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 DOTCOLON = 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 URI_START = auto() 84 85 BLOCK_START = auto() 86 BLOCK_END = auto() 87 88 SPACE = auto() 89 BREAK = auto() 90 91 STRING = auto() 92 NUMBER = auto() 93 IDENTIFIER = auto() 94 DATABASE = auto() 95 COLUMN = auto() 96 COLUMN_DEF = auto() 97 SCHEMA = auto() 98 TABLE = auto() 99 WAREHOUSE = auto() 100 STAGE = auto() 101 STREAMLIT = auto() 102 VAR = auto() 103 BIT_STRING = auto() 104 HEX_STRING = auto() 105 BYTE_STRING = auto() 106 NATIONAL_STRING = auto() 107 RAW_STRING = auto() 108 HEREDOC_STRING = auto() 109 UNICODE_STRING = auto() 110 111 # types 112 BIT = auto() 113 BOOLEAN = auto() 114 TINYINT = auto() 115 UTINYINT = auto() 116 SMALLINT = auto() 117 USMALLINT = auto() 118 MEDIUMINT = auto() 119 UMEDIUMINT = auto() 120 INT = auto() 121 UINT = auto() 122 BIGINT = auto() 123 UBIGINT = auto() 124 INT128 = auto() 125 UINT128 = auto() 126 INT256 = auto() 127 UINT256 = auto() 128 FLOAT = auto() 129 DOUBLE = auto() 130 UDOUBLE = auto() 131 DECIMAL = auto() 132 DECIMAL32 = auto() 133 DECIMAL64 = auto() 134 DECIMAL128 = auto() 135 DECIMAL256 = auto() 136 UDECIMAL = auto() 137 BIGDECIMAL = auto() 138 CHAR = auto() 139 NCHAR = auto() 140 VARCHAR = auto() 141 NVARCHAR = auto() 142 BPCHAR = auto() 143 TEXT = auto() 144 MEDIUMTEXT = auto() 145 LONGTEXT = auto() 146 BLOB = auto() 147 MEDIUMBLOB = auto() 148 LONGBLOB = auto() 149 TINYBLOB = auto() 150 TINYTEXT = auto() 151 NAME = auto() 152 BINARY = auto() 153 VARBINARY = auto() 154 JSON = auto() 155 JSONB = auto() 156 TIME = auto() 157 TIMETZ = auto() 158 TIMESTAMP = auto() 159 TIMESTAMPTZ = auto() 160 TIMESTAMPLTZ = auto() 161 TIMESTAMPNTZ = auto() 162 TIMESTAMP_S = auto() 163 TIMESTAMP_MS = auto() 164 TIMESTAMP_NS = auto() 165 DATETIME = auto() 166 DATETIME2 = auto() 167 DATETIME64 = auto() 168 SMALLDATETIME = auto() 169 DATE = auto() 170 DATE32 = auto() 171 INT4RANGE = auto() 172 INT4MULTIRANGE = auto() 173 INT8RANGE = auto() 174 INT8MULTIRANGE = auto() 175 NUMRANGE = auto() 176 NUMMULTIRANGE = auto() 177 TSRANGE = auto() 178 TSMULTIRANGE = auto() 179 TSTZRANGE = auto() 180 TSTZMULTIRANGE = auto() 181 DATERANGE = auto() 182 DATEMULTIRANGE = auto() 183 UUID = auto() 184 GEOGRAPHY = auto() 185 NULLABLE = auto() 186 GEOMETRY = auto() 187 POINT = auto() 188 RING = auto() 189 LINESTRING = auto() 190 MULTILINESTRING = auto() 191 POLYGON = auto() 192 MULTIPOLYGON = auto() 193 HLLSKETCH = auto() 194 HSTORE = auto() 195 SUPER = auto() 196 SERIAL = auto() 197 SMALLSERIAL = auto() 198 BIGSERIAL = auto() 199 XML = auto() 200 YEAR = auto() 201 USERDEFINED = auto() 202 MONEY = auto() 203 SMALLMONEY = auto() 204 ROWVERSION = auto() 205 IMAGE = auto() 206 VARIANT = auto() 207 OBJECT = auto() 208 INET = auto() 209 IPADDRESS = auto() 210 IPPREFIX = auto() 211 IPV4 = auto() 212 IPV6 = auto() 213 ENUM = auto() 214 ENUM8 = auto() 215 ENUM16 = auto() 216 FIXEDSTRING = auto() 217 LOWCARDINALITY = auto() 218 NESTED = auto() 219 AGGREGATEFUNCTION = auto() 220 SIMPLEAGGREGATEFUNCTION = auto() 221 TDIGEST = auto() 222 UNKNOWN = auto() 223 VECTOR = auto() 224 DYNAMIC = auto() 225 226 # keywords 227 ALIAS = auto() 228 ALTER = auto() 229 ALWAYS = auto() 230 ALL = auto() 231 ANTI = auto() 232 ANY = auto() 233 APPLY = auto() 234 ARRAY = auto() 235 ASC = auto() 236 ASOF = auto() 237 ATTACH = auto() 238 AUTO_INCREMENT = auto() 239 BEGIN = auto() 240 BETWEEN = auto() 241 BULK_COLLECT_INTO = auto() 242 CACHE = auto() 243 CASE = auto() 244 CHARACTER_SET = auto() 245 CLUSTER_BY = auto() 246 COLLATE = auto() 247 COMMAND = auto() 248 COMMENT = auto() 249 COMMIT = auto() 250 CONNECT_BY = auto() 251 CONSTRAINT = auto() 252 COPY = auto() 253 CREATE = auto() 254 CROSS = auto() 255 CUBE = auto() 256 CURRENT_DATE = auto() 257 CURRENT_DATETIME = auto() 258 CURRENT_SCHEMA = auto() 259 CURRENT_TIME = auto() 260 CURRENT_TIMESTAMP = auto() 261 CURRENT_USER = auto() 262 DECLARE = auto() 263 DEFAULT = auto() 264 DELETE = auto() 265 DESC = auto() 266 DESCRIBE = auto() 267 DETACH = auto() 268 DICTIONARY = auto() 269 DISTINCT = auto() 270 DISTRIBUTE_BY = auto() 271 DIV = auto() 272 DROP = auto() 273 ELSE = auto() 274 END = auto() 275 ESCAPE = auto() 276 EXCEPT = auto() 277 EXECUTE = auto() 278 EXISTS = auto() 279 FALSE = auto() 280 FETCH = auto() 281 FILE_FORMAT = auto() 282 FILTER = auto() 283 FINAL = auto() 284 FIRST = auto() 285 FOR = auto() 286 FORCE = auto() 287 FOREIGN_KEY = auto() 288 FORMAT = auto() 289 FROM = auto() 290 FULL = auto() 291 FUNCTION = auto() 292 GLOB = auto() 293 GLOBAL = auto() 294 GRANT = auto() 295 GROUP_BY = auto() 296 GROUPING_SETS = auto() 297 HAVING = auto() 298 HINT = auto() 299 IGNORE = auto() 300 ILIKE = auto() 301 ILIKE_ANY = auto() 302 IN = auto() 303 INDEX = auto() 304 INNER = auto() 305 INSERT = auto() 306 INTERSECT = auto() 307 INTERVAL = auto() 308 INTO = auto() 309 INTRODUCER = auto() 310 IRLIKE = auto() 311 IS = auto() 312 ISNULL = auto() 313 JOIN = auto() 314 JOIN_MARKER = auto() 315 KEEP = auto() 316 KEY = auto() 317 KILL = auto() 318 LANGUAGE = auto() 319 LATERAL = auto() 320 LEFT = auto() 321 LIKE = auto() 322 LIKE_ANY = auto() 323 LIMIT = auto() 324 LIST = auto() 325 LOAD = auto() 326 LOCK = auto() 327 MAP = auto() 328 MATCH_CONDITION = auto() 329 MATCH_RECOGNIZE = auto() 330 MEMBER_OF = auto() 331 MERGE = auto() 332 MOD = auto() 333 MODEL = auto() 334 NATURAL = auto() 335 NEXT = auto() 336 NOTHING = auto() 337 NOTNULL = auto() 338 NULL = auto() 339 OBJECT_IDENTIFIER = auto() 340 OFFSET = auto() 341 ON = auto() 342 ONLY = auto() 343 OPERATOR = auto() 344 ORDER_BY = auto() 345 ORDER_SIBLINGS_BY = auto() 346 ORDERED = auto() 347 ORDINALITY = auto() 348 OUTER = auto() 349 OVER = auto() 350 OVERLAPS = auto() 351 OVERWRITE = auto() 352 PARTITION = auto() 353 PARTITION_BY = auto() 354 PERCENT = auto() 355 PIVOT = auto() 356 PLACEHOLDER = auto() 357 POSITIONAL = auto() 358 PRAGMA = auto() 359 PREWHERE = auto() 360 PRIMARY_KEY = auto() 361 PROCEDURE = auto() 362 PROPERTIES = auto() 363 PSEUDO_TYPE = auto() 364 PUT = auto() 365 QUALIFY = auto() 366 QUOTE = auto() 367 RANGE = auto() 368 RECURSIVE = auto() 369 REFRESH = auto() 370 RENAME = auto() 371 REPLACE = auto() 372 RETURNING = auto() 373 REFERENCES = auto() 374 RIGHT = auto() 375 RLIKE = auto() 376 ROLLBACK = auto() 377 ROLLUP = auto() 378 ROW = auto() 379 ROWS = auto() 380 SELECT = auto() 381 SEMI = auto() 382 SEPARATOR = auto() 383 SEQUENCE = auto() 384 SERDE_PROPERTIES = auto() 385 SET = auto() 386 SETTINGS = auto() 387 SHOW = auto() 388 SIMILAR_TO = auto() 389 SOME = auto() 390 SORT_BY = auto() 391 START_WITH = auto() 392 STORAGE_INTEGRATION = auto() 393 STRAIGHT_JOIN = auto() 394 STRUCT = auto() 395 SUMMARIZE = auto() 396 TABLE_SAMPLE = auto() 397 TAG = auto() 398 TEMPORARY = auto() 399 TOP = auto() 400 THEN = auto() 401 TRUE = auto() 402 TRUNCATE = auto() 403 UNCACHE = auto() 404 UNION = auto() 405 UNNEST = auto() 406 UNPIVOT = auto() 407 UPDATE = auto() 408 USE = auto() 409 USING = auto() 410 VALUES = auto() 411 VIEW = auto() 412 VOLATILE = auto() 413 WHEN = auto() 414 WHERE = auto() 415 WINDOW = auto() 416 WITH = auto() 417 UNIQUE = auto() 418 VERSION_SNAPSHOT = auto() 419 TIMESTAMP_SNAPSHOT = auto() 420 OPTION = auto() 421 SINK = auto() 422 SOURCE = auto() 423 ANALYZE = auto() 424 NAMESPACE = auto() 425 EXPORT = auto() 426 427 428_ALL_TOKEN_TYPES = list(TokenType) 429_TOKEN_TYPE_TO_INDEX = {token_type: i for i, token_type in enumerate(_ALL_TOKEN_TYPES)} 430 431 432class Token: 433 __slots__ = ("token_type", "text", "line", "col", "start", "end", "comments") 434 435 @classmethod 436 def number(cls, number: int) -> Token: 437 """Returns a NUMBER token with `number` as its text.""" 438 return cls(TokenType.NUMBER, str(number)) 439 440 @classmethod 441 def string(cls, string: str) -> Token: 442 """Returns a STRING token with `string` as its text.""" 443 return cls(TokenType.STRING, string) 444 445 @classmethod 446 def identifier(cls, identifier: str) -> Token: 447 """Returns an IDENTIFIER token with `identifier` as its text.""" 448 return cls(TokenType.IDENTIFIER, identifier) 449 450 @classmethod 451 def var(cls, var: str) -> Token: 452 """Returns an VAR token with `var` as its text.""" 453 return cls(TokenType.VAR, var) 454 455 def __init__( 456 self, 457 token_type: TokenType, 458 text: str, 459 line: int = 1, 460 col: int = 1, 461 start: int = 0, 462 end: int = 0, 463 comments: t.Optional[t.List[str]] = None, 464 ) -> None: 465 """Token initializer. 466 467 Args: 468 token_type: The TokenType Enum. 469 text: The text of the token. 470 line: The line that the token ends on. 471 col: The column that the token ends on. 472 start: The start index of the token. 473 end: The ending index of the token. 474 comments: The comments to attach to the token. 475 """ 476 self.token_type = token_type 477 self.text = text 478 self.line = line 479 self.col = col 480 self.start = start 481 self.end = end 482 self.comments = [] if comments is None else comments 483 484 def __repr__(self) -> str: 485 attributes = ", ".join(f"{k}: {getattr(self, k)}" for k in self.__slots__) 486 return f"<Token {attributes}>" 487 488 489class _Tokenizer(type): 490 def __new__(cls, clsname, bases, attrs): 491 klass = super().