 word = None 1172 chars = self._text 1173 char = chars 1174 prev_space = False 1175 skip = False 1176 trie = self._KEYWORD_TRIE 1177 single_token = char in self.SINGLE_TOKENS 1178 1179 while chars: 1180 if skip: 1181 result = TrieResult.PREFIX 1182 else: 1183 result, trie = in_trie(trie, char.upper()) 1184 1185 if result == TrieResult.FAILED: 1186 break 1187 if result == TrieResult.EXISTS: 1188 word = chars 1189 1190 end = self._current + size 1191 size += 1 1192 1193 if end < self.size: 1194 char = self.sql[end] 1195 single_token = single_token or char in self.SINGLE_TOKENS 1196 is_space = char.isspace() 1197 1198 if not is_space or not prev_space: 1199 if is_space: 1200 char = " " 1201 chars += char 1202 prev_space = is_space 1203 skip = False 1204 else: 1205 skip = True 1206 else: 1207 char = "" 1208 break 1209 1210 if word: 1211 if self._scan_string(word): 1212 return 1213 if self._scan_comment(word): 1214 return 1215 if prev_space or single_token or not char: 1216 self._advance(size - 1) 1217 word = word.upper() 1218 self._add(self.KEYWORDS[word], text=word) 1219 return 1220 1221 if self._char in self.SINGLE_TOKENS: 1222 self._add(self.SINGLE_TOKENS[self._char], text=self._char) 1223 return 1224 1225 self._scan_var() 1226 1227 def _scan_comment(self, comment_start: str) -> bool: 1228 if comment_start not in self._COMMENTS: 1229 return False 1230 1231 comment_start_line = self._line 1232 comment_start_size = len(comment_start) 1233 comment_end = self._COMMENTS[comment_start] 1234 1235 if comment_end: 1236 # Skip the comment's start delimiter 1237 self._advance(comment_start_size) 1238 1239 comment_count = 1 1240 comment_end_size = len(comment_end) 1241 1242 while not self._end: 1243 if self._chars(comment_end_size) == comment_end: 1244 comment_count -= 1 1245 if not comment_count: 1246 break 1247 1248 self._advance(alnum=True) 1249 1250 # Nested comments are allowed by some dialects, e.g. databricks, duckdb, postgres 1251 if ( 1252 self.NESTED_COMMENTS 1253 and not self._end 1254 and self._chars(comment_end_size) == comment_start 1255 ): 1256 self._advance(comment_start_size) 1257 comment_count += 1 1258 1259 self._comments.append(self._text[comment_start_size : -comment_end_size + 1]) 1260 self._advance(comment_end_size - 1) 1261 else: 1262 while not self._end and self.WHITE_SPACE.get(self._peek) is not TokenType.BREAK: 1263 self._advance(alnum=True) 1264 self._comments.append(self._text[comment_start_size:]) 1265 1266 if ( 1267 comment_start == self.HINT_START 1268 and self.tokens 1269 and self.tokens[-1].token_type in self.TOKENS_PRECEDING_HINT 1270 ): 1271 self._add(TokenType.HINT) 1272 1273 # Leading comment is attached to the succeeding token, whilst trailing comment to the preceding. 1274 # Multiple consecutive comments are preserved by appending them to the current comments list. 1275 if comment_start_line == self._prev_token_line: 1276 self.tokens[-1].comments.extend(self._comments) 1277 self._comments = [] 1278 self._prev_token_line = self._line 1279 1280 return True 1281 1282 def _scan_number(self) -> None: 1283 if self._char == "0": 1284 peek = self._peek.upper() 1285 if peek == "B": 1286 return self._scan_bits() if self.BIT_STRINGS else self._add(TokenType.NUMBER) 1287 elif peek == "X": 1288 return self._scan_hex() if self.HEX_STRINGS else self._add(TokenType.NUMBER) 1289 1290 decimal = False 1291 scientific = 0 1292 1293 while True: 1294 if self._peek.isdigit(): 1295 self._advance() 1296 elif self._peek == "." and not decimal: 1297 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER: 1298 return self._add(TokenType.NUMBER) 1299 decimal = True 1300 self._advance() 1301 elif self._peek in ("-", "+") and scientific == 1: 1302 scientific += 1 1303 self._advance() 1304 elif self._peek.upper() == "E" and not scientific: 1305 scientific += 1 1306 self._advance() 1307 elif self._peek.isidentifier(): 1308 number_text = self._text 1309 literal = "" 1310 1311 while self._peek.strip() and self._peek not in self.SINGLE_TOKENS: 1312 literal += self._peek 1313 self._advance() 1314 1315 token_type = self.KEYWORDS.get(self.NUMERIC_LITERALS.get(literal.upper(), "")) 1316 1317 if token_type: 1318 self._add(TokenType.NUMBER, number_text) 1319 self._add(TokenType.DCOLON, "::") 1320 return self._add(token_type, literal) 1321 else: 1322 replaced = literal.replace("_", "") 1323 if self.dialect.NUMBERS_CAN_BE_UNDERSCORE_SEPARATED and replaced.isdigit(): 1324 return self._add(TokenType.NUMBER, number_text + replaced) 1325 if self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT: 1326 return self._add(TokenType.VAR) 1327 1328 self._advance(-len(literal)) 1329 return self._add(TokenType.NUMBER, number_text) 1330 else: 1331 return self._add(TokenType.NUMBER) 1332 1333 def _scan_bits(self) -> None: 1334 self._advance() 1335 value = self._extract_value() 1336 try: 1337 # If `value` can't be converted to a binary, fallback to tokenizing it as an identifier 1338 int(value, 2) 1339 self._add(TokenType.BIT_STRING, value[2:]) # Drop the 0b 1340 except ValueError: 1341 self._add(TokenType.IDENTIFIER) 1342 1343 def _scan_hex(self) -> None: 1344 self._advance() 1345 value = self._extract_value() 1346 try: 1347 # If `value` can't be converted to a hex, fallback to tokenizing it as an identifier 1348 int(value, 16) 1349 self._add(TokenType.HEX_STRING, value[2:]) # Drop the 0x 1350 except ValueError: 1351 self._add(TokenType.IDENTIFIER) 1352 1353 def _extract_value(self) -> str: 1354 while True: 1355 char = self._peek.strip() 1356 if char and char not in self.SINGLE_TOKENS: 1357 self._advance(alnum=True) 1358 else: 1359 break 1360 1361 return self._text 1362 1363 def _scan_string(self, start: str) -> bool: 1364 base = None 1365 token_type = TokenType.STRING 1366 1367 if start in self._QUOTES: 1368 end = self._QUOTES[start] 1369 elif start in self._FORMAT_STRINGS: 1370 end, token_type = self._FORMAT_STRINGS[start] 1371 1372 if token_type == TokenType.HEX_STRING: 1373 base = 16 1374 elif token_type == TokenType.BIT_STRING: 1375 base = 2 1376 elif token_type == TokenType.HEREDOC_STRING: 1377 self._advance() 1378 1379 if self._char == end: 1380 tag = "" 1381 else: 1382 tag = self._extract_string( 1383 end, 1384 raw_string=True, 1385 raise_unmatched=not self.HEREDOC_TAG_IS_IDENTIFIER, 1386 ) 1387 1388 if tag and self.HEREDOC_TAG_IS_IDENTIFIER and (self._end or not tag.isidentifier()): 1389 if not self._end: 1390 self._advance(-1) 1391 1392 self._advance(-len(tag)) 1393 self._add(self.HEREDOC_STRING_ALTERNATIVE) 1394 return True 1395 1396 end = f"{start}{tag}{end}" 1397 else: 1398 return False 1399 1400 self._advance(len(start)) 1401 text = self._extract_string(end, raw_string=token_type == TokenType.RAW_STRING) 1402 1403 if base: 1404 try: 1405 int(text, base) 1406 except Exception: 1407 raise TokenError( 1408 f"Numeric string contains invalid characters from {self._line}:{self._start}" 1409 ) 1410 1411 self._add(token_type, text) 1412 return True 1413 1414 def _scan_identifier(self, identifier_end: str) -> None: 1415 self._advance() 1416 text = self._extract_string( 1417 identifier_end, escapes=self._IDENTIFIER_ESCAPES | {identifier_end} 1418 ) 1419 self._add(TokenType.IDENTIFIER, text) 1420 1421 def _scan_var(self) -> None: 1422 while True: 1423 char = self._peek.strip() 1424 if char and (char in self.VAR_SINGLE_TOKENS or char not