__new__(cls, clsname, bases, attrs) 492 493 def _convert_quotes(arr: t.List[str | t.Tuple[str, str]]) -> t.Dict[str, str]: 494 return dict( 495 (item, item) if isinstance(item, str) else (item[0], item[1]) for item in arr 496 ) 497 498 def _quotes_to_format( 499 token_type: TokenType, arr: t.List[str | t.Tuple[str, str]] 500 ) -> t.Dict[str, t.Tuple[str, TokenType]]: 501 return {k: (v, token_type) for k, v in _convert_quotes(arr).items()} 502 503 klass._QUOTES = _convert_quotes(klass.QUOTES) 504 klass._IDENTIFIERS = _convert_quotes(klass.IDENTIFIERS) 505 506 klass._FORMAT_STRINGS = { 507 **{ 508 p + s: (e, TokenType.NATIONAL_STRING) 509 for s, e in klass._QUOTES.items() 510 for p in ("n", "N") 511 }, 512 **_quotes_to_format(TokenType.BIT_STRING, klass.BIT_STRINGS), 513 **_quotes_to_format(TokenType.BYTE_STRING, klass.BYTE_STRINGS), 514 **_quotes_to_format(TokenType.HEX_STRING, klass.HEX_STRINGS), 515 **_quotes_to_format(TokenType.RAW_STRING, klass.RAW_STRINGS), 516 **_quotes_to_format(TokenType.HEREDOC_STRING, klass.HEREDOC_STRINGS), 517 **_quotes_to_format(TokenType.UNICODE_STRING, klass.UNICODE_STRINGS), 518 } 519 520 klass._STRING_ESCAPES = set(klass.STRING_ESCAPES) 521 klass._IDENTIFIER_ESCAPES = set(klass.IDENTIFIER_ESCAPES) 522 klass._COMMENTS = { 523 **dict( 524 (comment, None) if isinstance(comment, str) else (comment[0], comment[1]) 525 for comment in klass.COMMENTS 526 ), 527 "{#": "#}", # Ensure Jinja comments are tokenized correctly in all dialects 528 } 529 if klass.HINT_START in klass.KEYWORDS: 530 klass._COMMENTS[klass.HINT_START] = "*/" 531 532 klass._KEYWORD_TRIE = new_trie( 533 key.upper() 534 for key in ( 535 *klass.KEYWORDS, 536 *klass._COMMENTS, 537 *klass._QUOTES, 538 *klass._FORMAT_STRINGS, 539 ) 540 if " " in key or any(single in key for single in klass.SINGLE_TOKENS) 541 ) 542 543 if USE_RS_TOKENIZER: 544 settings = RsTokenizerSettings( 545 white_space={k: _TOKEN_TYPE_TO_INDEX[v] for k, v in klass.WHITE_SPACE.items()}, 546 single_tokens={k: _TOKEN_TYPE_TO_INDEX[v] for k, v in klass.SINGLE_TOKENS.items()}, 547 keywords={k: _TOKEN_TYPE_TO_INDEX[v] for k, v in klass.KEYWORDS.items()}, 548 numeric_literals=klass.NUMERIC_LITERALS, 549 identifiers=klass._IDENTIFIERS, 550 identifier_escapes=klass._IDENTIFIER_ESCAPES, 551 string_escapes=klass._STRING_ESCAPES, 552 quotes=klass._QUOTES, 553 format_strings={ 554 k: (v1, _TOKEN_TYPE_TO_INDEX[v2]) 555 for k, (v1, v2) in klass._FORMAT_STRINGS.items() 556 }, 557 has_bit_strings=bool(klass.BIT_STRINGS), 558 has_hex_strings=bool(klass.HEX_STRINGS), 559 comments=klass._COMMENTS, 560 var_single_tokens=klass.VAR_SINGLE_TOKENS, 561 commands={_TOKEN_TYPE_TO_INDEX[v] for v in klass.COMMANDS}, 562 command_prefix_tokens={ 563 _TOKEN_TYPE_TO_INDEX[v] for v in klass.COMMAND_PREFIX_TOKENS 564 }, 565 heredoc_tag_is_identifier=klass.HEREDOC_TAG_IS_IDENTIFIER, 566 string_escapes_allowed_in_raw_strings=klass.STRING_ESCAPES_ALLOWED_IN_RAW_STRINGS, 567 nested_comments=klass.NESTED_COMMENTS, 568 hint_start=klass.HINT_START, 569 tokens_preceding_hint={ 570 _TOKEN_TYPE_TO_INDEX[v] for v in klass.TOKENS_PRECEDING_HINT 571 }, 572 ) 573 token_types = RsTokenTypeSettings( 574 bit_string=_TOKEN_TYPE_TO_INDEX[TokenType.BIT_STRING], 575 break_=_TOKEN_TYPE_TO_INDEX[TokenType.BREAK], 576 dcolon=_TOKEN_TYPE_TO_INDEX[TokenType.DCOLON], 577 heredoc_string=_TOKEN_TYPE_TO_INDEX[TokenType.HEREDOC_STRING], 578 raw_string=_TOKEN_TYPE_TO_INDEX[TokenType.RAW_STRING], 579 hex_string=_TOKEN_TYPE_TO_INDEX[TokenType.HEX_STRING], 580 identifier=_TOKEN_TYPE_TO_INDEX[TokenType.IDENTIFIER], 581 number=_TOKEN_TYPE_TO_INDEX[TokenType.NUMBER], 582 parameter=_TOKEN_TYPE_TO_INDEX[TokenType.PARAMETER], 583 semicolon=_TOKEN_TYPE_TO_INDEX[TokenType.SEMICOLON], 584 string=_TOKEN_TYPE_TO_INDEX[TokenType.STRING], 585 var=_TOKEN_TYPE_TO_INDEX[TokenType.VAR], 586 heredoc_string_alternative=_TOKEN_TYPE_TO_INDEX[klass.HEREDOC_STRING_ALTERNATIVE], 587 hint=_TOKEN_TYPE_TO_INDEX[TokenType.HINT], 588 ) 589 klass._RS_TOKENIZER = RsTokenizer(settings, token_types) 590 else: 591 klass._RS_TOKENIZER = None 592 593 return klass 594 595 596class Tokenizer(metaclass=_Tokenizer): 597 SINGLE_TOKENS = { 598 "(": TokenType.L_PAREN, 599 ")": TokenType.R_PAREN, 600 "[": TokenType.L_BRACKET, 601 "]": TokenType.R_BRACKET, 602 "{": TokenType.L_BRACE, 603 "}": TokenType.R_BRACE, 604 "&": TokenType.AMP, 605 "^": TokenType.CARET, 606 ":": TokenType.COLON, 607 ",": TokenType.COMMA, 608 ".": TokenType.DOT, 609 "-": TokenType.DASH, 610 "=": TokenType.EQ, 611 ">": TokenType.GT, 612 "<": TokenType.LT, 613 "%": TokenType.MOD, 614 "!": TokenType.NOT, 615 "|": TokenType.PIPE, 616 "+": TokenType.PLUS, 617 ";": TokenType.SEMICOLON, 618 "/": TokenType.SLASH, 619 "\\": TokenType.BACKSLASH, 620 "*": TokenType.STAR, 621 "~": TokenType.TILDA, 622 "?": TokenType.PLACEHOLDER, 623 "@": TokenType.PARAMETER, 624 "#": TokenType.HASH, 625 # Used for breaking a var like x'y' but nothing else the token type doesn't matter 626 "'": TokenType.UNKNOWN, 627 "`": TokenType.UNKNOWN, 628 '"': TokenType.UNKNOWN, 629 } 630 631 BIT_STRINGS: t.List[str | t.Tuple[str, str]] = [] 632 BYTE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 633 HEX_STRINGS: t.List[str | t.Tuple[str, str]] = [] 634 RAW_STRINGS: t.List[str | t.Tuple[str, str]] = [] 635 HEREDOC_STRINGS: t.List[str | t.Tuple[str, str]] = [] 636 UNICODE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 637 IDENTIFIERS: t.List[str | t.Tuple[str, str]] = ['"'] 638 QUOTES: t.List[t.Tuple[str, str] | str] = ["'"] 639 STRING_ESCAPES = ["'"] 640 VAR_SINGLE_TOKENS: t.Set[str] = set() 641 642 # The strings in this list can always be used as escapes, regardless of the surrounding 643 # identifier delimiters. By default, the closing delimiter is assumed to also act as an 644 # identifier escape, e.g. if we use double-quotes, then they also act as escapes: "x""" 645 IDENTIFIER_ESCAPES: t.List[str] = [] 646 647 # Whether the heredoc tags follow the same lexical rules as unquoted identifiers 648 HEREDOC_TAG_IS_IDENTIFIER = False 649 650 # Token that we'll generate as a fallback if the heredoc prefix doesn't correspond to a heredoc 651 HEREDOC_STRING_ALTERNATIVE = TokenType.VAR 652 653 # Whether string escape characters function as such when placed within raw strings 654 STRING_ESCAPES_ALLOWED_IN_RAW_STRINGS = True 655 656 NESTED_COMMENTS = True 657 658 HINT_START = "/*+" 659 660 TOKENS_PRECEDING_HINT = {TokenType.SELECT, TokenType.INSERT, TokenType.UPDATE, TokenType.DELETE} 661 662 # Autofilled 663 _COMMENTS: t.Dict[str, str] = {} 664 _FORMAT_STRINGS: t.Dict[str, t.Tuple[str, TokenType]] = {} 665 _IDENTIFIERS: t.Dict[str, str] = {} 666 _IDENTIFIER_ESCAPES: t.Set[str] = set() 667 _QUOTES: t.Dict[str, str] = {} 668 _STRING_ESCAPES: t.Set[str] = set() 669 _KEYWORD_TRIE: t.Dict = {} 670 _RS_TOKENIZER: t.Optional[t.Any] = None 671 672 KEYWORDS: t.Dict[str, TokenType] = { 673 **{f"{{%{postfix}": TokenType.BLOCK_START for postfix in ("", "+", "-")}, 674 **{f"{prefix}%}}": TokenType.BLOCK_END for prefix in ("", "+", "-")}, 675 **{f"{{{{{postfix}": TokenType.BLOCK_START for postfix in ("+", "-")}, 676 **{f"{prefix}}}}}": TokenType.BLOCK_END for prefix in ("+", "-")}, 677 HINT_START: TokenType.HINT, 678 "==": TokenType.EQ, 679 "::": TokenType.DCOLON, 680 "||": TokenType.DPIPE, 681 ">=": TokenType.GTE, 682 "<=": TokenType.LTE, 683 "<>": TokenType.NEQ, 684 "!=": TokenType.NEQ, 685 ":=": TokenType.COLON_EQ, 686 "<=>": TokenType.NULLSAFE_EQ, 687 "->": TokenType.ARROW, 688 "->>": TokenType.DARROW, 689 "=>": TokenType.FARROW, 690 "#>": TokenType.HASH_ARROW, 691 "#>>": TokenType.DHASH_ARROW, 692 "<->": TokenType.LR_ARROW, 693 "&&": TokenType.DAMP, 694 "??": TokenType.DQMARK, 695 "~~~": TokenType.GLOB, 696 "~~": TokenType.LIKE, 697 "~~*": TokenType.ILIKE, 698 "~*": TokenType.IRLIKE, 699 "ALL": TokenType.ALL, 700 "ALWAYS": TokenType.ALWAYS, 701 "AND": TokenType.AND, 702 "ANTI": TokenType.ANTI, 703 "ANY": TokenType.ANY, 704 "ASC": TokenType.ASC, 705 "AS": TokenType.ALIAS, 706 "ASOF": TokenType.ASOF, 707 "AUTOINCREMENT": TokenType.AUTO_INCREMENT, 708 "AUTO_INCREMENT": TokenType.AUTO_INCREMENT, 709 "BEGIN": TokenType.BEGIN, 710 "BETWEEN": TokenType.BETWEEN, 711 "CACHE": TokenType.CACHE, 712 "UNCACHE": TokenType.UNCACHE, 713 "CASE": TokenType.CASE, 714 "CHARACTER SET": TokenType.CHARACTER_SET, 715 "CLUSTER BY": TokenType.CLUSTER_BY, 716 "COLLATE": TokenType.COLLATE, 717 "COLUMN": TokenType.COLUMN, 718 "COMMIT": TokenType.COMMIT, 719 "CONNECT BY": TokenType.CONNECT_BY, 720 "CONSTRAINT": TokenType.CONSTRAINT, 721 "COPY": TokenType.COPY, 722 "CREATE": TokenType.CREATE, 723 "CROSS": TokenType.CROSS, 724 "CUBE": TokenType.CUBE, 725 "CURRENT_DATE": TokenType.CURRENT_DATE, 726 "CURRENT_SCHEMA": TokenType.CURRENT_SCHEMA, 727 "CURRENT_TIME": TokenType.CURRENT_TIME, 728 "CURRENT_TIMESTAMP": TokenType.CURRENT_TIMESTAMP, 729 "CURRENT_USER": TokenType.CURRENT_USER, 730 "DATABASE": TokenType.DATABASE, 731 "DEFAULT": TokenType.DEFAULT, 732 "DELETE": TokenType.DELETE, 733 "DESC": TokenType.DESC, 734 "DESCRIBE": TokenType.DESCRIBE, 735 "DISTINCT": TokenType.DISTINCT, 736 "DISTRIBUTE BY": TokenType.DISTRIBUTE_BY, 737 "DIV": TokenType.DIV, 738 "DROP": TokenType.DROP, 739 "ELSE": TokenType.ELSE, 740 "END": TokenType.END, 741 "ENUM": TokenType.ENUM, 742 "ESCAPE": TokenType.ESCAPE, 743 "EXCEPT": TokenType.EXCEPT, 744 "EXECUTE": TokenType.EXECUTE, 745 "EXISTS": TokenType.EXISTS, 746 "FALSE": TokenType.FALSE, 747 "FETCH": TokenType.FETCH, 748 "FILTER": TokenType.FILTER, 749 "FIRST": TokenType.FIRST, 750 "FULL": TokenType.FULL, 751 "FUNCTION": TokenType.FUNCTION, 752 "FOR": TokenType.FOR, 753 "FOREIGN KEY": TokenType.FOREIGN_KEY, 754 "FORMAT": TokenType.FORMAT, 755 "FROM": TokenType.FROM, 756 "GEOGRAPHY": TokenType.GEOGRAPHY, 757 "GEOMETRY": TokenType.GEOMETRY, 758 "GLOB": TokenType.GLOB, 759 "GROUP BY": TokenType.GROUP_BY, 760 "GROUPING SETS": TokenType.GROUPING_SETS, 761 "HAVING": TokenType.HAVING, 762 "ILIKE": TokenType.ILIKE, 763 "IN": TokenType.IN, 764 "INDEX": TokenType.INDEX, 765 "INET": TokenType.INET, 766 "INNER": TokenType.INNER, 767 "INSERT": TokenType.INSERT, 768 "INTERVAL": TokenType.INTERVAL, 769 "INTERSECT": TokenType.INTERSECT, 770 "INTO": TokenType.INTO, 771 "IS": TokenType.IS, 772 "ISNULL": TokenType.ISNULL, 773 "JOIN": TokenType.JOIN, 774 "KEEP": TokenType.KEEP, 775 "KILL": TokenType.KILL, 776 "LATERAL": TokenType.LATERAL, 777 "LEFT": TokenType.LEFT, 778 "LIKE": TokenType.LIKE, 779 "LIMIT": TokenType.LIMIT, 780 "LOAD": TokenType.LOAD, 781 "LOCK": TokenType.LOCK, 782 "MERGE": TokenType.MERGE, 783 "NAMESPACE": TokenType.NAMESPACE, 784 "NATURAL": TokenType.NATURAL, 785 "NEXT": TokenType.NEXT, 786 "NOT": TokenType.NOT, 787 "NOTNULL": TokenType.NOTNULL, 788 "NULL": TokenType.NULL, 789 "OBJECT": TokenType.OBJECT, 790 "OFFSET": TokenType.OFFSET, 791 "ON": TokenType.ON, 792 "OR": TokenType.OR, 793 "XOR": TokenType.XOR, 794 "ORDER BY": TokenType.ORDER_BY, 795 "ORDINALITY": TokenType.ORDINALITY, 796 "OUTER": TokenType.OUTER, 797 "OVER": TokenType.OVER, 798 "OVERLAPS": TokenType.OVERLAPS, 799 "OVERWRITE": TokenType.OVERWRITE, 800 "PARTITION": TokenType.PARTITION, 801 "PARTITION BY": TokenType.PARTITION_BY, 802 "PARTITIONED BY": TokenType.PARTITION_BY, 803 "PARTITIONED_BY": TokenType.PARTITION_BY, 804 "PERCENT": TokenType.PERCENT, 805 "PIVOT": TokenType.PIVOT, 806 "PRAGMA": TokenType.PRAGMA, 807 "PRIMARY KEY": TokenType.PRIMARY_KEY, 808 "PROCEDURE": TokenType.PROCEDURE, 809 "QUALIFY": TokenType.QUALIFY, 810 "RANGE": TokenType.RANGE, 811 "RECURSIVE": TokenType.RECURSIVE, 812 "REGEXP": TokenType.RLIKE, 813 "RENAME": TokenType.RENAME, 814 "REPLACE": TokenType.REPLACE, 815 "RETURNING": TokenType.RETURNING, 816 "REFERENCES": TokenType.REFERENCES, 817 "RIGHT": TokenType.RIGHT, 818 "RLIKE": TokenType.RLIKE, 819 "ROLLBACK": TokenType.ROLLBACK, 820 "ROLLUP": TokenType.ROLLUP, 821 "ROW": TokenType.ROW, 822 "ROWS": TokenType.ROWS, 823 "SCHEMA": TokenType.SCHEMA, 824 "SELECT": TokenType.SELECT, 825 "SEMI": TokenType.SEMI, 826 "SET": TokenType.SET, 827 "SETTINGS": TokenType.SETTINGS, 828 "SHOW": TokenType.SHOW, 829 "SIMILAR TO": TokenType.SIMILAR_TO, 830 "SOME": TokenType.SOME, 831 "SORT BY": TokenType.SORT_BY, 832 "START WITH": TokenType.START_WITH, 833 "STRAIGHT_JOIN": TokenType.STRAIGHT_JOIN, 834 "TABLE": TokenType.TABLE, 835 "TABLESAMPLE": TokenType.TABLE_SAMPLE, 836 "TEMP": TokenType.TEMPORARY, 837 "TEMPORARY": TokenType.TEMPORARY, 838 "THEN": TokenType.THEN, 839 "TRUE": TokenType.TRUE, 840 "TRUNCATE": TokenType.TRUNCATE, 841 "UNION": TokenType.UNION, 842 "UNKNOWN": TokenType.UNKNOWN, 843 "UNNEST": TokenType.UNNEST, 844 "UNPIVOT": TokenType.UNPIVOT, 845 "UPDATE": TokenType.UPDATE, 846 "USE": TokenType.USE, 847 "USING": TokenType.USING, 848 "UUID": TokenType.UUID, 849 "VALUES": TokenType.VALUES, 850 "VIEW": TokenType.VIEW, 851 "VOLATILE": TokenType.VOLATILE, 852 "WHEN": TokenType.WHEN, 853 "WHERE": TokenType.WHERE, 854 "WINDOW": TokenType.WINDOW, 855 "WITH": TokenType.WITH, 856 "APPLY": TokenType.APPLY, 857 "ARRAY": TokenType.ARRAY, 858 "BIT": TokenType.BIT, 859 "BOOL": TokenType.BOOLEAN, 860 "BOOLEAN": TokenType.BOOLEAN, 861 "BYTE": TokenType.TINYINT, 862 "MEDIUMINT": TokenType.MEDIUMINT, 863 "INT1": TokenType.TINYINT, 864 "TINYINT": TokenType.TINYINT, 865 "INT16": TokenType.SMALLINT, 866 "SHORT": TokenType.SMALLINT, 867 "SMALLINT": TokenType.SMALLINT, 868 "HUGEINT": TokenType.INT128, 869 "UHUGEINT": TokenType.UINT128, 870 "INT2": TokenType.SMALLINT, 871 "INTEGER": TokenType.INT, 872 "INT": TokenType.INT, 873 "INT4": TokenType.INT, 874 "INT32": TokenType.INT, 875 "INT64": TokenType.BIGINT, 876 "INT128": TokenType.INT128, 877 "INT256": TokenType.INT256, 878 "LONG": TokenType.BIGINT, 879 "BIGINT": TokenType.BIGINT, 880 "INT8": TokenType.TINYINT, 881 "UINT": TokenType.UINT, 882 "UINT128": TokenType.UINT128, 883 "UINT256": TokenType.UINT256, 884 "DEC": TokenType.DECIMAL, 885 "DECIMAL": TokenType.DECIMAL, 886 "DECIMAL32": TokenType.DECIMAL32, 887 "DECIMAL64": TokenType.DECIMAL64, 888 "DECIMAL128": TokenType.DECIMAL128, 889 "DECIMAL256": TokenType.DECIMAL256, 890 "BIGDECIMAL": TokenType.BIGDECIMAL, 891 "BIGNUMERIC": TokenType.BIGDECIMAL, 892 "LIST": TokenType.LIST, 893 "MAP": TokenType.MAP, 894 "NULLABLE": TokenType.NULLABLE, 895 "NUMBER": TokenType.DECIMAL, 896 "NUMERIC": TokenType.DECIMAL, 897 "FIXED": TokenType.DECIMAL, 898 "REAL": TokenType.FLOAT, 899 "FLOAT": TokenType.FLOAT, 900 "FLOAT4": TokenType.FLOAT, 901 "FLOAT8": TokenType.DOUBLE, 902 "DOUBLE": TokenType.DOUBLE, 903 "DOUBLE PRECISION": TokenType.DOUBLE, 904 "JSON": TokenType.JSON, 905 "JSONB": TokenType.JSONB, 906 "CHAR": TokenType.CHAR, 907 "CHARACTER": TokenType.CHAR, 908 "NCHAR": TokenType.NCHAR, 909 "VARCHAR": TokenType.VARCHAR, 910 "VARCHAR2": TokenType.VARCHAR, 911 "NVARCHAR": TokenType.NVARCHAR, 912 "NVARCHAR2": TokenType.NVARCHAR, 913 "BPCHAR": TokenType.BPCHAR, 914 "STR": TokenType.TEXT, 915 "STRING": TokenType.TEXT, 916 "TEXT": TokenType.TEXT, 917 "LONGTEXT": TokenType.LONGTEXT, 918 "MEDIUMTEXT": TokenType.MEDIUMTEXT, 919 "TINYTEXT": TokenType.TINYTEXT, 920 "CLOB": TokenType.TEXT, 921 "LONGVARCHAR": TokenType.TEXT, 922 "BINARY": TokenType.BINARY, 923 "BLOB": TokenType.VARBINARY, 924 "LONGBLOB": TokenType.LONGBLOB, 925 "MEDIUMBLOB": TokenType.MEDIUMBLOB, 926 "TINYBLOB": TokenType.TINYBLOB, 927 "BYTEA": TokenType.VARBINARY, 928 "VARBINARY": TokenType.VARBINARY, 929 "TIME": TokenType.TIME, 930 "TIMETZ": TokenType.TIMETZ, 931 "TIMESTAMP": TokenType.TIMESTAMP, 932 "TIMESTAMPTZ": TokenType.TIMESTAMPTZ, 933 "TIMESTAMPLTZ": TokenType.TIMESTAMPLTZ, 934 "TIMESTAMP_LTZ": TokenType.TIMESTAMPLTZ, 935 "TIMESTAMPNTZ": TokenType.TIMESTAMPNTZ, 936 "TIMESTAMP_NTZ": TokenType.TIMESTAMPNTZ, 937 "DATE": TokenType.DATE, 938 "DATETIME": TokenType.DATETIME, 939 "INT4RANGE": TokenType.INT4RANGE, 940 "INT4MULTIRANGE": TokenType.INT4MULTIRANGE, 941 "INT8RANGE": TokenType.INT8RANGE, 942 "INT8MULTIRANGE": TokenType.INT8MULTIRANGE, 943 "NUMRANGE": TokenType.NUMRANGE, 944 "NUMMULTIRANGE": TokenType.NUMMULTIRANGE, 945 "TSRANGE": TokenType.TSRANGE, 946 "TSMULTIRANGE": TokenType.TSMULTIRANGE, 947 "TSTZRANGE": TokenType.TSTZRANGE, 948 "TSTZMULTIRANGE": TokenType.TSTZMULTIRANGE, 949 "DATERANGE": TokenType.DATERANGE, 950 "DATEMULTIRANGE": TokenType.DATEMULTIRANGE, 951 "UNIQUE": TokenType.UNIQUE, 952 "VECTOR": TokenType.VECTOR, 953 "STRUCT": TokenType.STRUCT, 954 "SEQUENCE": TokenType.SEQUENCE, 955 "VARIANT": TokenType.VARIANT, 956 "ALTER": TokenType.ALTER, 957 "ANALYZE": TokenType.ANALYZE, 958 "CALL": TokenType.COMMAND, 959 "COMMENT": TokenType.COMMENT, 960 "EXPLAIN": TokenType.COMMAND, 961 "GRANT": TokenType.GRANT, 962 "OPTIMIZE": TokenType.COMMAND, 963 "PREPARE": TokenType.COMMAND, 964 "VACUUM": TokenType.COMMAND, 965 "USER-DEFINED": TokenType.USERDEFINED, 966 "FOR VERSION": TokenType.VERSION_SNAPSHOT, 967 "FOR TIMESTAMP": TokenType.TIMESTAMP_SNAPSHOT, 968 } 969 970 WHITE_SPACE: t.Dict[t.Optional[str], TokenType] = { 971 " ": TokenType.SPACE, 972 "\t": TokenType.SPACE, 973 "\n": TokenType.BREAK, 974 "\r": TokenType.BREAK, 975 } 976 977 COMMANDS = { 978 TokenType.COMMAND, 979 TokenType.EXECUTE, 980 TokenType.FETCH, 981 TokenType.SHOW, 982 TokenType.RENAME, 983 } 984 985 COMMAND_PREFIX_TOKENS = {TokenType.SEMICOLON, TokenType.BEGIN} 986 987 # Handle numeric literals like in hive (3L = BIGINT) 988 NUMERIC_LITERALS: t.Dict[str, str] = {} 989 990 COMMENTS = ["--", ("/*", "*/")] 991 992 __slots__ = ( 993 "sql", 994 "size", 995 "tokens", 996 "dialect", 997 "use_rs_tokenizer", 998 "_start", 999 "_current", 1000 "_line", 1001 "_col", 1002 "_comments", 1003 "_char", 1004 "_end", 1005 "_peek", 1006 "_prev_token_line", 1007 "_rs_dialect_settings", 1008 ) 1009 1010 def __init__( 1011 self, dialect: DialectType = None, use_rs_tokenizer: t.Optional[bool] = None 1012 ) -> None: 1013 from sqlglot.dialects import Dialect 1014 1015 self.dialect = Dialect.get_or_raise(dialect) 1016 1017 # initialize `use_rs_tokenizer`, and allow it to be overwritten per Tokenizer instance 1018 self.use_rs_tokenizer = ( 1019 use_rs_tokenizer if use_rs_tokenizer is not None else USE_RS_TOKENIZER 1020 ) 1021 1022 if self.use_rs_tokenizer: 1023 self._rs_dialect_settings = RsTokenizerDialectSettings( 1024 unescaped_sequences=self.dialect.UNESCAPED_SEQUENCES, 1025 identifiers_can_start_with_digit=self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT, 1026 numbers_can_be_underscore_separated=self.dialect.NUMBERS_CAN_BE_UNDERSCORE_SEPARATED, 1027 ) 1028 1029 self.reset() 1030 1031 def reset(self) -> None: 1032 self.sql = "" 1033 self.size = 0 1034 self.tokens: t.List[Token] = [] 1035 self._start = 0 1036 self._current = 0 1037 self._line = 1 1038 self._col = 0 1039 self._comments: t.List[str] = [] 1040 1041 self._char = "" 1042 self._end = False 1043 self._peek = "" 1044 self._prev_token_line = -1 1045 1046 def tokenize(self, sql: str) -> t.List[Token]: 1047 """Returns a list of tokens corresponding to the SQL string `sql`.""" 1048 if self.use_rs_tokenizer: 1049 return self.tokenize_rs(sql) 1050 1051 self.reset() 1052 self.sql = sql 1053 self.size = len(sql) 1054 1055 try: 1056 self._scan() 1057 except Exception as e: 1058 start = max(self._current - 50, 0) 1059 end = min(self._current + 50, self.size - 1) 1060 context = self.sql[start:end] 1061 raise TokenError(f"Error tokenizing '{context}'") from e 1062 1063 return self.tokens 1064 1065 def _scan(self, until: t.Optional[t.Callable] = None) -> None: 1066 while self.size and not self._end: 1067 current = self._current 1068 1069 # Skip spaces here rather than iteratively calling advance() for performance reasons 1070 while current < self.size: 1071 char = self.sql[current] 1072 1073 if char.isspace() and (char == " " or char == "\t"): 1074 current += 1 1075 else: 1076 break 1077 1078 offset = current - self._current if current > self._current else 1 1079 1080 self._start = current 1081 self._advance(offset) 1082 1083 if not self._char.isspace(): 1084 if self._char.isdigit(): 1085 self._scan_number() 1086 elif self._char in self._IDENTIFIERS: 1087 self._scan_identifier(self._IDENTIFIERS[self._char]) 1088 else: 1089 self._scan_keywords() 1090 1091 if until and until(): 1092 break 1093 1094 if self.tokens and self._comments: 1095 self.tokens[-1].comments.extend(self._comments) 1096 1097 def _chars(self, size: int) -> str: 1098 if size == 1: 1099 return self._char 1100 1101 start = self._current - 1 1102 end = start + size 1103 1104 return self.sql[start:end] if end <= self.size else "" 1105 1106 def _advance(self, i: int = 1, alnum: bool = False) -> None: 1107 if self.WHITE_SPACE.get(self._char) is TokenType.BREAK: 1108 # Ensures we don't count an extra line if we get a \r\n line break sequence 1109 if not (self._char == "\r" and self._peek == "\n"): 1110 self._col = 1 1111 self._line += 1 1112 else: 1113 self._col += i 1114 1115 self._current += i 1116 self._end = self._current >= self.size 1117 self._char = self.sql[self._current - 1] 1118 self._peek = "" if self._end else self.sql[self._current] 1119 1120 if alnum and self._char.isalnum(): 1121 # Here we use local variables instead of attributes for better performance 1122 _col = self._col 1123 _current = self._current 1124 _end = self._end 1125 _peek = self._peek 1126 1127 while _peek.isalnum(): 1128 _col += 1 1129 _current += 1 1130 _end = _current >= self.size 1131 _peek = "" if _end else self.sql[_current] 1132 1133 self._col = _col 1134 self._current = _current 1135 self._end = _end 1136 self._peek = _peek 1137 self._char = self.sql[_current - 1] 1138 1139 @property 1140 def _text(self) -> str: 1141 return self.sql[self._start : self._current] 1142 1143 def _add(self, token_type: TokenType, text: t.Optional[str] = None) -> None: 1144 self._prev_token_line = self._line 1145 1146 if self._comments and token_type == TokenType.SEMICOLON and self.tokens: 1147 self.tokens[-1].comments.extend(self._comments) 1148 self._comments = [] 1149 1150 self.tokens.append( 1151 Token( 1152 token_type, 1153 text=self._text if text is None else text, 1154 line=self._line, 1155 col=self._col, 1156 start=self._start, 1157 end=self._current - 1, 1158 comments=self._comments, 1159 ) 1160 ) 1161 self._comments = [] 1162 1163 # If we have either a semicolon or a begin token before the command's token, we'll parse 1164 # whatever follows the command's token as a string 1165 if ( 1166 token_type in self.COMMANDS 1167 and self._peek != ";" 1168 and (len(self.tokens) == 1 or self.tokens[-2].token_type in self.COMMAND_PREFIX_TOKENS) 1169 ): 1170 start = self._current 