in self.SINGLE_TOKENS): 1425 self._advance(alnum=True) 1426 else: 1427 break 1428 1429 self._add( 1430 TokenType.VAR 1431 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER 1432 else self.KEYWORDS.get(self._text.upper(), TokenType.VAR) 1433 ) 1434 1435 def _extract_string( 1436 self, 1437 delimiter: str, 1438 escapes: t.Optional[t.Set[str]] = None, 1439 raw_string: bool = False, 1440 raise_unmatched: bool = True, 1441 ) -> str: 1442 text = "" 1443 delim_size = len(delimiter) 1444 escapes = self._STRING_ESCAPES if escapes is None else escapes 1445 1446 while True: 1447 if ( 1448 not raw_string 1449 and self.dialect.UNESCAPED_SEQUENCES 1450 and self._peek 1451 and self._char in self.STRING_ESCAPES 1452 ): 1453 unescaped_sequence = self.dialect.UNESCAPED_SEQUENCES.get(self._char + self._peek) 1454 if unescaped_sequence: 1455 self._advance(2) 1456 text += unescaped_sequence 1457 continue 1458 if ( 1459 (self.STRING_ESCAPES_ALLOWED_IN_RAW_STRINGS or not raw_string) 1460 and self._char in escapes 1461 and (self._peek == delimiter or self._peek in escapes) 1462 and (self._char not in self._QUOTES or self._char == self._peek) 1463 ): 1464 if self._peek == delimiter: 1465 text += self._peek 1466 else: 1467 text += self._char + self._peek 1468 1469 if self._current + 1 < self.size: 1470 self._advance(2) 1471 else: 1472 raise TokenError(f"Missing {delimiter} from {self._line}:{self._current}") 1473 else: 1474 if self._chars(delim_size) == delimiter: 1475 if delim_size > 1: 1476 self._advance(delim_size - 1) 1477 break 1478 1479 if self._end: 1480 if not raise_unmatched: 1481 return text + self._char 1482 1483 raise TokenError(f"Missing {delimiter} from {self._line}:{self._start}") 1484 1485 current = self._current - 1 1486 self._advance(alnum=True) 1487 text += self.sql[current : self._current - 1] 1488 1489 return text 1490 1491 def tokenize_rs(self, sql: str) -> t.List[Token]: 1492 if not self._RS_TOKENIZER: 1493 raise SqlglotError("Rust tokenizer is not available") 1494 1495 try: 1496 tokens = self._RS_TOKENIZER.tokenize(sql, self._rs_dialect_settings) 1497 for token in tokens: 1498 token.token_type = _ALL_TOKEN_TYPES[token.token_type_index] 1499 return tokens 1500 except Exception as e: 1501 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 CARET_AT = auto() 65 TILDA = auto() 66 ARROW = auto() 67 DARROW = auto() 68 FARROW = auto() 69 HASH = auto() 70 HASH_ARROW = auto() 71 DHASH_ARROW = auto() 72 LR_ARROW = auto() 73 DAT = auto() 74 LT_AT = auto() 75 AT_GT = auto() 76 DOLLAR = auto() 77 PARAMETER = auto() 78 SESSION_PARAMETER = auto() 79 DAMP = auto() 80 XOR = auto() 81 DSTAR = auto() 82 83 BLOCK_START = auto() 84 BLOCK_END = auto() 85 86 SPACE = auto() 87 BREAK = auto() 88 89 STRING = auto() 90 NUMBER = auto() 91 IDENTIFIER = auto() 92 DATABASE = auto() 93 COLUMN = auto() 94 COLUMN_DEF = auto() 95 SCHEMA = auto() 96 TABLE = auto() 97 WAREHOUSE = auto() 98 STREAMLIT = auto() 99 VAR = auto() 100 BIT_STRING = auto() 101 HEX_STRING = auto() 102 BYTE_STRING = auto() 103 NATIONAL_STRING = auto() 104 RAW_STRING = auto() 105 HEREDOC_STRING = auto() 106 UNICODE_STRING = auto() 107 108 # types 109 BIT = auto() 110 BOOLEAN = auto() 111 TINYINT = auto() 112 UTINYINT = auto() 113 SMALLINT = auto() 114 USMALLINT = auto() 115 MEDIUMINT = auto() 116 UMEDIUMINT = auto() 117 INT = auto() 118 UINT = auto() 119 BIGINT = auto() 120 UBIGINT = auto() 121 INT128 = auto() 122 UINT128 = auto() 123 INT256 = auto() 124 UINT256 = auto() 125 FLOAT = auto() 126 DOUBLE = auto() 127 DECIMAL = auto() 128 DECIMAL32 = auto() 129 DECIMAL64 = auto() 130 DECIMAL128 = auto() 131 DECIMAL256 = auto() 132 UDECIMAL = auto() 133 BIGDECIMAL = auto() 134 CHAR = auto() 135 NCHAR = auto() 136 VARCHAR = auto() 137 NVARCHAR = auto() 138 BPCHAR = auto() 139 TEXT = auto() 140 MEDIUMTEXT = auto() 141 LONGTEXT = auto() 142 MEDIUMBLOB = auto() 143 LONGBLOB = auto() 144 TINYBLOB = auto() 145 TINYTEXT = auto() 146 NAME = auto() 147 BINARY = auto() 148 VARBINARY = auto() 149 JSON = auto() 150 JSONB = auto() 151 TIME = auto() 152 TIMETZ = auto() 153 TIMESTAMP = auto() 154 TIMESTAMPTZ = auto() 155 TIMESTAMPLTZ = auto() 156 TIMESTAMPNTZ = auto() 157 TIMESTAMP_S = auto() 158 TIMESTAMP_MS = auto() 159 TIMESTAMP_NS = auto() 160 DATETIME = auto() 161 DATETIME2 = auto() 162 DATETIME64 = auto() 163 SMALLDATETIME = auto() 164 DATE = auto() 165 DATE32 = auto() 166 INT4RANGE = auto() 167 INT4MULTIRANGE = auto() 168 INT8RANGE = auto() 169 INT8MULTIRANGE = auto() 170 NUMRANGE = auto() 171 NUMMULTIRANGE = auto() 172 TSRANGE = auto() 173 TSMULTIRANGE = auto() 174 TSTZRANGE = auto() 175 TSTZMULTIRANGE = auto() 176 DATERANGE = auto() 177 DATEMULTIRANGE = auto() 178 UUID = auto() 179 GEOGRAPHY = auto() 180 NULLABLE = auto() 181 GEOMETRY = auto() 182 POINT = auto() 183 RING = auto() 184 LINESTRING = auto() 185 MULTILINESTRING = auto() 186 POLYGON = auto() 187 MULTIPOLYGON = auto() 188 HLLSKETCH = auto() 189 HSTORE = auto() 190 SUPER = auto() 191 SERIAL = auto() 192 SMALLSERIAL = auto() 193 BIGSERIAL = auto() 194 XML = auto() 195 YEAR = auto() 196 USERDEFINED = auto() 197 MONEY = auto() 198 SMALLMONEY = auto() 199 ROWVERSION = auto() 200 IMAGE = auto() 201 VARIANT = auto() 202 OBJECT = auto() 203 INET = auto() 204 IPADDRESS = auto() 205 IPPREFIX = auto() 206 IPV4 = auto() 207 IPV6 = auto() 208 ENUM = auto() 209 ENUM8 = auto() 210 ENUM16 = auto() 211 FIXEDSTRING = auto() 212 LOWCARDINALITY = auto() 213 NESTED = auto() 214 AGGREGATEFUNCTION = auto() 215 SIMPLEAGGREGATEFUNCTION = auto() 216 TDIGEST = auto() 217 UNKNOWN = auto() 218 VECTOR = auto() 219 DYNAMIC = auto() 220 221 # keywords 222 ALIAS = auto() 223 ALTER = auto() 224 ALWAYS = auto() 225 ALL = auto() 226 ANTI = auto() 227 ANY = auto() 228 APPLY = auto() 229 ARRAY = auto() 230 ASC = auto() 231 ASOF = auto() 232 ATTACH = auto() 233 AUTO_INCREMENT = auto() 234 BEGIN = auto() 235 BETWEEN = auto() 236 BULK_COLLECT_INTO = auto() 237 CACHE = auto() 238 CASE = auto() 239 CHARACTER_SET = auto() 240 CLUSTER_BY = auto() 241 COLLATE = auto() 242 COMMAND = auto() 243 COMMENT = auto() 244 COMMIT = auto() 245 CONNECT_BY = auto() 246 CONSTRAINT = auto() 247 COPY = auto() 248 CREATE = auto() 249 CROSS = auto() 250 CUBE = auto() 251 CURRENT_DATE = auto() 252 CURRENT_DATETIME = auto() 253 CURRENT_SCHEMA = auto() 254 CURRENT_TIME = auto() 255 CURRENT_TIMESTAMP = auto() 256 CURRENT_USER = auto() 257 DECLARE = auto() 258 DEFAULT = auto() 259 DELETE = auto() 260 DESC = auto() 261 DESCRIBE = auto() 262 DETACH = auto() 263 DICTIONARY = auto() 264 DISTINCT = auto() 265 DISTRIBUTE_BY = auto() 266 DIV = auto() 267 DROP = auto() 268 ELSE = auto() 269 END = auto() 270 ESCAPE = auto() 271 EXCEPT = auto() 272 EXECUTE = auto() 273 EXISTS = auto() 274 FALSE = auto() 275 FETCH = auto() 276 FILTER = auto() 277 FINAL = auto() 278 FIRST = auto() 279 FOR = auto() 280 FORCE = auto() 281 FOREIGN_KEY = auto() 282 FORMAT = auto() 283 FROM = auto() 284 FULL = auto() 285 FUNCTION = auto() 286 GLOB = auto() 287 GLOBAL = auto() 288 GRANT = auto() 289 GROUP_BY = auto() 290 GROUPING_SETS = auto() 291 HAVING = auto() 292 HINT = auto() 293 IGNORE = auto() 294 ILIKE = auto() 295 ILIKE_ANY = auto() 296 IN = auto() 297 INDEX = auto() 298 INNER = auto() 299 INSERT = auto() 300 INTERSECT = auto() 301 INTERVAL = auto() 302 INTO = auto() 303 INTRODUCER = auto() 304 IRLIKE = auto() 305 IS = auto() 306 ISNULL = auto() 307 JOIN = auto() 308 JOIN_MARKER = auto() 309 KEEP = auto() 310 KEY = auto() 311 KILL = auto() 312 LANGUAGE = auto() 313 LATERAL = auto() 314 LEFT = auto() 315 LIKE = auto() 316 LIKE_ANY = auto() 317 LIMIT = auto() 318 LIST = auto() 319 LOAD = auto() 320 LOCK = auto() 321 MAP = auto() 322 MATCH_CONDITION = auto() 323 MATCH_RECOGNIZE = auto() 324 MEMBER_OF = auto() 325 MERGE = auto() 326 MOD = auto() 327 MODEL = auto() 328 NATURAL = auto() 329 NEXT = auto() 330 NOTNULL = auto() 331 NULL = auto() 332 OBJECT_IDENTIFIER = auto() 333 OFFSET = auto() 334 ON = auto() 335 ONLY = auto() 336 OPERATOR = auto() 337 ORDER_BY = auto() 338 ORDER_SIBLINGS_BY = auto() 339 ORDERED = auto() 340 ORDINALITY = auto() 341 OUTER = auto() 342 OVER = auto() 343 OVERLAPS = auto() 344 OVERWRITE = auto() 345 PARTITION = auto() 346 PARTITION_BY = auto() 347 PERCENT = auto() 348 PIVOT = auto() 349 PLACEHOLDER = auto() 350 POSITIONAL = auto() 351 PRAGMA = auto() 352 PREWHERE = auto() 353 PRIMARY_KEY = auto() 354 PROCEDURE = auto() 355 PROPERTIES = auto() 356 PSEUDO_TYPE = auto() 357 QUALIFY = auto() 358 QUOTE = auto() 359 RANGE = auto() 360 RECURSIVE = auto() 361 REFRESH = auto() 362 RENAME = auto() 363 REPLACE = auto() 364 RETURNING = auto() 365 REFERENCES = auto() 366 RIGHT = auto() 367 RLIKE = auto() 368 ROLLBACK = auto() 369 ROLLUP = auto() 370 ROW = auto() 371 ROWS = auto() 372 SELECT = auto() 373 SEMI = auto() 374 SEPARATOR = auto() 375 SEQUENCE = auto() 376 SERDE_PROPERTIES = auto() 377 SET = auto() 378 SETTINGS = auto() 379 SHOW = auto() 380 SIMILAR_TO = auto() 381 SOME = auto() 382 SORT_BY = auto() 383 START_WITH = auto() 384 STORAGE_INTEGRATION = auto() 385 STRAIGHT_JOIN = auto() 386 STRUCT = auto() 387 SUMMARIZE = auto() 388 TABLE_SAMPLE = auto() 389 TAG = auto() 390 TEMPORARY = auto() 391 TOP = auto() 392 THEN = auto() 393 TRUE = auto() 394 TRUNCATE = auto() 395 UNCACHE = auto() 396 UNION = auto() 397 UNNEST = auto() 398 UNPIVOT = auto() 399 UPDATE = auto() 400 USE = auto() 401 USING = auto() 402 VALUES = auto() 403 VIEW = auto() 404 VOLATILE = auto() 405 WHEN = auto() 406 WHERE = auto() 407 WINDOW = auto() 408 WITH = auto() 409 UNIQUE = auto() 410 VERSION_SNAPSHOT = auto() 411 TIMESTAMP_SNAPSHOT = auto() 412 OPTION = auto() 413 SINK = auto() 414 SOURCE = auto() 415 ANALYZE = auto() 416 NAMESPACE = auto() 417 EXPORT = 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'>
CARET_AT =
<TokenType.CARET_AT: 'CARET_AT'>
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'>
WAREHOUSE =
<TokenType.WAREHOUSE: 'WAREHOUSE'>
STREAMLIT =
<TokenType.STREAMLIT: 'STREAMLIT'>
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'>
DECIMAL32 =
<TokenType.DECIMAL32: 'DECIMAL32'>
DECIMAL64 =
<TokenType.DECIMAL64: 'DECIMAL64'>
DECIMAL128 =
<TokenType.DECIMAL128: 'DECIMAL128'>
DECIMAL256 =
<TokenType.DECIMAL256: 'DECIMAL256'>
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'>
DATETIME2 =
<TokenType.DATETIME2: 'DATETIME2'>
DATETIME64 =
<TokenType.DATETIME64: 'DATETIME64'>
SMALLDATETIME =
<TokenType.SMALLDATETIME: 'SMALLDATETIME'>
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'>
POINT =
<TokenType.POINT: 'POINT'>
RING =
<TokenType.RING: 'RING'>
LINESTRING =
<TokenType.LINESTRING: 'LINESTRING'>
MULTILINESTRING =
<TokenType.MULTILINESTRING: 'MULTILINESTRING'>
POLYGON =
<TokenType.POLYGON: 'POLYGON'>
MULTIPOLYGON =
<TokenType.MULTIPOLYGON: 'MULTIPOLYGON'>
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'>
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'>
VECTOR =
<TokenType.VECTOR: 'VECTOR'>
DYNAMIC =
<TokenType.DYNAMIC: 'DYNAMIC'>
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'>
ATTACH =
<TokenType.ATTACH: 'ATTACH'>
AUTO_INCREMENT =
<TokenType.AUTO_INCREMENT: 'AUTO_INCREMENT'>
BEGIN =
<TokenType.BEGIN: 'BEGIN'>
BETWEEN =
<TokenType.BETWEEN: 'BETWEEN'>
BULK_COLLECT_INTO =
<TokenType.BULK_COLLECT_INTO: 'BULK_COLLECT_INTO'>
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_SCHEMA =
<TokenType.CURRENT_SCHEMA: 'CURRENT_SCHEMA'>
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'>
DETACH =
<TokenType.DETACH: 'DETACH'>
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'>
GRANT =
<TokenType.GRANT: 'GRANT'>
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'>
KEY =
<TokenType.KEY: 'KEY'>
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'>
LIST =
<TokenType.LIST: 'LIST'>
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'>
RENAME =
<TokenType.RENAME: 'RENAME'>
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'>
STRAIGHT_JOIN =
<TokenType.STRAIGHT_JOIN: 'STRAIGHT_JOIN'>
STRUCT =
<TokenType.STRUCT: 'STRUCT'>
SUMMARIZE =
<TokenType.SUMMARIZE: 'SUMMARIZE'>
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'>
SINK =
<TokenType.SINK: 'SINK'>
SOURCE =
<TokenType.SOURCE: 'SOURCE'>
ANALYZE =
<TokenType.ANALYZE: 'ANALYZE'>
NAMESPACE =
<TokenType.NAMESPACE: 'NAMESPACE'>
EXPORT =
<TokenType.EXPORT: 'EXPORT'>
class
Token:
424class Token: 425 __slots__ = ("token_type", "text", "line", "col", "start", "end", "comments") 426 427 @classmethod 428 def number(cls, number: int) -> Token: 429 """Returns a NUMBER token with `number` as its text.""" 430 return cls(TokenType.NUMBER, str(number)) 431 432 @classmethod 433 def string(cls, string: str) -> Token: 434 """Returns a STRING token with `string` as its text.""" 435 return cls(TokenType.STRING, string) 436 437 @classmethod 438 def identifier(cls, identifier: str) -> Token: 439 """Returns an IDENTIFIER token with `identifier` as its text.""" 440 return cls(TokenType.IDENTIFIER, identifier) 441 442 @classmethod 443 def var(cls, var: str) -> Token: 444 """Returns an VAR token with `var` as its text.""" 445 return cls(TokenType.VAR, var) 446 447 def __init__( 448 self, 449 token_type: TokenType, 450 text: str, 451 line: int = 1, 452 col: int = 1, 453 start: int = 0, 454 end: int = 0, 455 comments: t.Optional[t.List[str]] = None, 456 ) -> None: 457 """Token initializer. 458 459 Args: 460 token_type: The TokenType Enum. 461 text: The text of the token. 462 line: The line that the token ends on. 463 col: The column that the token ends on. 464 start: The start index of the token. 465 end: The ending index of the token. 466 comments: The comments to attach to the token. 467 """ 468 self.token_type = token_type 469 self.text = text 470 self.line = line 471 self.col = col 472 self.start = start 473 self.end = end 474 self.comments = [] if comments is None else comments 475 476 def __repr__(self) -> str: 477 attributes = ", ".join(f"{k}: {getattr(self, k)}" for k in self.__slots__) 478 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)
447 def __init__( 448 self, 449 token_type: TokenType, 450 text: str, 451 line: int = 1, 452 col: int = 1, 453 start: int = 0, 454 end: int = 0, 455 comments: t.Optional[t.List[str]] = None, 456 ) -> None: 457 """Token initializer. 458 459 Args: 460 token_type: The TokenType Enum. 461 text: The text of the token. 462 line: The line that the token ends on. 463 col: The column that the token ends on. 464 start: The start index of the token. 465 end: The ending index of the token. 466 comments: The comments to attach to the token. 467 """ 468 self.token_type = token_type 469 self.text = text 470 self.line = line 471 self.col = col 472 self.start = start 473 self.end = end 474 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.