1171 tokens = len(self.tokens) 1172 self._scan(lambda: self._peek == ";") 1173 self.tokens = self.tokens[:tokens] 1174 text = self.sql[start : self._current].strip() 1175 if text: 1176 self._add(TokenType.STRING, text) 1177 1178 def _scan_keywords(self) -> None: 1179 size = 0 1180 word = None 1181 chars = self._text 1182 char = chars 1183 prev_space = False 1184 skip = False 1185 trie = self._KEYWORD_TRIE 1186 single_token = char in self.SINGLE_TOKENS 1187 1188 while chars: 1189 if skip: 1190 result = TrieResult.PREFIX 1191 else: 1192 result, trie = in_trie(trie, char.upper()) 1193 1194 if result == TrieResult.FAILED: 1195 break 1196 if result == TrieResult.EXISTS: 1197 word = chars 1198 1199 end = self._current + size 1200 size += 1 1201 1202 if end < self.size: 1203 char = self.sql[end] 1204 single_token = single_token or char in self.SINGLE_TOKENS 1205 is_space = char.isspace() 1206 1207 if not is_space or not prev_space: 1208 if is_space: 1209 char = " " 1210 chars += char 1211 prev_space = is_space 1212 skip = False 1213 else: 1214 skip = True 1215 else: 1216 char = "" 1217 break 1218 1219 if word: 1220 if self._scan_string(word): 1221 return 1222 if self._scan_comment(word): 1223 return 1224 if prev_space or single_token or not char: 1225 self._advance(size - 1) 1226 word = word.upper() 1227 self._add(self.KEYWORDS[word], text=word) 1228 return 1229 1230 if self._char in self.SINGLE_TOKENS: 1231 self._add(self.SINGLE_TOKENS[self._char], text=self._char) 1232 return 1233 1234 self._scan_var() 1235 1236 def _scan_comment(self, comment_start: str) -> bool: 1237 if comment_start not in self._COMMENTS: 1238 return False 1239 1240 comment_start_line = self._line 1241 comment_start_size = len(comment_start) 1242 comment_end = self._COMMENTS[comment_start] 1243 1244 if comment_end: 1245 # Skip the comment's start delimiter 1246 self._advance(comment_start_size) 1247 1248 comment_count = 1 1249 comment_end_size = len(comment_end) 1250 1251 while not self._end: 1252 if self._chars(comment_end_size) == comment_end: 1253 comment_count -= 1 1254 if not comment_count: 1255 break 1256 1257 self._advance(alnum=True) 1258 1259 # Nested comments are allowed by some dialects, e.g. databricks, duckdb, postgres 1260 if ( 1261 self.NESTED_COMMENTS 1262 and not self._end 1263 and self._chars(comment_end_size) == comment_start 1264 ): 1265 self._advance(comment_start_size) 1266 comment_count += 1 1267 1268 self._comments.append(self._text[comment_start_size : -comment_end_size + 1]) 1269 self._advance(comment_end_size - 1) 1270 else: 1271 while not self._end and self.WHITE_SPACE.get(self._peek) is not TokenType.BREAK: 1272 self._advance(alnum=True) 1273 self._comments.append(self._text[comment_start_size:]) 1274 1275 if ( 1276 comment_start == self.HINT_START 1277 and self.tokens 1278 and self.tokens[-1].token_type in self.TOKENS_PRECEDING_HINT 1279 ): 1280 self._add(TokenType.HINT) 1281 1282 # Leading comment is attached to the succeeding token, whilst trailing comment to the preceding. 1283 # Multiple consecutive comments are preserved by appending them to the current comments list. 1284 if comment_start_line == self._prev_token_line: 1285 self.tokens[-1].comments.extend(self._comments) 1286 self._comments = [] 1287 self._prev_token_line = self._line 1288 1289 return True 1290 1291 def _scan_number(self) -> None: 1292 if self._char == "0": 1293 peek = self._peek.upper() 1294 if peek == "B": 1295 return self._scan_bits() if self.BIT_STRINGS else self._add(TokenType.NUMBER) 1296 elif peek == "X": 1297 return self._scan_hex() if self.HEX_STRINGS else self._add(TokenType.NUMBER) 1298 1299 decimal = False 1300 scientific = 0 1301 1302 while True: 1303 if self._peek.isdigit(): 1304 self._advance() 1305 elif self._peek == "." and not decimal: 1306 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER: 1307 return self._add(TokenType.NUMBER) 1308 decimal = True 1309 self._advance() 1310 elif self._peek in ("-", "+") and scientific == 1: 1311 scientific += 1 1312 self._advance() 1313 elif self._peek.upper() == "E" and not scientific: 1314 scientific += 1 1315 self._advance() 1316 elif self._peek.isidentifier(): 1317 number_text = self._text 1318 literal = "" 1319 1320 while self._peek.strip() and self._peek not in self.SINGLE_TOKENS: 1321 literal += self._peek 1322 self._advance() 1323 1324 token_type = self.KEYWORDS.get(self.NUMERIC_LITERALS.get(literal.upper(), "")) 1325 1326 if token_type: 1327 self._add(TokenType.NUMBER, number_text) 1328 self._add(TokenType.DCOLON, "::") 1329 return self._add(token_type, literal) 1330 else: 1331 replaced = literal.replace("_", "") 1332 if self.dialect.NUMBERS_CAN_BE_UNDERSCORE_SEPARATED and replaced.isdigit(): 1333 return self._add(TokenType.NUMBER, number_text + replaced) 1334 if self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT: 1335 return self._add(TokenType.VAR) 1336 1337 self._advance(-len(literal)) 1338 return self._add(TokenType.NUMBER, number_text) 1339 else: 1340 return self._add(TokenType.NUMBER) 1341 1342 def _scan_bits(self) -> None: 1343 self._advance() 1344 value = self._extract_value() 1345 try: 1346 # If `value` can't be converted to a binary, fallback to tokenizing it as an identifier 1347 int(value, 2) 1348 self._add(TokenType.BIT_STRING, value[2:]) # Drop the 0b 1349 except ValueError: 1350 self._add(TokenType.IDENTIFIER) 1351 1352 def _scan_hex(self) -> None: 1353 self._advance() 1354 value = self._extract_value() 1355 try: 1356 # If `value` can't be converted to a hex, fallback to tokenizing it as an identifier 1357 int(value, 16) 1358 self._add(TokenType.HEX_STRING, value[2:]) # Drop the 0x 1359 except ValueError: 1360 self._add(TokenType.IDENTIFIER) 1361 1362 def _extract_value(self) -> str: 1363 while True: 1364 char = self._peek.strip() 1365 if char and char not in self.SINGLE_TOKENS: 1366 self._advance(alnum=True) 1367 else: 1368 break 1369 1370 return self._text 1371 1372 def _scan_string(self, start: str) -> bool: 1373 base = None 1374 token_type = TokenType.STRING 1375 1376 if start in self._QUOTES: 1377 end = self._QUOTES[start] 1378 elif start in self._FORMAT_STRINGS: 1379 end, token_type = self._FORMAT_STRINGS[start] 1380 1381 if token_type == TokenType.HEX_STRING: 1382 base = 16 1383 elif token_type == TokenType.BIT_STRING: 1384 base = 2 1385 elif token_type == TokenType.HEREDOC_STRING: 1386 self._advance() 1387 1388 if self._char == end: 1389 tag = "" 1390 else: 1391 tag = self._extract_string( 1392 end, 1393 raw_string=True, 1394 raise_unmatched=not self.HEREDOC_TAG_IS_IDENTIFIER, 1395 ) 1396 1397 if tag and self.HEREDOC_TAG_IS_IDENTIFIER and (self._end or not tag.isidentifier()): 1398 if not self._end: 1399 self._advance(-1) 1400 1401 self._advance(-len(tag)) 1402 self._add(self.HEREDOC_STRING_ALTERNATIVE) 1403 return True 1404 1405 end = f"{start}{tag}{end}" 1406 else: 1407 return False 1408 1409 self._advance(len(start)) 1410 text = self._extract_string(end, raw_string=token_type == TokenType.RAW_STRING) 1411 1412 if base: 1413 try: 1414 int(text, base) 1415 except Exception: 1416 raise TokenError( 1417 f"Numeric string contains invalid characters from {self._line}:{self._start}" 1418 ) 1419 1420 self._add(token_type, text) 1421 return True 1422 1423 def _scan_identifier(self, identifier_end: str) -> None: 1424 self._advance() 1425 text = self._extract_string( 1426 identifier_end, escapes=self._IDENTIFIER_ESCAPES | {identifier_end} 1427 ) 1428 self._add(TokenType.IDENTIFIER, text) 1429 1430 def _scan_var(self) -> None: 1431 while True: 1432 char = self._peek.strip() 1433 if char and (char in self.VAR_SINGLE_TOKENS or char not in self.SINGLE_TOKENS): 1434 self._advance(alnum=True) 1435 else: 1436 break 1437 1438 self._add( 1439 TokenType.VAR 1440 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER 1441 else self.KEYWORDS.get(self._text.upper(), TokenType.VAR) 1442 ) 1443 1444 def _extract_string( 1445 self, 1446 delimiter: str, 1447 escapes: t.Optional[t.Set[str]] = None, 1448 raw_string: bool = False, 1449 raise_unmatched: bool = True, 1450 ) -> str: 1451 text = "" 1452 delim_size = len(delimiter) 1453 escapes = self._STRING_ESCAPES if escapes is None else escapes 1454 1455 while True: 1456 if ( 1457 not raw_string 1458 and self.dialect.UNESCAPED_SEQUENCES 1459 and self._peek 1460 and self._char in self.STRING_ESCAPES 1461 ): 1462 unescaped_sequence = self.dialect.UNESCAPED_SEQUENCES.get(self._char + self._peek) 1463 if unescaped_sequence: 1464 self._advance(2) 1465 text += unescaped_sequence 1466 continue 1467 if ( 1468 (self.STRING_ESCAPES_ALLOWED_IN_RAW_STRINGS or not raw_string) 1469 and self._char in escapes 1470 and (self._peek == delimiter or self._peek in escapes) 1471 and (self._char not in self._QUOTES or self._char == self._peek) 1472 ): 1473 if self._peek == delimiter: 1474 text += self._peek 1475 else: 1476 text += self._char + self._peek 1477 1478 if self._current + 1 < self.size: 1479 self._advance(2) 1480 else: 1481 raise TokenError(f"Missing {delimiter} from {self._line}:{self._current}") 1482 else: 1483 if self._chars(delim_size) == delimiter: 1484 if delim_size > 1: 1485 self._advance(delim_size - 1) 1486 break 1487 1488 if self._end: 1489 if not raise_unmatched: 1490 return text + self._char 1491 1492 raise TokenError(f"Missing {delimiter} from {self._line}:{self._start}") 1493 1494 current = self._current - 1 1495 self._advance(alnum=True) 1496 text += self.sql[current : self._current - 1] 1497 1498 return text 1499 1500 def tokenize_rs(self, sql: str) -> t.List[Token]: 1501 if not self._RS_TOKENIZER: 1502 raise SqlglotError("Rust tokenizer is not available") 1503 1504 try: 1505 tokens = self._RS_TOKENIZER.tokenize(sql, self._rs_dialect_settings) 1506 for token in tokens: 1507 token.token_type = _ALL_TOKEN_TYPES[token.token_type_index] 1508 return tokens 1509 except Exception as e: 1510 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 DOTCOLON = auto() 42 DCOLON = auto() 43 DQMARK = auto() 44 SEMICOLON = auto() 45 STAR = auto() 46 BACKSLASH = auto() 47 SLASH = auto() 48 LT = auto() 49 LTE = auto() 50 GT = auto() 51 GTE = auto() 52 NOT = auto() 53 EQ = auto() 54 NEQ = auto() 55 NULLSAFE_EQ = auto() 56 COLON_EQ = auto() 57 AND = auto() 58 OR = auto() 59 AMP = auto() 60 DPIPE = auto() 61 PIPE = auto() 62 PIPE_SLASH = auto() 63 DPIPE_SLASH = auto() 64 CARET = auto() 65 CARET_AT = auto() 66 TILDA = auto() 67 ARROW = auto() 68 DARROW = auto() 69 FARROW = auto() 70 HASH = auto() 71 HASH_ARROW = auto() 72 DHASH_ARROW = auto() 73 LR_ARROW = auto() 74 DAT = auto() 75 LT_AT = auto() 76 AT_GT = auto() 77 DOLLAR = auto() 78 PARAMETER = auto() 79 SESSION_PARAMETER = auto() 80 DAMP = auto() 81 XOR = auto() 82 DSTAR = auto() 83 84 URI_START = auto() 85 86 BLOCK_START = auto() 87 BLOCK_END = auto() 88 89 SPACE = auto() 90 BREAK = auto() 91 92 STRING = auto() 93 NUMBER = auto() 94 IDENTIFIER = auto() 95 DATABASE = auto() 96 COLUMN = auto() 97 COLUMN_DEF = auto() 98 SCHEMA = auto() 99 TABLE = auto() 100 WAREHOUSE = auto() 101 STAGE = auto() 102 STREAMLIT = auto() 103 VAR = auto() 104 BIT_STRING = auto() 105 HEX_STRING = auto() 106 BYTE_STRING = auto() 107 NATIONAL_STRING = auto() 108 RAW_STRING = auto() 109 HEREDOC_STRING = auto() 110 UNICODE_STRING = auto() 111 112 # types 113 BIT = auto() 114 BOOLEAN = auto() 115 TINYINT = auto() 116 UTINYINT = auto() 117 SMALLINT = auto() 118 USMALLINT = auto() 119 MEDIUMINT = auto() 120 UMEDIUMINT = auto() 121 INT = auto() 122 UINT = auto() 123 BIGINT = auto() 124 UBIGINT = auto() 125 INT128 = auto() 126 UINT128 = auto() 127 INT256 = auto() 128 UINT256 = auto() 129 FLOAT = auto() 130 DOUBLE = auto() 131 UDOUBLE = auto() 132 DECIMAL = auto() 133 DECIMAL32 = auto() 134 DECIMAL64 = auto() 135 DECIMAL128 = auto() 136 DECIMAL256 = auto() 137 UDECIMAL = auto() 138 BIGDECIMAL = auto() 139 CHAR = auto() 140 NCHAR = auto() 141 VARCHAR = auto() 142 NVARCHAR = auto() 143 BPCHAR = auto() 144 TEXT = auto() 145 MEDIUMTEXT = auto() 146 LONGTEXT = auto() 147 BLOB = auto() 148 MEDIUMBLOB = auto() 149 LONGBLOB = auto() 150 TINYBLOB = auto() 151 TINYTEXT = auto() 152 NAME = auto() 153 BINARY = auto() 154 VARBINARY = auto() 155 JSON = auto() 156 JSONB = auto() 157 TIME = auto() 158 TIMETZ = auto() 159 TIMESTAMP = auto() 160 TIMESTAMPTZ = auto() 161 TIMESTAMPLTZ = auto() 162 TIMESTAMPNTZ = auto() 163 TIMESTAMP_S = auto() 164 TIMESTAMP_MS = auto() 165 TIMESTAMP_NS = auto() 166 DATETIME = auto() 167 DATETIME2 = auto() 168 DATETIME64 = auto() 169 SMALLDATETIME = auto() 170 DATE = auto() 171 DATE32 = auto() 172 INT4RANGE = auto() 173 INT4MULTIRANGE = auto() 174 INT8RANGE = auto() 175 INT8MULTIRANGE = auto() 176 NUMRANGE = auto() 177 NUMMULTIRANGE = auto() 178 TSRANGE = auto() 179 TSMULTIRANGE = auto() 180 TSTZRANGE = auto() 181 TSTZMULTIRANGE = auto() 182 DATERANGE = auto() 183 DATEMULTIRANGE = auto() 184 UUID = auto() 185 GEOGRAPHY = auto() 186 NULLABLE = auto() 187 GEOMETRY = auto() 188 POINT = auto() 189 RING = auto() 190 LINESTRING = auto() 191 MULTILINESTRING = auto() 192 POLYGON = auto() 193 MULTIPOLYGON = auto() 194 HLLSKETCH = auto() 195 HSTORE = auto() 196 SUPER = auto() 197 SERIAL = auto() 198 SMALLSERIAL = auto() 199 BIGSERIAL = auto() 200 XML = auto() 201 YEAR = auto() 202 USERDEFINED = auto() 203 MONEY = auto() 204 SMALLMONEY = auto() 205 ROWVERSION = auto() 206 IMAGE = auto() 207 VARIANT = auto() 208 OBJECT = auto() 209 INET = auto() 210 IPADDRESS = auto() 211 IPPREFIX = auto() 212 IPV4 = auto() 213 IPV6 = auto() 214 ENUM = auto() 215 ENUM8 = auto() 216 ENUM16 = auto() 217 FIXEDSTRING = auto() 218 LOWCARDINALITY = auto() 219 NESTED = auto() 220 AGGREGATEFUNCTION = auto() 221 SIMPLEAGGREGATEFUNCTION = auto() 222 TDIGEST = auto() 223 UNKNOWN = auto() 224 VECTOR = auto() 225 DYNAMIC = auto() 226 227 # keywords 228 ALIAS = auto() 229 ALTER = auto() 230 ALWAYS = auto() 231 ALL = auto() 232 ANTI = auto() 233 ANY = auto() 234 APPLY = auto() 235 ARRAY = auto() 236 ASC = auto() 237 ASOF = auto() 238 ATTACH = auto() 239 AUTO_INCREMENT = auto() 240 BEGIN = auto() 241 BETWEEN = auto() 242 BULK_COLLECT_INTO = auto() 243 CACHE = auto() 244 CASE = auto() 245 CHARACTER_SET = auto() 246 CLUSTER_BY = auto() 247 COLLATE = auto() 248 COMMAND = auto() 249 COMMENT = auto() 250 COMMIT = auto() 251 CONNECT_BY = auto() 252 CONSTRAINT = auto() 253 COPY = auto() 254 CREATE = auto() 255 CROSS = auto() 256 CUBE = auto() 257 CURRENT_DATE = auto() 258 CURRENT_DATETIME = auto() 259 CURRENT_SCHEMA = auto() 260 CURRENT_TIME = auto() 261 CURRENT_TIMESTAMP = auto() 262 CURRENT_USER = auto() 263 DECLARE = auto() 264 DEFAULT = auto() 265 DELETE = auto() 266 DESC = auto() 267 DESCRIBE = auto() 268 DETACH = auto() 269 DICTIONARY = auto() 270 DISTINCT = auto() 271 DISTRIBUTE_BY = auto() 272 DIV = auto() 273 DROP = auto() 274 ELSE = auto() 275 END = auto() 276 ESCAPE = auto() 277 EXCEPT = auto() 278 EXECUTE = auto() 279 EXISTS = auto() 280 FALSE = auto() 281 FETCH = auto() 282 FILE_FORMAT = auto() 283 FILTER = auto() 284 FINAL = auto() 285 FIRST = auto() 286 FOR = auto() 287 FORCE = auto() 288 FOREIGN_KEY = auto() 289 FORMAT = auto() 290 FROM = auto() 291 FULL = auto() 292 FUNCTION = auto() 293 GLOB = auto() 294 GLOBAL = auto() 295 GRANT = auto() 296 GROUP_BY = auto() 297 GROUPING_SETS = auto() 298 HAVING = auto() 299 HINT = auto() 300 IGNORE = auto() 301 ILIKE = auto() 302 ILIKE_ANY = auto() 303 IN = auto() 304 INDEX = auto() 305 INNER = auto() 306 INSERT = auto() 307 INTERSECT = auto() 308 INTERVAL = auto() 309 INTO = auto() 310 INTRODUCER = auto() 311 IRLIKE = auto() 312 IS = auto() 313 ISNULL = auto() 314 JOIN = auto() 315 JOIN_MARKER = auto() 316 KEEP = auto() 317 KEY = auto() 318 KILL = auto() 319 LANGUAGE = auto() 320 LATERAL = auto() 321 LEFT = auto() 322 LIKE = auto() 323 LIKE_ANY = auto() 324 LIMIT = auto() 325 LIST = auto() 326 LOAD = auto() 327 LOCK = auto() 328 MAP = auto() 329 MATCH_CONDITION = auto() 330 MATCH_RECOGNIZE = auto() 331 MEMBER_OF = auto() 332 MERGE = auto() 333 MOD = auto() 334 MODEL = auto() 335 NATURAL = auto() 336 NEXT = auto() 337 NOTHING = auto() 338 NOTNULL = auto() 339 NULL = auto() 340 OBJECT_IDENTIFIER = auto() 341 OFFSET = auto() 342 ON = auto() 343 ONLY = auto() 344 OPERATOR = auto() 345 ORDER_BY = auto() 346 ORDER_SIBLINGS_BY = auto() 347 ORDERED = auto() 348 ORDINALITY = auto() 349 OUTER = auto() 350 OVER = auto() 351 OVERLAPS = auto() 352 OVERWRITE = auto() 353 PARTITION = auto() 354 PARTITION_BY = auto() 355 PERCENT = auto() 356 PIVOT = auto() 357 PLACEHOLDER = auto() 358 POSITIONAL = auto() 359 PRAGMA = auto() 360 PREWHERE = auto() 361 PRIMARY_KEY = auto() 362 PROCEDURE = auto() 363 PROPERTIES = auto() 364 PSEUDO_TYPE = auto() 365 PUT = auto() 366 QUALIFY = auto() 367 QUOTE = auto() 368 RANGE = auto() 369 RECURSIVE = auto() 370 REFRESH = auto() 371 RENAME = auto() 372 REPLACE = auto() 373 RETURNING = auto() 374 REFERENCES = auto() 375 RIGHT = auto() 376 RLIKE = auto() 377 ROLLBACK = auto() 378 ROLLUP = auto() 379 ROW = auto() 380 ROWS = auto() 381 SELECT = auto() 382 SEMI = auto() 383 SEPARATOR = auto() 384 SEQUENCE = auto() 385 SERDE_PROPERTIES = auto() 386 SET = auto() 387 SETTINGS = auto() 388 SHOW = auto() 389 SIMILAR_TO = auto() 390 SOME = auto() 391 SORT_BY = auto() 392 START_WITH = auto() 393 STORAGE_INTEGRATION = auto() 394 STRAIGHT_JOIN = auto() 395 STRUCT = auto() 396 SUMMARIZE = auto() 397 TABLE_SAMPLE = auto() 398 TAG = auto() 399 TEMPORARY = auto() 400 TOP = auto() 401 THEN = auto() 402 TRUE = auto() 403 TRUNCATE = auto() 404 UNCACHE = auto() 405 UNION = auto() 406 UNNEST = auto() 407 UNPIVOT = auto() 408 UPDATE = auto() 409 USE = auto() 410 USING = auto() 411 VALUES = auto() 412 VIEW = auto() 413 VOLATILE = auto() 414 WHEN = auto() 415 WHERE = auto() 416 WINDOW = auto() 417 WITH = auto() 418 UNIQUE = auto() 419 VERSION_SNAPSHOT = auto() 420 TIMESTAMP_SNAPSHOT = auto() 421 OPTION = auto() 422 SINK = auto() 423 SOURCE = auto() 424 ANALYZE = auto() 425 NAMESPACE = auto() 426 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'>
DOTCOLON =
<TokenType.DOTCOLON: 'DOTCOLON'>
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'>
URI_START =
<TokenType.URI_START: 'URI_START'>
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'>
STAGE =
<TokenType.STAGE: 'STAGE'>
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'>
UDOUBLE =
<TokenType.UDOUBLE: 'UDOUBLE'>
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'>
BLOB =
<TokenType.BLOB: 'BLOB'>
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'>
FILE_FORMAT =
<TokenType.FILE_FORMAT: 'FILE_FORMAT'>
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'>
NOTHING =
<TokenType.NOTHING: 'NOTHING'>
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'>
PUT =
<TokenType.PUT: 'PUT'>
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:
433class Token: 434 __slots__ = ("token_type", "text", "line", "col", "start", "end", "comments") 435 436 @classmethod 437 def number(cls, number: int) -> Token: 438 """Returns a NUMBER token with `number` as its text.""" 439 return cls(TokenType.NUMBER, str(number)) 440 441 @classmethod 442 def string(cls, string: str) -> Token: 443 """Returns a STRING token with `string` as its text.""" 444 return cls(TokenType.STRING, string) 445 446 @classmethod 447 def identifier(cls, identifier: str) -> Token: 448 """Returns an IDENTIFIER token with `identifier` as its text.""" 449 return cls(TokenType.IDENTIFIER, identifier) 450 451 @classmethod 452 def var(cls, var: str) -> Token: 453 """Returns an VAR token with `var` as its text.""" 454 return cls(TokenType.VAR, var) 455 456 def __init__( 457 self, 458 token_type: TokenType, 459 text: str, 460 line: int = 1, 461 col: int = 1, 462 start: int = 0, 463 end: int = 0, 464 comments: t.Optional[t.List[str]] = None, 465 ) -> None: 466 """Token initializer. 467 468 Args: 469 token_type: The TokenType Enum. 470 text: The text of the token. 471 line: The line that the token ends on. 472 col: The column that the token ends on. 473 start: The start index of the token. 474 end: The ending index of the token. 475 comments: The comments to attach to the token. 476 """ 477 self.token_type = token_type 478 self.text = text 479 self.line = line 480 self.col = col 481 self.start = start 482 self.end = end 483 self.comments = [] if comments is None else comments 484 485 def __repr__(self) -> str: 486 attributes = ", ".join(f"{k}: {getattr(self, k)}" for k in self.__slots__) 487 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)
456 def __init__( 457 self, 458 token_type: TokenType, 459 text: str, 460 line: int = 1, 461 col: int = 1, 462 start: int = 0, 463 end: int = 0, 464 comments: t.Optional[t.List[str]] = None, 465 ) -> None: 466 """Token initializer. 467 468 Args: 469 token_type: The TokenType Enum. 470 text: The text of the token. 471 line: The line that the token ends on. 472 col: The column that the token ends on. 473 start: The start index of the token. 474 end: The ending index of the token. 475 comments: The comments to attach to the token. 476 """ 477 self.token_type = token_type 478 self.text = text 479 self.line = line 480 self.col = col 481 self.start = start 482 self.end = end 483 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.