427 @classmethod 428 def number(cls, number: int) -> Token: 429 """Returns a NUMBER token with `number` as its text.""" 430 return cls(TokenType.NUMBER, str(number))
Returns a NUMBER token with number
as its text.
432 @classmethod 433 def string(cls, string: str) -> Token: 434 """Returns a STRING token with `string` as its text.""" 435 return cls(TokenType.STRING, string)
Returns a STRING token with string
as its text.
437 @classmethod 438 def identifier(cls, identifier: str) -> Token: 439 """Returns an IDENTIFIER token with `identifier` as its text.""" 440 return cls(TokenType.IDENTIFIER, identifier)
Returns an IDENTIFIER token with identifier
as its text.
class
Tokenizer:
588class Tokenizer(metaclass=_Tokenizer): 589 SINGLE_TOKENS = { 590 "(": TokenType.L_PAREN, 591 ")": TokenType.R_PAREN, 592 "[": TokenType.L_BRACKET, 593 "]": TokenType.R_BRACKET, 594 "{": TokenType.L_BRACE, 595 "}": TokenType.R_BRACE, 596 "&": TokenType.AMP, 597 "^": TokenType.CARET, 598 ":": TokenType.COLON, 599 ",": TokenType.COMMA, 600 ".": TokenType.DOT, 601 "-": TokenType.DASH, 602 "=": TokenType.EQ, 603 ">": TokenType.GT, 604 "<": TokenType.LT, 605 "%": TokenType.MOD, 606 "!": TokenType.NOT, 607 "|": TokenType.PIPE, 608 "+": TokenType.PLUS, 609 ";": TokenType.SEMICOLON, 610 "/": TokenType.SLASH, 611 "\\": TokenType.BACKSLASH, 612 "*": TokenType.STAR, 613 "~": TokenType.TILDA, 614 "?": TokenType.PLACEHOLDER, 615 "@": TokenType.PARAMETER, 616 "#": TokenType.HASH, 617 # Used for breaking a var like x'y' but nothing else the token type doesn't matter 618 "'": TokenType.UNKNOWN, 619 "`": TokenType.UNKNOWN, 620 '"': TokenType.UNKNOWN, 621 } 622 623 BIT_STRINGS: t.List[str | t.Tuple[str, str]] = [] 624 BYTE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 625 HEX_STRINGS: t.List[str | t.Tuple[str, str]] = [] 626 RAW_STRINGS: t.List[str | t.Tuple[str, str]] = [] 627 HEREDOC_STRINGS: t.List[str | t.Tuple[str, str]] = [] 628 UNICODE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 629 IDENTIFIERS: t.List[str | t.Tuple[str, str]] = ['"'] 630 QUOTES: t.List[t.Tuple[str, str] | str] = ["'"] 631 STRING_ESCAPES = ["'"] 632 VAR_SINGLE_TOKENS: t.Set[str] = set() 633 634 # The strings in this list can always be used as escapes, regardless of the surrounding 635 # identifier delimiters. By default, the closing delimiter is assumed to also act as an 636 # identifier escape, e.g. if we use double-quotes, then they also act as escapes: "x""" 637 IDENTIFIER_ESCAPES: t.List[str] = [] 638 639 # Whether the heredoc tags follow the same lexical rules as unquoted identifiers 640 HEREDOC_TAG_IS_IDENTIFIER = False 641 642 # Token that we'll generate as a fallback if the heredoc prefix doesn't correspond to a heredoc 643 HEREDOC_STRING_ALTERNATIVE = TokenType.VAR 644 645 # Whether string escape characters function as such when placed within raw strings 646 STRING_ESCAPES_ALLOWED_IN_RAW_STRINGS = True 647 648 NESTED_COMMENTS = True 649 650 HINT_START = "/*+" 651 652 TOKENS_PRECEDING_HINT = {TokenType.SELECT, TokenType.INSERT, TokenType.UPDATE, TokenType.DELETE} 653 654 # Autofilled 655 _COMMENTS: t.Dict[str, str] = {} 656 _FORMAT_STRINGS: t.Dict[str, t.Tuple[str, TokenType]] = {} 657 _IDENTIFIERS: t.Dict[str, str] = {} 658 _IDENTIFIER_ESCAPES: t.Set[str] = set() 659 _QUOTES: t.Dict[str, str] = {} 660 _STRING_ESCAPES: t.Set[str] = set() 661 _KEYWORD_TRIE: t.Dict = {} 662 _RS_TOKENIZER: t.Optional[t.Any] = None 663 664 KEYWORDS: t.Dict[str, TokenType] = { 665 **{f"{{%{postfix}": TokenType.BLOCK_START for postfix in ("", "+", "-")}, 666 **{f"{prefix}%}}": TokenType.BLOCK_END for prefix in ("", "+", "-")}, 667 **{f"{{{{{postfix}": TokenType.BLOCK_START for postfix in ("+", "-")}, 668 **{f"{prefix}}}}}": TokenType.BLOCK_END for prefix in ("+", "-")}, 669 HINT_START: TokenType.HINT, 670 "==": TokenType.EQ, 671 "::": TokenType.DCOLON, 672 "||": TokenType.DPIPE, 673 ">=": TokenType.GTE, 674 "<=": TokenType.LTE, 675 "<>": TokenType.NEQ, 676 "!=": TokenType.NEQ, 677 ":=": TokenType.COLON_EQ, 678 "<=>": TokenType.NULLSAFE_EQ, 679 "->": TokenType.ARROW, 680 "->>": TokenType.DARROW, 681 "=>": TokenType.FARROW, 682 "#>": TokenType.HASH_ARROW, 683 "#>>": TokenType.DHASH_ARROW, 684 "<->": TokenType.LR_ARROW, 685 "&&": TokenType.DAMP, 686 "??": TokenType.DQMARK, 687 "~~~": TokenType.GLOB, 688 "~~": TokenType.LIKE, 689 "~~*": TokenType.ILIKE, 690 "~*": TokenType.IRLIKE, 691 "ALL": TokenType.ALL, 692 "ALWAYS": TokenType.ALWAYS, 693 "AND": TokenType.AND, 694 "ANTI": TokenType.ANTI, 695 "ANY": TokenType.ANY, 696 "ASC": TokenType.ASC, 697 "AS": TokenType.ALIAS, 698 "ASOF": TokenType.ASOF, 699 "AUTOINCREMENT": TokenType.AUTO_INCREMENT, 700 "AUTO_INCREMENT": TokenType.AUTO_INCREMENT, 701 "BEGIN": TokenType.BEGIN, 702 "BETWEEN": TokenType.BETWEEN, 703 "CACHE": TokenType.CACHE, 704 "UNCACHE": TokenType.UNCACHE, 705 "CASE": TokenType.CASE, 706 "CHARACTER SET": TokenType.CHARACTER_SET, 707 "CLUSTER BY": TokenType.CLUSTER_BY, 708 "COLLATE": TokenType.COLLATE, 709 "COLUMN": TokenType.COLUMN, 710 "COMMIT": TokenType.COMMIT, 711 "CONNECT BY": TokenType.CONNECT_BY, 712 "CONSTRAINT": TokenType.CONSTRAINT, 713 "COPY": TokenType.COPY, 714 "CREATE": TokenType.CREATE, 715 "CROSS": TokenType.CROSS, 716 "CUBE": TokenType.CUBE, 717 "CURRENT_DATE": TokenType.CURRENT_DATE, 718 "CURRENT_SCHEMA": TokenType.CURRENT_SCHEMA, 719 "CURRENT_TIME": TokenType.CURRENT_TIME, 720 "CURRENT_TIMESTAMP": TokenType.CURRENT_TIMESTAMP, 721 "CURRENT_USER": TokenType.CURRENT_USER, 722 "DATABASE": TokenType.DATABASE, 723 "DEFAULT": TokenType.DEFAULT, 724 "DELETE": TokenType.DELETE, 725 "DESC": TokenType.DESC, 726 "DESCRIBE": TokenType.DESCRIBE, 727 "DISTINCT": TokenType.DISTINCT, 728 "DISTRIBUTE BY": TokenType.DISTRIBUTE_BY, 729 "DIV": TokenType.DIV, 730 "DROP": TokenType.DROP, 731 "ELSE": TokenType.ELSE, 732 "END": TokenType.END, 733 "ENUM": TokenType.ENUM, 734 "ESCAPE": TokenType.ESCAPE, 735 "EXCEPT": TokenType.EXCEPT, 736 "EXECUTE": TokenType.EXECUTE, 737 "EXISTS": TokenType.EXISTS, 738 "FALSE": TokenType.FALSE, 739 "FETCH": TokenType.FETCH, 740 "FILTER": TokenType.FILTER, 741 "FIRST": TokenType.FIRST, 742 "FULL": TokenType.FULL, 743 "FUNCTION": TokenType.FUNCTION, 744 "FOR": TokenType.FOR, 745 "FOREIGN KEY": TokenType.FOREIGN_KEY, 746 "FORMAT": TokenType.FORMAT, 747 "FROM": TokenType.FROM, 748 "GEOGRAPHY": TokenType.GEOGRAPHY, 749 "GEOMETRY": TokenType.GEOMETRY, 750 "GLOB": TokenType.GLOB, 751 "GROUP BY": TokenType.GROUP_BY, 752 "GROUPING SETS": TokenType.GROUPING_SETS, 753 "HAVING": TokenType.HAVING, 754 "ILIKE": TokenType.ILIKE, 755 "IN": TokenType.IN, 756 "INDEX": TokenType.INDEX, 757 "INET": TokenType.INET, 758 "INNER": TokenType.INNER, 759 "INSERT": TokenType.INSERT, 760 "INTERVAL": TokenType.INTERVAL, 761 "INTERSECT": TokenType.INTERSECT, 762 "INTO": TokenType.INTO, 763 "IS": TokenType.IS, 