436 @classmethod 437 def number(cls, number: int) -> Token: 438 """Returns a NUMBER token with `number` as its text.""" 439 return cls(TokenType.NUMBER, str(number))
Returns a NUMBER token with number
as its text.
441 @classmethod 442 def string(cls, string: str) -> Token: 443 """Returns a STRING token with `string` as its text.""" 444 return cls(TokenType.STRING, string)
Returns a STRING token with string
as its text.
446 @classmethod 447 def identifier(cls, identifier: str) -> Token: 448 """Returns an IDENTIFIER token with `identifier` as its text.""" 449 return cls(TokenType.IDENTIFIER, identifier)
Returns an IDENTIFIER token with identifier
as its text.
class
Tokenizer:
597class Tokenizer(metaclass=_Tokenizer): 598 SINGLE_TOKENS = { 599 "(": TokenType.L_PAREN, 600 ")": TokenType.R_PAREN, 601 "[": TokenType.L_BRACKET, 602 "]": TokenType.R_BRACKET, 603 "{": TokenType.L_BRACE, 604 "}": TokenType.R_BRACE, 605 "&": TokenType.AMP, 606 "^": TokenType.CARET, 607 ":": TokenType.COLON, 608 ",": TokenType.COMMA, 609 ".": TokenType.DOT, 610 "-": TokenType.DASH, 611 "=": TokenType.EQ, 612 ">": TokenType.GT, 613 "<": TokenType.LT, 614 "%": TokenType.MOD, 615 "!": TokenType.NOT, 616 "|": TokenType.PIPE, 617 "+": TokenType.PLUS, 618 ";": TokenType.SEMICOLON, 619 "/": TokenType.SLASH, 620 "\\": TokenType.BACKSLASH, 621 "*": TokenType.STAR, 622 "~": TokenType.TILDA, 623 "?": TokenType.PLACEHOLDER, 624 "@": TokenType.PARAMETER, 625 "#": TokenType.HASH, 626 # Used for breaking a var like x'y' but nothing else the token type doesn't matter 627 "'": TokenType.UNKNOWN, 628 "`": TokenType.UNKNOWN, 629 '"': TokenType.UNKNOWN, 630 } 631 632 BIT_STRINGS: t.List[str | t.Tuple[str, str]] = [] 633 BYTE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 634 HEX_STRINGS: t.List[str | t.Tuple[str, str]] = [] 635 RAW_STRINGS: t.List[str | t.Tuple[str, str]] = [] 636 HEREDOC_STRINGS: t.List[str | t.Tuple[str, str]] = [] 637 UNICODE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 638 IDENTIFIERS: t.List[str | t.Tuple[str, str]] = ['"'] 639 QUOTES: t.List[t.Tuple[str, str] | str] = ["'"] 640 STRING_ESCAPES = ["'"] 641 VAR_SINGLE_TOKENS: t.Set[str] = set() 642 643 # The strings in this list can always be used as escapes, regardless of the surrounding 644 # identifier delimiters. By default, the closing delimiter is assumed to also act as an 645 # identifier escape, e.g. if we use double-quotes, then they also act as escapes: "x""" 646 IDENTIFIER_ESCAPES: t.List[str] = [] 647 648 # Whether the heredoc tags follow the same lexical rules as unquoted identifiers 649 HEREDOC_TAG_IS_IDENTIFIER = False 650 651 # Token that we'll generate as a fallback if the heredoc prefix doesn't correspond to a heredoc 652 HEREDOC_STRING_ALTERNATIVE = TokenType.VAR 653 654 # Whether string escape characters function as such when placed within raw strings 655 STRING_ESCAPES_ALLOWED_IN_RAW_STRINGS = True 656 657 NESTED_COMMENTS = True 658 659 HINT_START = "/*+" 660 661 TOKENS_PRECEDING_HINT = {TokenType.SELECT, TokenType.INSERT, TokenType.UPDATE, TokenType.DELETE} 662 663 # Autofilled 664 _COMMENTS: t.Dict[str, str] = {} 665 _FORMAT_STRINGS: t.Dict[str, t.Tuple[str, TokenType]] = {} 666 _IDENTIFIERS: t.Dict[str, str] = {} 667 _IDENTIFIER_ESCAPES: t.Set[str] = set() 668 _QUOTES: t.Dict[str, str] = {} 669 _STRING_ESCAPES: t.Set[str] = set() 670 _KEYWORD_TRIE: t.Dict = {} 671 _RS_TOKENIZER: t.Optional[t.Any] = None 672 673 KEYWORDS: t.Dict[str, TokenType] = { 674 **{f"{{%{postfix}": TokenType.BLOCK_START for postfix in ("", "+", "-")}, 675 **{f"{prefix}%}}": TokenType.BLOCK_END for prefix in ("", "+", "-")}, 676 **{f"{{{{{postfix}": TokenType.BLOCK_START for postfix in ("+", "-")}, 677 **{f"{prefix}}}}}": TokenType.BLOCK_END for prefix in ("+", "-")}, 678 HINT_START: TokenType.HINT, 679 "==": TokenType.EQ, 680 "::": TokenType.DCOLON, 681 "||": TokenType.DPIPE, 682 ">=": TokenType.GTE, 683 "<=": TokenType.LTE, 684 "<>": TokenType.NEQ, 685 "!=": TokenType.NEQ, 686 ":=": TokenType.COLON_EQ, 687 "<=>": TokenType.NULLSAFE_EQ, 688 "->": TokenType.ARROW, 689 "->>": TokenType.DARROW, 690 "=>": TokenType.FARROW, 691 "#>": TokenType.HASH_ARROW, 692 "#>>": TokenType.DHASH_ARROW, 693 "<->": TokenType.LR_ARROW, 694 "&&": TokenType.DAMP, 695 "??": TokenType.DQMARK, 696 "~~~": TokenType.GLOB, 697 "~~": TokenType.LIKE, 698 "~~*": TokenType.ILIKE, 699 "~*": TokenType.IRLIKE, 700 "ALL": TokenType.ALL, 701 "ALWAYS": TokenType.ALWAYS, 702 "AND": TokenType.AND, 703 "ANTI": TokenType.ANTI, 704 "ANY": TokenType.ANY, 705 "ASC": TokenType.ASC, 706 "AS": TokenType.ALIAS, 707 "ASOF": TokenType.ASOF, 708 "AUTOINCREMENT": TokenType.AUTO_INCREMENT, 709 "AUTO_INCREMENT": TokenType.AUTO_INCREMENT, 710 "BEGIN": TokenType.BEGIN, 711 "BETWEEN": TokenType.BETWEEN, 712 "CACHE": TokenType.CACHE, 713 "UNCACHE": TokenType.UNCACHE, 714 "CASE": TokenType.CASE, 715 "CHARACTER SET": TokenType.CHARACTER_SET, 716 "CLUSTER BY": TokenType.CLUSTER_BY, 717 "COLLATE": TokenType.COLLATE, 718 "COLUMN": TokenType.COLUMN, 719 "COMMIT": TokenType.COMMIT, 720 "CONNECT BY": TokenType.CONNECT_BY, 721 "CONSTRAINT": TokenType.CONSTRAINT, 722 "COPY": TokenType.COPY, 723 "CREATE": TokenType.CREATE, 724 "CROSS": TokenType.CROSS, 725 "CUBE": TokenType.CUBE, 726 "CURRENT_DATE": TokenType.CURRENT_DATE, 727 "CURRENT_SCHEMA": TokenType.CURRENT_SCHEMA, 728 "CURRENT_TIME": TokenType.CURRENT_TIME, 729 "CURRENT_TIMESTAMP": TokenType.CURRENT_TIMESTAMP, 730 "CURRENT_USER": TokenType.CURRENT_USER, 731 "DATABASE": TokenType.DATABASE, 732 "DEFAULT": TokenType.DEFAULT, 733 "DELETE": TokenType.DELETE, 734 "DESC": TokenType.DESC, 735 "DESCRIBE": TokenType.DESCRIBE, 736 "DISTINCT": TokenType.DISTINCT, 737 "DISTRIBUTE BY": TokenType.DISTRIBUTE_BY, 738 "DIV": TokenType.DIV, 739 "DROP": TokenType.DROP, 740 "ELSE": TokenType.ELSE, 741 "END": TokenType.END, 742 "ENUM": TokenType.ENUM, 743 "ESCAPE": TokenType.ESCAPE, 744 "EXCEPT": TokenType.EXCEPT, 745 "EXECUTE": TokenType.EXECUTE, 746 "EXISTS": TokenType.EXISTS, 747 "FALSE": TokenType.FALSE, 748 "FETCH": TokenType.FETCH, 749 "FILTER": TokenType.FILTER, 750 "FIRST": TokenType.FIRST, 751 "FULL": TokenType.FULL, 752 "FUNCTION": TokenType.FUNCTION, 753 "FOR": TokenType.FOR, 754 "FOREIGN KEY": TokenType.FOREIGN_KEY, 755 "FORMAT": TokenType.FORMAT, 756 "FROM": TokenType.FROM, 757 "GEOGRAPHY": TokenType.GEOGRAPHY, 758 "GEOMETRY": TokenType.GEOMETRY, 759 "GLOB": TokenType.GLOB, 760 "GROUP BY": TokenType.GROUP_BY, 761 "GROUPING SETS": TokenType.GROUPING_SETS, 762 "HAVING": TokenType.HAVING, 763 "ILIKE": TokenType.ILIKE, 764 "IN": TokenType.IN, 765 "INDEX": TokenType.INDEX, 766 "INET": TokenType.INET, 767 "INNER": TokenType.INNER, 768 "INSERT": TokenType.INSERT, 769 "INTERVAL": TokenType.INTERVAL, 770 "INTERSECT": TokenType.INTERSECT, 771 "INTO": TokenType.INTO, 772 "IS": TokenType.IS, 773 "ISNULL": TokenType.ISNULL, 774 "JOIN": TokenType.JOIN, 775 "KEEP": TokenType.KEEP, 776 "KILL": TokenType.KILL, 777 "LATERAL": TokenType.LATERAL, 778 "LEFT": TokenType.LEFT, 779 "LIKE": TokenType.LIKE, 780 "LIMIT": TokenType.LIMIT, 781 "LOAD": TokenType.LOAD, 782 "LOCK": TokenType.LOCK, 783 "MERGE": TokenType.MERGE, 784 "NAMESPACE": TokenType.NAMESPACE, 785 "NATURAL": TokenType.NATURAL, 786 "NEXT": TokenType.NEXT, 787 "NOT": TokenType.NOT, 788 "NOTNULL": TokenType.NOTNULL, 789 "NULL": TokenType.NULL, 790 "OBJECT": TokenType.OBJECT, 791 "OFFSET": TokenType.OFFSET, 792 "ON": TokenType.ON, 793 "OR": TokenType.OR, 794 "XOR": TokenType.XOR, 795 "ORDER BY": TokenType.ORDER_BY, 796 "ORDINALITY": TokenType.ORDINALITY, 797 "OUTER": TokenType.OUTER, 798 "OVER": TokenType.OVER, 799 "OVERLAPS": TokenType.OVERLAPS, 800 "OVERWRITE": TokenType.OVERWRITE, 801 "PARTITION": TokenType.PARTITION, 802 "PARTITION BY": TokenType.PARTITION_BY, 803 "PARTITIONED BY": TokenType.PARTITION_BY, 804 "PARTITIONED_BY": TokenType.PARTITION_BY, 805 "PERCENT": TokenType.PERCENT, 806 "PIVOT": TokenType.PIVOT, 807 "PRAGMA": TokenType.PRAGMA, 808 "PRIMARY KEY": TokenType.PRIMARY_KEY, 809 "PROCEDURE": TokenType.PROCEDURE, 810 "QUALIFY": TokenType.QUALIFY, 811 "RANGE": TokenType.RANGE, 812 "RECURSIVE": TokenType.RECURSIVE, 813 "REGEXP": TokenType.RLIKE, 814 "RENAME": TokenType.RENAME, 815 "REPLACE": TokenType.REPLACE, 816 "RETURNING": TokenType.RETURNING, 817 "REFERENCES": TokenType.REFERENCES, 818 "RIGHT": TokenType.RIGHT, 819 "RLIKE": TokenType.RLIKE, 820 "ROLLBACK": TokenType.ROLLBACK, 821 "ROLLUP": TokenType.ROLLUP, 822 "ROW": TokenType.ROW, 823 "ROWS": TokenType.ROWS, 824 "SCHEMA": TokenType.SCHEMA, 825 "SELECT": TokenType.SELECT, 826 "SEMI": TokenType.SEMI, 827 "SET": TokenType.SET, 828 "SETTINGS": TokenType.SETTINGS, 829 "SHOW": TokenType.SHOW, 830 "SIMILAR TO": TokenType.SIMILAR_TO, 831 "SOME": TokenType.SOME, 832 "SORT BY": TokenType.SORT_BY, 833 "START WITH": TokenType.START_WITH, 834 "STRAIGHT_JOIN": TokenType.STRAIGHT_JOIN, 835 "TABLE": TokenType.TABLE, 836 "TABLESAMPLE": TokenType.TABLE_SAMPLE, 837 "TEMP": TokenType.TEMPORARY, 838 "TEMPORARY": TokenType.TEMPORARY, 839 "THEN": TokenType.THEN, 840 "TRUE": TokenType.TRUE, 841 "TRUNCATE": TokenType.TRUNCATE, 842 "UNION": TokenType.UNION, 843 "UNKNOWN": TokenType.UNKNOWN, 844 "UNNEST": TokenType.UNNEST, 845 "UNPIVOT": TokenType.UNPIVOT, 846 "UPDATE": TokenType.UPDATE, 847 "USE": TokenType.USE, 848 "USING": TokenType.USING, 849 "UUID": TokenType.UUID, 850 "VALUES": TokenType.VALUES, 851 "VIEW": TokenType.VIEW, 852 "VOLATILE": TokenType.VOLATILE, 853 "WHEN": TokenType.WHEN, 854 "WHERE": TokenType.WHERE, 855 "WINDOW": TokenType.WINDOW, 856 "WITH": TokenType.WITH, 857 "APPLY": TokenType.APPLY, 858 "ARRAY": TokenType.ARRAY, 859 "BIT": TokenType.BIT, 860 "BOOL": TokenType.BOOLEAN, 861 "BOOLEAN": TokenType.BOOLEAN, 862 "BYTE": TokenType.TINYINT, 863 "MEDIUMINT": TokenType.MEDIUMINT, 864 "INT1": TokenType.TINYINT, 865 "TINYINT": TokenType.TINYINT, 866 "INT16": TokenType.SMALLINT, 867 "SHORT": TokenType.SMALLINT, 868 "SMALLINT": TokenType.SMALLINT, 869 "HUGEINT": TokenType.INT128, 870 "UHUGEINT": TokenType.UINT128, 871 "INT2": TokenType.SMALLINT, 872 "INTEGER": TokenType.INT, 