764 "ISNULL": TokenType.ISNULL, 765 "JOIN": TokenType.JOIN, 766 "KEEP": TokenType.KEEP, 767 "KILL": TokenType.KILL, 768 "LATERAL": TokenType.LATERAL, 769 "LEFT": TokenType.LEFT, 770 "LIKE": TokenType.LIKE, 771 "LIMIT": TokenType.LIMIT, 772 "LOAD": TokenType.LOAD, 773 "LOCK": TokenType.LOCK, 774 "MERGE": TokenType.MERGE, 775 "NAMESPACE": TokenType.NAMESPACE, 776 "NATURAL": TokenType.NATURAL, 777 "NEXT": TokenType.NEXT, 778 "NOT": TokenType.NOT, 779 "NOTNULL": TokenType.NOTNULL, 780 "NULL": TokenType.NULL, 781 "OBJECT": TokenType.OBJECT, 782 "OFFSET": TokenType.OFFSET, 783 "ON": TokenType.ON, 784 "OR": TokenType.OR, 785 "XOR": TokenType.XOR, 786 "ORDER BY": TokenType.ORDER_BY, 787 "ORDINALITY": TokenType.ORDINALITY, 788 "OUTER": TokenType.OUTER, 789 "OVER": TokenType.OVER, 790 "OVERLAPS": TokenType.OVERLAPS, 791 "OVERWRITE": TokenType.OVERWRITE, 792 "PARTITION": TokenType.PARTITION, 793 "PARTITION BY": TokenType.PARTITION_BY, 794 "PARTITIONED BY": TokenType.PARTITION_BY, 795 "PARTITIONED_BY": TokenType.PARTITION_BY, 796 "PERCENT": TokenType.PERCENT, 797 "PIVOT": TokenType.PIVOT, 798 "PRAGMA": TokenType.PRAGMA, 799 "PRIMARY KEY": TokenType.PRIMARY_KEY, 800 "PROCEDURE": TokenType.PROCEDURE, 801 "QUALIFY": TokenType.QUALIFY, 802 "RANGE": TokenType.RANGE, 803 "RECURSIVE": TokenType.RECURSIVE, 804 "REGEXP": TokenType.RLIKE, 805 "RENAME": TokenType.RENAME, 806 "REPLACE": TokenType.REPLACE, 807 "RETURNING": TokenType.RETURNING, 808 "REFERENCES": TokenType.REFERENCES, 809 "RIGHT": TokenType.RIGHT, 810 "RLIKE": TokenType.RLIKE, 811 "ROLLBACK": TokenType.ROLLBACK, 812 "ROLLUP": TokenType.ROLLUP, 813 "ROW": TokenType.ROW, 814 "ROWS": TokenType.ROWS, 815 "SCHEMA": TokenType.SCHEMA, 816 "SELECT": TokenType.SELECT, 817 "SEMI": TokenType.SEMI, 818 "SET": TokenType.SET, 819 "SETTINGS": TokenType.SETTINGS, 820 "SHOW": TokenType.SHOW, 821 "SIMILAR TO": TokenType.SIMILAR_TO, 822 "SOME": TokenType.SOME, 823 "SORT BY": TokenType.SORT_BY, 824 "START WITH": TokenType.START_WITH, 825 "STRAIGHT_JOIN": TokenType.STRAIGHT_JOIN, 826 "TABLE": TokenType.TABLE, 827 "TABLESAMPLE": TokenType.TABLE_SAMPLE, 828 "TEMP": TokenType.TEMPORARY, 829 "TEMPORARY": TokenType.TEMPORARY, 830 "THEN": TokenType.THEN, 831 "TRUE": TokenType.TRUE, 832 "TRUNCATE": TokenType.TRUNCATE, 833 "UNION": TokenType.UNION, 834 "UNKNOWN": TokenType.UNKNOWN, 835 "UNNEST": TokenType.UNNEST, 836 "UNPIVOT": TokenType.UNPIVOT, 837 "UPDATE": TokenType.UPDATE, 838 "USE": TokenType.USE, 839 "USING": TokenType.USING, 840 "UUID": TokenType.UUID, 841 "VALUES": TokenType.VALUES, 842 "VIEW": TokenType.VIEW, 843 "VOLATILE": TokenType.VOLATILE, 844 "WHEN": TokenType.WHEN, 845 "WHERE": TokenType.WHERE, 846 "WINDOW": TokenType.WINDOW, 847 "WITH": TokenType.WITH, 848 "APPLY": TokenType.APPLY, 849 "ARRAY": TokenType.ARRAY, 850 "BIT": TokenType.BIT, 851 "BOOL": TokenType.BOOLEAN, 852 "BOOLEAN": TokenType.BOOLEAN, 853 "BYTE": TokenType.TINYINT, 854 "MEDIUMINT": TokenType.MEDIUMINT, 855 "INT1": TokenType.TINYINT, 856 "TINYINT": TokenType.TINYINT, 857 "INT16": TokenType.SMALLINT, 858 "SHORT": TokenType.SMALLINT, 859 "SMALLINT": TokenType.SMALLINT, 860 "HUGEINT": TokenType.INT128, 861 "UHUGEINT": TokenType.UINT128, 862 "INT2": TokenType.SMALLINT, 863 "INTEGER": TokenType.INT, 864 "INT": TokenType.INT, 865 "INT4": TokenType.INT, 866 "INT32": TokenType.INT, 867 "INT64": TokenType.BIGINT, 868 "INT128": TokenType.INT128, 869 "INT256": TokenType.INT256, 870 "LONG": TokenType.BIGINT, 871 "BIGINT": TokenType.BIGINT, 872 "INT8": TokenType.TINYINT, 873 "UINT": TokenType.UINT, 874 "UINT128": TokenType.UINT128, 875 "UINT256": TokenType.UINT256, 876 "DEC": TokenType.DECIMAL, 877 "DECIMAL": TokenType.DECIMAL, 878 "DECIMAL32": TokenType.DECIMAL32, 879 "DECIMAL64": TokenType.DECIMAL64, 880 "DECIMAL128": TokenType.DECIMAL128, 881 "DECIMAL256": TokenType.DECIMAL256, 882 "BIGDECIMAL": TokenType.BIGDECIMAL, 883 "BIGNUMERIC": TokenType.BIGDECIMAL, 884 "LIST": TokenType.LIST, 885 "MAP": TokenType.MAP, 886 "NULLABLE": TokenType.NULLABLE, 887 "NUMBER": TokenType.DECIMAL, 888 "NUMERIC": TokenType.DECIMAL, 889 "FIXED": TokenType.DECIMAL, 890 "REAL": TokenType.FLOAT, 891 "FLOAT": TokenType.FLOAT, 892 "FLOAT4": TokenType.FLOAT, 893 "FLOAT8": TokenType.DOUBLE, 894 "DOUBLE": TokenType.DOUBLE, 895 "DOUBLE PRECISION": TokenType.DOUBLE, 896 "JSON": TokenType.JSON, 897 "JSONB": TokenType.JSONB, 898 "CHAR": TokenType.CHAR, 899 "CHARACTER": TokenType.CHAR, 900 "NCHAR": TokenType.NCHAR, 901 "VARCHAR": TokenType.VARCHAR, 902 "VARCHAR2": TokenType.VARCHAR, 903 "NVARCHAR": TokenType.NVARCHAR, 904 "NVARCHAR2": TokenType.NVARCHAR, 905 "BPCHAR": TokenType.BPCHAR, 906 "STR": TokenType.TEXT, 907 "STRING": TokenType.TEXT, 908 "TEXT": TokenType.TEXT, 909 "LONGTEXT": TokenType.LONGTEXT, 910 "MEDIUMTEXT": TokenType.MEDIUMTEXT, 911 "TINYTEXT": TokenType.TINYTEXT, 912 "CLOB": TokenType.TEXT, 913 "LONGVARCHAR": TokenType.TEXT, 914 "BINARY": TokenType.BINARY, 915 "BLOB": TokenType.VARBINARY, 916 "LONGBLOB": TokenType.LONGBLOB, 917 "MEDIUMBLOB": TokenType.MEDIUMBLOB, 918 "TINYBLOB": TokenType.TINYBLOB, 919 "BYTEA": TokenType.VARBINARY, 920 "VARBINARY": TokenType.VARBINARY, 921 "TIME": TokenType.TIME, 922 "TIMETZ": TokenType.TIMETZ, 923 "TIMESTAMP": TokenType.TIMESTAMP, 924 "TIMESTAMPTZ": TokenType.TIMESTAMPTZ, 925 "TIMESTAMPLTZ": TokenType.TIMESTAMPLTZ, 926 "TIMESTAMP_LTZ": TokenType.TIMESTAMPLTZ, 927 "TIMESTAMPNTZ": TokenType.TIMESTAMPNTZ, 928 "TIMESTAMP_NTZ": TokenType.TIMESTAMPNTZ, 929 "DATE": TokenType.DATE, 930 "DATETIME": TokenType.DATETIME, 931 "INT4RANGE": TokenType.INT4RANGE, 932 "INT4MULTIRANGE": TokenType.INT4MULTIRANGE, 933 "INT8RANGE": TokenType.INT8RANGE, 934 "INT8MULTIRANGE": TokenType.INT8MULTIRANGE, 935 "NUMRANGE": TokenType.NUMRANGE, 936 "NUMMULTIRANGE": TokenType.NUMMULTIRANGE, 937 "TSRANGE": TokenType.TSRANGE, 938 "TSMULTIRANGE": TokenType.TSMULTIRANGE, 939 "TSTZRANGE": TokenType.TSTZRANGE, 940 "TSTZMULTIRANGE": TokenType.TSTZMULTIRANGE, 941 "DATERANGE": TokenType.DATERANGE, 942 "DATEMULTIRANGE": TokenType.DATEMULTIRANGE, 943 "UNIQUE": TokenType.UNIQUE, 944 "VECTOR": TokenType.VECTOR, 945 "STRUCT": TokenType.STRUCT, 946 "SEQUENCE": TokenType.SEQUENCE, 947 "VARIANT": TokenType.VARIANT, 948 "ALTER": TokenType.ALTER, 949 "ANALYZE": TokenType.ANALYZE, 950 "CALL": TokenType.COMMAND, 951 "COMMENT": TokenType.COMMENT, 952 "EXPLAIN": TokenType.COMMAND, 953 "GRANT": TokenType.GRANT, 954 "OPTIMIZE": TokenType.COMMAND, 955 "PREPARE": TokenType.COMMAND, 956 "VACUUM": TokenType.COMMAND, 957 "USER-DEFINED": TokenType.USERDEFINED, 958 "FOR VERSION": TokenType.VERSION_SNAPSHOT, 959 "FOR TIMESTAMP": TokenType.TIMESTAMP_SNAPSHOT, 960 } 961 962 WHITE_SPACE: t.Dict[t.Optional[str], TokenType] = { 963 " ": TokenType.SPACE, 964 "\t": TokenType.SPACE, 965 "\n": TokenType.BREAK, 966 "\r": TokenType.BREAK, 967 } 968 969 COMMANDS = { 970 TokenType.COMMAND, 971 TokenType.EXECUTE, 972 TokenType.FETCH, 973 TokenType.SHOW, 974 TokenType.RENAME, 975 } 976 977 COMMAND_PREFIX_TOKENS = {TokenType.SEMICOLON, TokenType.BEGIN} 978 979 # Handle numeric literals like in hive (3L = BIGINT) 980 NUMERIC_LITERALS: t.Dict[str, str] = {} 981 982 COMMENTS = ["--", ("/*", "*/")] 983 984 __slots__ = ( 985 "sql", 986 "size", 987 "tokens", 988 "dialect", 989 "use_rs_tokenizer", 990 "_start", 991 "_current", 992 "_line", 993 "_col", 994 "_comments", 995 "_char", 996 "_end", 997 "_peek", 998 "_prev_token_line", 999 "_rs_dialect_settings", 1000 ) 1001 1002 def __init__( 1003 self, dialect: DialectType = None, use_rs_tokenizer: t.Optional[bool] = None 1004 ) -> None: 1005 from sqlglot.dialects import Dialect 1006 1007 self.dialect = Dialect.get_or_raise(dialect) 1008 1009 # initialize `use_rs_tokenizer`, and allow it to be overwritten per Tokenizer instance 1010 self.use_rs_tokenizer = ( 1011 use_rs_tokenizer if use_rs_tokenizer is not None else USE_RS_TOKENIZER 1012 ) 1013 1014 if self.use_rs_tokenizer: 1015 self._rs_dialect_settings = RsTokenizerDialectSettings( 1016 unescaped_sequences=self.dialect.UNESCAPED_SEQUENCES, 1017 identifiers_can_start_with_digit=self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT, 1018 numbers_can_be_underscore_separated=self.dialect.NUMBERS_CAN_BE_UNDERSCORE_SEPARATED, 1019 ) 1020 1021 self.reset() 1022 1023 def reset(self) -> None: 1024 self.sql = "" 1025 self.size = 0 1026 self.tokens: t.List[Token] = [] 1027 self._start = 0 1028 self._current = 0 1029 self._line = 1 1030 self._col = 0 1031 self._comments: t.List[str] = [] 1032 1033 self._char = "" 1034 self._end = False 1035 self._peek = "" 1036 self._prev_token_line = -1 1037 1038 def tokenize(self, sql: str) -> t.List[Token]: 1039 """Returns a list of tokens corresponding to the SQL string `sql`.""" 1040 if self.use_rs_tokenizer: 1041 return self.tokenize_rs(sql) 1042 1043 self.reset() 1044 self.sql = sql 1045 self.size = len(sql) 1046 1047 try: 1048 self._scan() 1049 except Exception as e: 1050 start = max(self._current - 50, 0) 1051 end = min(self._current + 50, self.size - 1) 1052 context = self.sql[start:end] 1053 raise TokenError(f"Error tokenizing '{context}'") from e 1054 1055 return self.tokens 1056 1057 def _scan(self, until: t.Optional[t.Callable] = None) -> None: 1058 while self.size and not self._end: 1059 current = self._current 1060 1061 # Skip spaces here rather than iteratively calling advance() for performance reasons 1062 while current < self.size: 1063 char = self.sql[current] 1064 1065 if char.isspace() and (char == " " or char == "\t"): 1066 current += 1 1067 else: 1068 break 1069 1070 offset = current - self._current if current > self._current else 1 1071 1072 self._start = current 1073 self._advance(offset) 1074 1075 if not self._char.isspace(): 1076 if self._char.isdigit(): 1077 self._scan_number() 1078 elif self._char in self._IDENTIFIERS: 1079 self._scan_identifier(self._IDENTIFIERS[self._char]) 1080 else: 1081 self._scan_keywords() 1082 1083 if until and until(): 1084 break 1085 1086 if self.tokens and self._comments: 1087 self.tokens[-1].comments.extend(self._comments) 1088 1089 def _chars(self, size: int) -> str: 1090 if size == 1: 1091 return self._char 1092 1093 start = self._current - 1 1094 end = start + size 1095 1096 return self.sql[start:end] if end <= self.size else "" 1097 1098 def _advance(self, i: int = 1, alnum: bool = False) -> None: 1099 if self.WHITE_SPACE.get(self._char) is TokenType.BREAK: 1100 # Ensures we don't count an extra line if we get a \r\n line break sequence 1101 if not (self._char == "\r" and self._peek == "\n"): 1102 self._col = 1 1103 self._line += 1 1104 else: 1105 self._col += i 1106 1107 self._current += i 1108 self._end = self._current >= self.size 1109 self._char = self.sql[self._current - 1] 1110 self._peek = "" if self._end else self.sql[self._current] 1111 1112 if alnum and self._char.isalnum(): 1113 # Here we use local variables instead of attributes for better performance 1114 _col = self._col 1115 _current = self._current 1116 _end = self._end 1117 _peek = self._peek 1118 1119 while _peek.isalnum(): 1120 _col += 1 1121 _current += 1 1122 _end = _current >= self.size 1123 _peek = "" if _end else self.sql[_current] 1124 1125 self._col = _col 1126 self._current = _current 1127 self._end = _end 1128 self._peek = _peek 1129 self._char = self.sql[_current - 1] 1130 1131 @property 1132 def _text(self) -> str: 1133 return self.sql[self._start : self._current] 1134 1135 def _add(self, token_type: TokenType, text: t.Optional[str] = None) -> None: 1136 self._prev_token_line = self._line 1137 1138 if self._comments and token_type == TokenType.SEMICOLON and self.tokens: 1139 self.tokens[-1].comments.extend(self._comments) 1140 self._comments = [] 1141 1142 self.tokens.append( 1143 Token( 1144 token_type, 1145 text=self._text if text is None else text, 1146 line=self._line, 1147 col=self._col, 1148 start=self._start, 1149 end=self._current - 1, 1150 comments=self._comments, 1151 ) 1152 ) 1153 self._comments = [] 1154 1155 # If we have either a semicolon or a begin token before the command's token, we'll parse 1156 # whatever follows the command's token as a string 1157 if ( 1158 token_type in self.COMMANDS 1159 and self._peek != ";" 1160 and (len(self.tokens) == 1 or self.tokens[-2].token_type in self.COMMAND_PREFIX_TOKENS) 1161 ): 1162 start = self._current 1163 tokens = len(self.tokens) 1164 self._scan(lambda: self._peek == ";") 1165 self.tokens = self.tokens[:tokens] 1166 text = self.sql[start : self._current].strip() 1167 if text: 1168 self._add(TokenType.STRING, text) 1169 1170 def _scan_keywords(self) -> None: 1171 size = 0 1172 word = None 1173 chars = self._text 1174 char = chars 1175 prev_space = False 1176 skip = False 1177 trie = self._KEYWORD_TRIE 1178 single_token = char in self.SINGLE_TOKENS 1179 1180 while chars: 1181 if skip: 1182 result = TrieResult.PREFIX 1183 else: 1184 result, trie = in_trie(trie, char.upper()) 1185 1186 if result == TrieResult.FAILED: 1187 break 1188 if result == TrieResult.EXISTS: 1189 word = chars 1190 1191 end = self._current + size 1192 size += 1 1193 1194 if end < self.size: 1195 char = self.sql[end] 1196 single_token = single_token or char in self.SINGLE_TOKENS 1197 is_space = char.isspace() 1198 1199 if not is_space or not prev_space: 1200 if is_space: 1201 char = " " 1202 chars += char 1203 prev_space = is_space 1204 skip = False 1205 else: 1206 skip = True 1207 else: 1208 char = "" 1209 break 1210 1211 if word: 1212 if self._scan_string(word): 1213 return 1214 if self._scan_comment(word): 1215 return 1216 if prev_space or single_token or not char: 1217 self._advance(size - 1) 1218 word = word.upper() 1219 self._add(self.KEYWORDS[word], text=word) 1220 return 1221 1222 if self._char in self.SINGLE_TOKENS: 1223 self._add(self.SINGLE_TOKENS[self._char], text=self._char) 1224 return 1225 1226 self._scan_var() 1227 1228 def _scan_comment(self, comment_start: str) -> bool: 1229 if comment_start not in self._COMMENTS: 1230 return False 1231 1232 comment_start_line = self._line 1233 comment_start_size = len(comment_start) 1234 comment_end = self._COMMENTS[comment_start] 1235 1236 if comment_end: 1237 # Skip the comment's start delimiter 1238 self._advance(comment_start_size) 1239 1240 comment_count = 1 1241 comment_end_size = len(comment_end) 1242 1243 while not self._end: 1244 if self._chars(comment_end_size) == comment_end: 1245 comment_count -= 1 1246 if not comment_count: 1247 break 1248 1249 self._advance(alnum=True) 1250 1251 # Nested comments are allowed by some dialects, e.g. databricks, duckdb, postgres 1252 if ( 1253 self.NESTED_COMMENTS 1254 and not self._end 1255 and self._chars(comment_end_size) == comment_start 1256 ): 1257 self._advance(comment_start_size) 1258 comment_count += 1 1259 1260 self._comments.append(self._text[comment_start_size : -comment_end_size + 1]) 1261 self._advance(comment_end_size - 1) 1262 else: 1263 while not self._end and self.WHITE_SPACE.get(self._peek) is not TokenType.BREAK: 1264 self._advance(alnum=True) 1265 self._comments.append(self._text[comment_start_size:]) 1266 1267 if ( 1268 comment_start == self.HINT_START 1269 and self.tokens 1270 and self.tokens[-1].token_type in self.TOKENS_PRECEDING_HINT 1271 ): 1272 self._add(TokenType.HINT) 1273 1274 # Leading comment is attached to the succeeding token, whilst trailing comment to the preceding. 1275 # Multiple consecutive comments are preserved by appending them to the current comments list. 1276 if comment_start_line == self._prev_token_line: 1277 self.tokens[-1].comments.extend(self._comments) 1278 self._comments = [] 1279 self._prev_token_line = self._line 1280 1281 return True 1282 1283 def _scan_number(self) -> None: 1284 if self._char == "0": 1285 peek = self._peek.upper() 1286 if peek == "B": 1287 return self._scan_bits() if self.BIT_STRINGS else self._add(TokenType.NUMBER) 1288 elif peek == "X": 1289 return self._scan_hex() if self.HEX_STRINGS else self._add(TokenType.NUMBER) 1290 1291 decimal = False 1292 scientific = 0 1293 1294 while True: 1295 if self._peek.isdigit(): 1296 self._advance() 1297 elif self._peek == "." and not decimal: 1298 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER: 1299 return self._add(TokenType.NUMBER) 1300 decimal = True 1301 self._advance() 1302 elif self._peek in ("-", "+") and scientific == 1: 1303 scientific += 1 1304 self._advance() 1305 elif self._peek.upper() == "E" and not scientific: 1306 scientific += 1 1307 self._advance() 1308 elif self._peek.isidentifier(): 1309 number_text = self._text 1310 literal = "" 1311 1312 while self._peek.strip() and self._peek not in self.SINGLE_TOKENS: 1313 literal += self._peek 1314 self._advance() 1315 1316 token_type = self.KEYWORDS.get(self.NUMERIC_LITERALS.get(literal.upper(), "")) 1317 1318 if token_type: 1319 self._add(TokenType.NUMBER, number_text) 1320 self._add(TokenType.DCOLON, "::") 1321 return self._add(token_type, literal) 1322 else: 1323 replaced = literal.replace("_", "") 1324 if self.dialect.NUMBERS_CAN_BE_UNDERSCORE_SEPARATED and replaced.isdigit(): 1325 return self._add(TokenType.NUMBER, number_text + replaced) 1326 if self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT: 1327 return self._add(TokenType.VAR) 1328 1329 self._advance(-len(literal)) 1330 return self._add(TokenType.NUMBER, number_text) 1331 else: 1332 return self._add(TokenType.NUMBER) 1333 1334 def _scan_bits(self) -> None: 1335 self._advance() 1336 value = self._extract_value() 1337 try: 1338 # If `value` can't be converted to a binary, fallback to tokenizing it as an identifier 1339 int(value, 2) 1340 self._add(TokenType.BIT_STRING, value[2:]) # Drop the 0b 1341 except ValueError: 1342 self._add(TokenType.IDENTIFIER) 1343 1344 def _scan_hex(self) -> None: 1345 self._advance() 1346 value = self._extract_value() 1347 try: 1348 # If `value` can't be converted to a hex, fallback to tokenizing it as an identifier 1349 int(value, 16) 1350 self._add(TokenType.HEX_STRING, value[2:]) # Drop the 0x 1351 except ValueError: 1352 self._add(TokenType.IDENTIFIER) 1353 1354 def _extract_value(self) -> str: 1355 while True: 1356 char = self._peek.strip() 1357 if char and char not in self.SINGLE_TOKENS: 1358 self._advance(alnum=True) 1359 else: 1360 break 1361 1362 return self._text 1363 1364 def _scan_string(self, start: str) -> bool: 1365 base = None 1366 token_type = TokenType.STRING 1367 1368 if start in self._QUOTES: 1369 end = self._QUOTES[start] 1370 elif start in self._FORMAT_STRINGS: 1371 end, token_type = self._FORMAT_STRINGS[start] 1372 1373 if token_type == TokenType.HEX_STRING: 1374 base = 16 1375 elif token_type == TokenType.BIT_STRING: 1376 base = 2 1377 elif token_type == TokenType.HEREDOC_STRING: 1378 self._advance() 1379 1380 if self._char == end: 1381 tag = "" 1382 else: 1383 tag = self._extract_string( 1384 end, 1385 raw_string=True, 1386 raise_unmatched=not self.HEREDOC_TAG_IS_IDENTIFIER, 1387 ) 1388 1389 if tag and self.HEREDOC_TAG_IS_IDENTIFIER and (self._end or not tag.isidentifier()): 1390 if not self._end: 1391 self._advance(-1) 1392 1393 self._advance(-len(tag)) 1394 self._add(self.HEREDOC_STRING_ALTERNATIVE) 1395 return True 1396 1397 end = f"{start}{tag}{end}" 1398 else: 1399 return False 1400 1401 self._advance(len(start)) 1402 text = self._extract_string(end, raw_string=token_type == TokenType.RAW_STRING) 1403 1404 if base: 1405 try: 1406 int(text, base) 1407 except Exception: 1408 raise TokenError( 1409 f"Numeric string contains invalid characters from {self._line}:{self._start}" 1410 ) 1411 1412 self._add(token_type, text) 1413 return True 1414 1415 def _scan_identifier(self, identifier_end: str) -> None: 1416 self._advance() 1417 text = self._extract_string( 1418 identifier_end, escapes=self._IDENTIFIER_ESCAPES | {identifier_end} 1419 ) 1420 self._add(TokenType.IDENTIFIER, text) 1421 1422 def _scan_var(self) -> None: 1423 while True: 1424 char = self._peek.strip() 1425 if char and (char in self.VAR_SINGLE_TOKENS or char not in self.SINGLE_TOKENS): 1426 self._advance(alnum=True) 1427 else: 1428 break 1429 1430 self._add( 1431 TokenType.VAR 1432 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER 1433 else self.KEYWORDS.get(self._text.upper(), TokenType.VAR) 1434 ) 1435 1436 def _extract_string( 1437 self, 1438 delimiter: str, 1439 escapes: t.Optional[t.Set[str]] = None, 1440 raw_string: bool = False, 1441 raise_unmatched: bool = True, 1442 ) -> str: 1443 text = "" 1444 delim_size = len(delimiter) 1445 escapes = self._STRING_ESCAPES if escapes is None else escapes 1446 1447 while True: 1448 if ( 1449 not raw_string 1450 and self.dialect.UNESCAPED_SEQUENCES 1451 and self._peek 1452 and self._char in self.STRING_ESCAPES 1453 ): 1454 unescaped_sequence = self.dialect.UNESCAPED_SEQUENCES.get(self._char + self._peek) 1455 if unescaped_sequence: 1456 self._advance(2) 1457 text += unescaped_sequence 1458 continue 1459 if ( 1460 (self.STRING_ESCAPES_ALLOWED_IN_RAW_STRINGS or not raw_string) 1461 and self._char in escapes 1462 and (self._peek == delimiter or self._peek in escapes) 1463 and (self._char not in self._QUOTES or self._char == self._peek) 1464 ): 1465 if self._peek == delimiter: 1466 text += self._peek 1467 else: 1468 text += self._char + self._peek 1469 1470 if self._current + 1 < self.size: 1471 self._advance(2) 1472 else: 1473 raise TokenError(f"Missing {delimiter} from {self._line}:{self._current}") 1474 else: 1475 if self._chars(delim_size) == delimiter: 1476 if delim_size > 1: 1477 self._advance(delim_size - 1) 1478 break 1479 1480 if self._end: 1481 if not raise_unmatched: 1482 return text + self._char 1483 1484 raise TokenError(f"Missing {delimiter} from {self._line}:{self._start}") 1485 1486 current = self._current - 1 1487 self._advance(alnum=True) 1488 text += self.sql[current : self._current - 1] 1489 1490 return text 1491 1492 def tokenize_rs(self, sql: str) -> t.List[Token]: 1493 if not self._RS_TOKENIZER: 1494 raise SqlglotError("Rust tokenizer is not available") 1495 1496 try: 1497 tokens = self._RS_TOKENIZER.tokenize(sql, self._rs_dialect_settings) 1498 for token in tokens: 1499 token.token_type = _ALL_TOKEN_TYPES[token.token_type_index] 1500 return tokens 1501 except Exception as e: 1502 raise TokenError(str(e))
Tokenizer( dialect: Union[str, sqlglot.dialects.Dialect, Type[sqlglot.dialects.Dialect], NoneType] = None, use_rs_tokenizer: Optional[bool] = None)
1002 def __init__( 1003 self, dialect: DialectType = None, use_rs_tokenizer: t.Optional[bool] = None 1004 ) -> None: 1005 from sqlglot.dialects import Dialect 1006 1007 self.dialect = Dialect.get_or_raise(dialect) 1008 1009 # initialize `use_rs_tokenizer`, and allow it to be overwritten per Tokenizer instance 1010 self.use_rs_tokenizer = ( 1011 use_rs_tokenizer if use_rs_tokenizer is not None else USE_RS_TOKENIZER 1012 ) 1013 1014 if self.use_rs_tokenizer: 1015 self._rs_dialect_settings = RsTokenizerDialectSettings( 1016 unescaped_sequences=self.dialect.UNESCAPED_SEQUENCES, 1017 identifiers_can_start_with_digit=self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT, 1018 numbers_can_be_underscore_separated=self.dialect.NUMBERS_CAN_BE_UNDERSCORE_SEPARATED, 1019 ) 1020 1021 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'>
TOKENS_PRECEDING_HINT =
{<TokenType.UPDATE: 'UPDATE'>, <TokenType.INSERT: 'INSERT'>, <TokenType.DELETE: 'DELETE'>, <TokenType.SELECT: 'SELECT'>}
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'>, '~~~': <TokenType.GLOB: 'GLOB'>, '~~': <TokenType.LIKE: 'LIKE'>, '~~*': <TokenType.ILIKE: 'ILIKE'>, '~*': <TokenType.IRLIKE: 'IRLIKE'>, '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_SCHEMA': <TokenType.CURRENT_SCHEMA: 'CURRENT_SCHEMA'>, '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'>, 'NAMESPACE': <TokenType.NAMESPACE: 'NAMESPACE'>, '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'>, 'RENAME': <TokenType.RENAME: 'RENAME'>, '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'>, 'STRAIGHT_JOIN': <TokenType.STRAIGHT_JOIN: 'STRAIGHT_JOIN'>, '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'>, 'HUGEINT': <TokenType.INT128: 'INT128'>, 'UHUGEINT': <TokenType.UINT128: 'UINT128'>, 'INT2': <TokenType.SMALLINT: 'SMALLINT'>, 'INTEGER': <TokenType.INT: 'INT'>, 'INT': <TokenType.INT: 'INT'>, 'INT4': <TokenType.INT: 'INT'>, 'INT32': <TokenType.INT: 'INT'>, 'INT64': <TokenType.BIGINT: 'BIGINT'>, 'INT128': <TokenType.INT128: 'INT128'>, 'INT256': <TokenType.INT256: 'INT256'>, 'LONG': <TokenType.BIGINT: 'BIGINT'>, 'BIGINT': <TokenType.BIGINT: 'BIGINT'>, 'INT8': <TokenType.TINYINT: 'TINYINT'>, 'UINT': <TokenType.UINT: 'UINT'>, 'UINT128': <TokenType.UINT128: 'UINT128'>, 'UINT256': <TokenType.UINT256: 'UINT256'>, 'DEC': <TokenType.DECIMAL: 'DECIMAL'>, 'DECIMAL': <TokenType.DECIMAL: 'DECIMAL'>, 'DECIMAL32': <TokenType.DECIMAL32: 'DECIMAL32'>, 'DECIMAL64': <TokenType.DECIMAL64: 'DECIMAL64'>, 'DECIMAL128': <TokenType.DECIMAL128: 'DECIMAL128'>, 'DECIMAL256': <TokenType.DECIMAL256: 'DECIMAL256'>, 'BIGDECIMAL': <TokenType.BIGDECIMAL: 'BIGDECIMAL'>, 'BIGNUMERIC': <TokenType.BIGDECIMAL: 'BIGDECIMAL'>, 'LIST': <TokenType.LIST: 'LIST'>, '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'>, 'JSONB': <TokenType.JSONB: 'JSONB'>, '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'>, 'VECTOR': <TokenType.VECTOR: 'VECTOR'>, 'STRUCT': <TokenType.STRUCT: 'STRUCT'>, 'SEQUENCE': <TokenType.SEQUENCE: 'SEQUENCE'>, 'VARIANT': <TokenType.VARIANT: 'VARIANT'>, 'ALTER': <TokenType.ALTER: 'ALTER'>, 'ANALYZE': <TokenType.ANALYZE: 'ANALYZE'>, 'CALL': <TokenType.COMMAND: 'COMMAND'>, 'COMMENT': <TokenType.COMMENT: 'COMMENT'>, 'EXPLAIN': <TokenType.COMMAND: 'COMMAND'>, 'GRANT': <TokenType.GRANT: 'GRANT'>, '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.FETCH: 'FETCH'>, <TokenType.RENAME: 'RENAME'>, <TokenType.COMMAND: 'COMMAND'>, <TokenType.EXECUTE: 'EXECUTE'>, <TokenType.SHOW: 'SHOW'>}
COMMAND_PREFIX_TOKENS =
{<TokenType.BEGIN: 'BEGIN'>, <TokenType.SEMICOLON: 'SEMICOLON'>}
def
reset(self) -> None:
1023 def reset(self) -> None: 1024 self.sql = "" 1025 self.size = 0 1026 self.tokens: t.List[Token] = [] 1027 self._start = 0 1028 self._current = 0 1029 self._line = 1 1030 self._col = 0 1031 self._comments: t.List[str] = [] 1032 1033 self._char = "" 1034 self._end = False 1035 self._peek = "" 1036 self._prev_token_line = -1
1038 def tokenize(self, sql: str) -> t.List[Token]: 1039 """Returns a list of tokens corresponding to the SQL string `sql`.""" 1040 if self.use_rs_tokenizer: 1041 return self.tokenize_rs(sql) 1042 1043 self.reset() 1044 self.sql = sql 1045 self.size = len(sql) 1046 1047 try: 1048 self._scan() 1049 except Exception as e: 1050 start = max(self._current - 50, 0) 1051 end = min(self._current + 50, self.size - 1) 1052 context = self.sql[start:end] 1053 raise TokenError(f"Error tokenizing '{context}'") from e 1054 1055 return self.tokens
Returns a list of tokens corresponding to the SQL string sql
.
1492 def tokenize_rs(self, sql: str) -> t.List[Token]: 1493 if not self._RS_TOKENIZER: 1494 raise SqlglotError("Rust tokenizer is not available") 1495 1496 try: 1497 tokens = self._RS_TOKENIZER.tokenize(sql, self._rs_dialect_settings) 1498 for token in tokens: 1499 token.token_type = _ALL_TOKEN_TYPES[token.token_type_index] 1500 return tokens 1501 except Exception as e: 1502 raise TokenError(str(e))