873 "INT": TokenType.INT, 874 "INT4": TokenType.INT, 875 "INT32": TokenType.INT, 876 "INT64": TokenType.BIGINT, 877 "INT128": TokenType.INT128, 878 "INT256": TokenType.INT256, 879 "LONG": TokenType.BIGINT, 880 "BIGINT": TokenType.BIGINT, 881 "INT8": TokenType.TINYINT, 882 "UINT": TokenType.UINT, 883 "UINT128": TokenType.UINT128, 884 "UINT256": TokenType.UINT256, 885 "DEC": TokenType.DECIMAL, 886 "DECIMAL": TokenType.DECIMAL, 887 "DECIMAL32": TokenType.DECIMAL32, 888 "DECIMAL64": TokenType.DECIMAL64, 889 "DECIMAL128": TokenType.DECIMAL128, 890 "DECIMAL256": TokenType.DECIMAL256, 891 "BIGDECIMAL": TokenType.BIGDECIMAL, 892 "BIGNUMERIC": TokenType.BIGDECIMAL, 893 "LIST": TokenType.LIST, 894 "MAP": TokenType.MAP, 895 "NULLABLE": TokenType.NULLABLE, 896 "NUMBER": TokenType.DECIMAL, 897 "NUMERIC": TokenType.DECIMAL, 898 "FIXED": TokenType.DECIMAL, 899 "REAL": TokenType.FLOAT, 900 "FLOAT": TokenType.FLOAT, 901 "FLOAT4": TokenType.FLOAT, 902 "FLOAT8": TokenType.DOUBLE, 903 "DOUBLE": TokenType.DOUBLE, 904 "DOUBLE PRECISION": TokenType.DOUBLE, 905 "JSON": TokenType.JSON, 906 "JSONB": TokenType.JSONB, 907 "CHAR": TokenType.CHAR, 908 "CHARACTER": TokenType.CHAR, 909 "NCHAR": TokenType.NCHAR, 910 "VARCHAR": TokenType.VARCHAR, 911 "VARCHAR2": TokenType.VARCHAR, 912 "NVARCHAR": TokenType.NVARCHAR, 913 "NVARCHAR2": TokenType.NVARCHAR, 914 "BPCHAR": TokenType.BPCHAR, 915 "STR": TokenType.TEXT, 916 "STRING": TokenType.TEXT, 917 "TEXT": TokenType.TEXT, 918 "LONGTEXT": TokenType.LONGTEXT, 919 "MEDIUMTEXT": TokenType.MEDIUMTEXT, 920 "TINYTEXT": TokenType.TINYTEXT, 921 "CLOB": TokenType.TEXT, 922 "LONGVARCHAR": TokenType.TEXT, 923 "BINARY": TokenType.BINARY, 924 "BLOB": TokenType.VARBINARY, 925 "LONGBLOB": TokenType.LONGBLOB, 926 "MEDIUMBLOB": TokenType.MEDIUMBLOB, 927 "TINYBLOB": TokenType.TINYBLOB, 928 "BYTEA": TokenType.VARBINARY, 929 "VARBINARY": TokenType.VARBINARY, 930 "TIME": TokenType.TIME, 931 "TIMETZ": TokenType.TIMETZ, 932 "TIMESTAMP": TokenType.TIMESTAMP, 933 "TIMESTAMPTZ": TokenType.TIMESTAMPTZ, 934 "TIMESTAMPLTZ": TokenType.TIMESTAMPLTZ, 935 "TIMESTAMP_LTZ": TokenType.TIMESTAMPLTZ, 936 "TIMESTAMPNTZ": TokenType.TIMESTAMPNTZ, 937 "TIMESTAMP_NTZ": TokenType.TIMESTAMPNTZ, 938 "DATE": TokenType.DATE, 939 "DATETIME": TokenType.DATETIME, 940 "INT4RANGE": TokenType.INT4RANGE, 941 "INT4MULTIRANGE": TokenType.INT4MULTIRANGE, 942 "INT8RANGE": TokenType.INT8RANGE, 943 "INT8MULTIRANGE": TokenType.INT8MULTIRANGE, 944 "NUMRANGE": TokenType.NUMRANGE, 945 "NUMMULTIRANGE": TokenType.NUMMULTIRANGE, 946 "TSRANGE": TokenType.TSRANGE, 947 "TSMULTIRANGE": TokenType.TSMULTIRANGE, 948 "TSTZRANGE": TokenType.TSTZRANGE, 949 "TSTZMULTIRANGE": TokenType.TSTZMULTIRANGE, 950 "DATERANGE": TokenType.DATERANGE, 951 "DATEMULTIRANGE": TokenType.DATEMULTIRANGE, 952 "UNIQUE": TokenType.UNIQUE, 953 "VECTOR": TokenType.VECTOR, 954 "STRUCT": TokenType.STRUCT, 955 "SEQUENCE": TokenType.SEQUENCE, 956 "VARIANT": TokenType.VARIANT, 957 "ALTER": TokenType.ALTER, 958 "ANALYZE": TokenType.ANALYZE, 959 "CALL": TokenType.COMMAND, 960 "COMMENT": TokenType.COMMENT, 961 "EXPLAIN": TokenType.COMMAND, 962 "GRANT": TokenType.GRANT, 963 "OPTIMIZE": TokenType.COMMAND, 964 "PREPARE": TokenType.COMMAND, 965 "VACUUM": TokenType.COMMAND, 966 "USER-DEFINED": TokenType.USERDEFINED, 967 "FOR VERSION": TokenType.VERSION_SNAPSHOT, 968 "FOR TIMESTAMP": TokenType.TIMESTAMP_SNAPSHOT, 969 } 970 971 WHITE_SPACE: t.Dict[t.Optional[str], TokenType] = { 972 " ": TokenType.SPACE, 973 "\t": TokenType.SPACE, 974 "\n": TokenType.BREAK, 975 "\r": TokenType.BREAK, 976 } 977 978 COMMANDS = { 979 TokenType.COMMAND, 980 TokenType.EXECUTE, 981 TokenType.FETCH, 982 TokenType.SHOW, 983 TokenType.RENAME, 984 } 985 986 COMMAND_PREFIX_TOKENS = {TokenType.SEMICOLON, TokenType.BEGIN} 987 988 # Handle numeric literals like in hive (3L = BIGINT) 989 NUMERIC_LITERALS: t.Dict[str, str] = {} 990 991 COMMENTS = ["--", ("/*", "*/")] 992 993 __slots__ = ( 994 "sql", 995 "size", 996 "tokens", 997 "dialect", 998 "use_rs_tokenizer", 999 "_start", 1000 "_current", 1001 "_line", 1002 "_col", 1003 "_comments", 1004 "_char", 1005 "_end", 1006 "_peek", 1007 "_prev_token_line", 1008 "_rs_dialect_settings", 1009 ) 1010 1011 def __init__( 1012 self, dialect: DialectType = None, use_rs_tokenizer: t.Optional[bool] = None 1013 ) -> None: 1014 from sqlglot.dialects import Dialect 1015 1016 self.dialect = Dialect.get_or_raise(dialect) 1017 1018 # initialize `use_rs_tokenizer`, and allow it to be overwritten per Tokenizer instance 1019 self.use_rs_tokenizer = ( 1020 use_rs_tokenizer if use_rs_tokenizer is not None else USE_RS_TOKENIZER 1021 ) 1022 1023 if self.use_rs_tokenizer: 1024 self._rs_dialect_settings = RsTokenizerDialectSettings( 1025 unescaped_sequences=self.dialect.UNESCAPED_SEQUENCES, 1026 identifiers_can_start_with_digit=self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT, 1027 numbers_can_be_underscore_separated=self.dialect.NUMBERS_CAN_BE_UNDERSCORE_SEPARATED, 1028 ) 1029 1030 self.reset() 1031 1032 def reset(self) -> None: 1033 self.sql = "" 1034 self.size = 0 1035 self.tokens: t.List[Token] = [] 1036 self._start = 0 1037 self._current = 0 1038 self._line = 1 1039 self._col = 0 1040 self._comments: t.List[str] = [] 1041 1042 self._char = "" 1043 self._end = False 1044 self._peek = "" 1045 self._prev_token_line = -1 1046 1047 def tokenize(self, sql: str) -> t.List[Token]: 1048 """Returns a list of tokens corresponding to the SQL string `sql`.""" 1049 if self.use_rs_tokenizer: 1050 return self.tokenize_rs(sql) 1051 1052 self.reset() 1053 self.sql = sql 1054 self.size = len(sql) 1055 1056 try: 1057 self._scan() 1058 except Exception as e: 1059 start = max(self._current - 50, 0) 1060 end = min(self._current + 50, self.size - 1) 1061 context = self.sql[start:end] 1062 raise TokenError(f"Error tokenizing '{context}'") from e 1063 1064 return self.tokens 1065 1066 def _scan(self, until: t.Optional[t.Callable] = None) -> None: 1067 while self.size and not self._end: 1068 current = self._current 1069 1070 # Skip spaces here rather than iteratively calling advance() for performance reasons 1071 while current < self.size: 1072 char = self.sql[current] 1073 1074 if char.isspace() and (char == " " or char == "\t"): 1075 current += 1 1076 else: 1077 break 1078 1079 offset = current - self._current if current > self._current else 1 1080 1081 self._start = current 1082 self._advance(offset) 1083 1084 if not self._char.isspace(): 1085 if self._char.isdigit(): 1086 self._scan_number() 1087 elif self._char in self._IDENTIFIERS: 1088 self._scan_identifier(self._IDENTIFIERS[self._char]) 1089 else: 1090 self._scan_keywords() 1091 1092 if until and until(): 1093 break 1094 1095 if self.tokens and self._comments: 1096 self.tokens[-1].comments.extend(self._comments) 1097 1098 def _chars(self, size: int) -> str: 1099 if size == 1: 1100 return self._char 1101 1102 start = self._current - 1 1103 end = start + size 1104 1105 return self.sql[start:end] if end <= self.size else "" 1106 1107 def _advance(self, i: int = 1, alnum: bool = False) -> None: 1108 if self.WHITE_SPACE.get(self._char) is TokenType.BREAK: 1109 # Ensures we don't count an extra line if we get a \r\n line break sequence 1110 if not (self._char == "\r" and self._peek == "\n"): 1111 self._col = 1 1112 self._line += 1 1113 else: 1114 self._col += i 1115 1116 self._current += i 1117 self._end = self._current >= self.size 1118 self._char = self.sql[self._current - 1] 1119 self._peek = "" if self._end else self.sql[self._current] 1120 1121 if alnum and self._char.isalnum(): 1122 # Here we use local variables instead of attributes for better performance 1123 _col = self._col 1124 _current = self._current 1125 _end = self._end 1126 _peek = self._peek 1127 1128 while _peek.isalnum(): 1129 _col += 1 1130 _current += 1 1131 _end = _current >= self.size 1132 _peek = "" if _end else self.sql[_current] 1133 1134 self._col = _col 1135 self._current = _current 1136 self._end = _end 1137 self._peek = _peek 1138 self._char = self.sql[_current - 1] 1139 1140 @property 1141 def _text(self) -> str: 1142 return self.sql[self._start : self._current] 1143 1144 def _add(self, token_type: TokenType, text: t.Optional[str] = None) -> None: 1145 self._prev_token_line = self._line 1146 1147 if self._comments and token_type == TokenType.SEMICOLON and self.tokens: 1148 self.tokens[-1].comments.extend(self._comments) 1149 self._comments = [] 1150 1151 self.tokens.append( 1152 Token( 1153 token_type, 1154 text=self._text if text is None else text, 1155 line=self._line, 1156 col=self._col, 1157 start=self._start, 1158 end=self._current - 1, 1159 comments=self._comments, 1160 ) 1161 ) 1162 self._comments = [] 1163 1164 # If we have either a semicolon or a begin token before the command's token, we'll parse 1165 # whatever follows the command's token as a string 1166 if ( 1167 token_type in self.COMMANDS 1168 and self._peek != ";" 1169 and (len(self.tokens) == 1 or self.tokens[-2].token_type in self.COMMAND_PREFIX_TOKENS) 1170 ): 1171 start = self._current 1172 tokens = len(self.tokens) 1173 self._scan(lambda: self._peek == ";") 1174 self.tokens = self.tokens[:tokens] 1175 text = self.sql[start : self._current].strip() 1176 if text: 1177 self._add(TokenType.STRING, text) 1178 1179 def _scan_keywords(self) -> None: 1180 size = 0 1181 word = None 1182 chars = self._text 1183 char = chars 1184 prev_space = False 1185 skip = False 1186 trie = self._KEYWORD_TRIE 1187 single_token = char in self.SINGLE_TOKENS 1188 1189 while chars: 1190 if skip: 1191 result = TrieResult.PREFIX 1192 else: 1193 result, trie = in_trie(trie, char.upper()) 1194 1195 if result == TrieResult.FAILED: 1196 break 1197 if result == TrieResult.EXISTS: 1198 word = chars 1199 1200 end = self._current + size 1201 size += 1 1202 1203 if end < self.size: 1204 char = self.sql[end] 1205 single_token = single_token or char in self.SINGLE_TOKENS 1206 is_space = char.isspace() 1207 1208 if not is_space or not prev_space: 1209 if is_space: 1210 char = " " 1211 chars += char 1212 prev_space = is_space 1213 skip = False 1214 else: 1215 skip = True 1216 else: 1217 char = "" 1218 break 1219 1220 if word: 1221 if self._scan_string(word): 1222 return 1223 if self._scan_comment(word): 1224 return 1225 if prev_space or single_token or not char: 1226 self._advance(size - 1) 1227 word = word.upper() 1228 self._add(self.KEYWORDS[word], text=word) 1229 return 1230 1231 if self._char in self.SINGLE_TOKENS: 1232 self._add(self.SINGLE_TOKENS[self._char], text=self._char) 1233 return 1234 1235 self._scan_var() 1236 1237 def _scan_comment(self, comment_start: str) -> bool: 1238 if comment_start not in self._COMMENTS: 1239 return False 1240 1241 comment_start_line = self._line 1242 comment_start_size = len(comment_start) 1243 comment_end = self._COMMENTS[comment_start] 1244 1245 if comment_end: 1246 # Skip the comment's start delimiter 1247 self._advance(comment_start_size) 1248 1249 comment_count = 1 1250 comment_end_size = len(comment_end) 1251 1252 while not self._end: 1253 if self._chars(comment_end_size) == comment_end: 1254 comment_count -= 1 1255 if not comment_count: 1256 break 1257 1258 self._advance(alnum=True) 1259 1260 # Nested comments are allowed by some dialects, e.g. databricks, duckdb, postgres 1261 if ( 1262 self.NESTED_COMMENTS 1263 and not self._end 1264 and self._chars(comment_end_size) == comment_start 1265 ): 1266 self._advance(comment_start_size) 1267 comment_count += 1 1268 1269 self._comments.append(self._text[comment_start_size : -comment_end_size + 1]) 1270 self._advance(comment_end_size - 1) 1271 else: 1272 while not self._end and self.WHITE_SPACE.get(self._peek) is not TokenType.BREAK: 1273 self._advance(alnum=True) 1274 self._comments.append(self._text[comment_start_size:]) 1275 1276 if ( 1277 comment_start == self.HINT_START 1278 and self.tokens 1279 and self.tokens[-1].token_type in self.TOKENS_PRECEDING_HINT 1280 ): 1281 self._add(TokenType.HINT) 1282 1283 # Leading comment is attached to the succeeding token, whilst trailing comment to the preceding. 1284 # Multiple consecutive comments are preserved by appending them to the current comments list. 1285 if comment_start_line == self._prev_token_line: 1286 self.tokens[-1].comments.extend(self._comments) 1287 self._comments = [] 1288 self._prev_token_line = self._line 1289 1290 return True 1291 1292 def _scan_number(self) -> None: 1293 if self._char == "0": 1294 peek = self._peek.upper() 1295 if peek == "B": 1296 return self._scan_bits() if self.BIT_STRINGS else self._add(TokenType.NUMBER) 1297 elif peek == "X": 1298 return self._scan_hex() if self.HEX_STRINGS else self._add(TokenType.NUMBER) 1299 1300 decimal = False 1301 scientific = 0 1302 1303 while True: 1304 if self._peek.isdigit(): 1305 self._advance() 1306 elif self._peek == "." and not decimal: 1307 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER: 1308 return self._add(TokenType.NUMBER) 1309 decimal = True 1310 self._advance() 1311 elif self._peek in ("-", "+") and scientific == 1: 1312 scientific += 1 1313 self._advance() 1314 elif self._peek.upper() == "E" and not scientific: 1315 scientific += 1 1316 self._advance() 1317 elif self._peek.isidentifier(): 1318 number_text = self._text 1319 literal = "" 1320 1321 while self._peek.strip() and self._peek not in self.SINGLE_TOKENS: 1322 literal += self._peek 1323 self._advance() 1324 1325 token_type = self.KEYWORDS.get(self.NUMERIC_LITERALS.get(literal.upper(), "")) 1326 1327 if token_type: 1328 self._add(TokenType.NUMBER, number_text) 1329 self._add(TokenType.DCOLON, "::") 1330 return self._add(token_type, literal) 1331 else: 1332 replaced = literal.replace("_", "") 1333 if self.dialect.NUMBERS_CAN_BE_UNDERSCORE_SEPARATED and replaced.isdigit(): 1334 return self._add(TokenType.NUMBER, number_text + replaced) 1335 if self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT: 1336 return self._add(TokenType.VAR) 1337 1338 self._advance(-len(literal)) 1339 return self._add(TokenType.NUMBER, number_text) 1340 else: 1341 return self._add(TokenType.NUMBER) 1342 1343 def _scan_bits(self) -> None: 1344 self._advance() 1345 value = self._extract_value() 1346 try: 1347 # If `value` can't be converted to a binary, fallback to tokenizing it as an identifier 1348 int(value, 2) 1349 self._add(TokenType.BIT_STRING, value[2:]) # Drop the 0b 1350 except ValueError: 1351 self._add(TokenType.IDENTIFIER) 1352 1353 def _scan_hex(self) -> None: 1354 self._advance() 1355 value = self._extract_value() 1356 try: 1357 # If `value` can't be converted to a hex, fallback to tokenizing it as an identifier 1358 int(value, 16) 1359 self._add(TokenType.HEX_STRING, value[2:]) # Drop the 0x 1360 except ValueError: 1361 self._add(TokenType.IDENTIFIER) 1362 1363 def _extract_value(self) -> str: 1364 while True: 1365 char = self._peek.strip() 1366 if char and char not in self.SINGLE_TOKENS: 1367 self._advance(alnum=True) 1368 else: 1369 break 1370 1371 return self._text 1372 1373 def _scan_string(self, start: str) -> bool: 1374 base = None 1375 token_type = TokenType.STRING 1376 1377 if start in self._QUOTES: 1378 end = self._QUOTES[start] 1379 elif start in self._FORMAT_STRINGS: 1380 end, token_type = self._FORMAT_STRINGS[start] 1381 1382 if token_type == TokenType.HEX_STRING: 1383 base = 16 1384 elif token_type == TokenType.BIT_STRING: 1385 base = 2 1386 elif token_type == TokenType.HEREDOC_STRING: 1387 self._advance() 1388 1389 if self._char == end: 1390 tag = "" 1391 else: 1392 tag = self._extract_string( 1393 end, 1394 raw_string=True, 1395 raise_unmatched=not self.HEREDOC_TAG_IS_IDENTIFIER, 1396 ) 1397 1398 if tag and self.HEREDOC_TAG_IS_IDENTIFIER and (self._end or not tag.isidentifier()): 1399 if not self._end: 1400 self._advance(-1) 1401 1402 self._advance(-len(tag)) 1403 self._add(self.HEREDOC_STRING_ALTERNATIVE) 1404 return True 1405 1406 end = f"{start}{tag}{end}" 1407 else: 1408 return False 1409 1410 self._advance(len(start)) 1411 text = self._extract_string(end, raw_string=token_type == TokenType.RAW_STRING) 1412 1413 if base: 1414 try: 1415 int(text, base) 1416 except Exception: 1417 raise TokenError( 1418 f"Numeric string contains invalid characters from {self._line}:{self._start}" 1419 ) 1420 1421 self._add(token_type, text) 1422 return True 1423 1424 def _scan_identifier(self, identifier_end: str) -> None: 1425 self._advance() 1426 text = self._extract_string( 1427 identifier_end, escapes=self._IDENTIFIER_ESCAPES | {identifier_end} 1428 ) 1429 self._add(TokenType.IDENTIFIER, text) 1430 1431 def _scan_var(self) -> None: 1432 while True: 1433 char = self._peek.strip() 1434 if char and (char in self.VAR_SINGLE_TOKENS or char not in self.SINGLE_TOKENS): 1435 self._advance(alnum=True) 1436 else: 1437 break 1438 1439 self._add( 1440 TokenType.VAR 1441 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER 1442 else self.KEYWORDS.get(self._text.upper(), TokenType.VAR) 1443 ) 1444 1445 def _extract_string( 1446 self, 1447 delimiter: str, 1448 escapes: t.Optional[t.Set[str]] = None, 1449 raw_string: bool = False, 1450 raise_unmatched: bool = True, 1451 ) -> str: 1452 text = "" 1453 delim_size = len(delimiter) 1454 escapes = self._STRING_ESCAPES if escapes is None else escapes 1455 1456 while True: 1457 if ( 1458 not raw_string 1459 and self.dialect.UNESCAPED_SEQUENCES 1460 and self._peek 1461 and self._char in self.STRING_ESCAPES 1462 ): 1463 unescaped_sequence = self.dialect.UNESCAPED_SEQUENCES.get(self._char + self._peek) 1464 if unescaped_sequence: 1465 self._advance(2) 1466 text += unescaped_sequence 1467 continue 1468 if ( 1469 (self.STRING_ESCAPES_ALLOWED_IN_RAW_STRINGS or not raw_string) 1470 and self._char in escapes 1471 and (self._peek == delimiter or self._peek in escapes) 1472 and (self._char not in self._QUOTES or self._char == self._peek) 1473 ): 1474 if self._peek == delimiter: 1475 text += self._peek 1476 else: 1477 text += self._char + self._peek 1478 1479 if self._current + 1 < self.size: 1480 self._advance(2) 1481 else: 1482 raise TokenError(f"Missing {delimiter} from {self._line}:{self._current}") 1483 else: 1484 if self._chars(delim_size) == delimiter: 1485 if delim_size > 1: 1486 self._advance(delim_size - 1) 1487 break 1488 1489 if self._end: 1490 if not raise_unmatched: 1491 return text + self._char 1492 1493 raise TokenError(f"Missing {delimiter} from {self._line}:{self._start}") 1494 1495 current = self._current - 1 1496 self._advance(alnum=True) 1497 text += self.sql[current : self._current - 1] 1498 1499 return text 1500 1501 def tokenize_rs(self, sql: str) -> t.List[Token]: 1502 if not self._RS_TOKENIZER: 1503 raise SqlglotError("Rust tokenizer is not available") 1504 1505 try: 1506 tokens = self._RS_TOKENIZER.tokenize(sql, self._rs_dialect_settings) 1507 for token in tokens: 1508 token.token_type = _ALL_TOKEN_TYPES[token.token_type_index] 1509 return tokens 1510 except Exception as e: 1511 raise TokenError(str(e))
Tokenizer( dialect: Union[str, sqlglot.dialects.Dialect, Type[sqlglot.dialects.Dialect], NoneType] = None, use_rs_tokenizer: Optional[bool] = None)
1011 def __init__( 1012 self, dialect: DialectType = None, use_rs_tokenizer: t.Optional[bool] = None 1013 ) -> None: 1014 from sqlglot.dialects import Dialect 1015 1016 self.dialect = Dialect.get_or_raise(dialect) 1017 1018 # initialize `use_rs_tokenizer`, and allow it to be overwritten per Tokenizer instance 1019 self.use_rs_tokenizer = ( 1020 use_rs_tokenizer if use_rs_tokenizer is not None else USE_RS_TOKENIZER 1021 ) 1022 1023 if self.use_rs_tokenizer: 1024 self._rs_dialect_settings = RsTokenizerDialectSettings( 1025 unescaped_sequences=self.dialect.UNESCAPED_SEQUENCES, 1026 identifiers_can_start_with_digit=self.dialect.IDENTIFIERS_CAN_START_WITH_DIGIT, 1027 numbers_can_be_underscore_separated=self.dialect.NUMBERS_CAN_BE_UNDERSCORE_SEPARATED, 1028 ) 1029 1030 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.SELECT: 'SELECT'>, <TokenType.DELETE: 'DELETE'>, <TokenType.INSERT: 'INSERT'>}
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.EXECUTE: 'EXECUTE'>, <TokenType.SHOW: 'SHOW'>, <TokenType.FETCH: 'FETCH'>, <TokenType.RENAME: 'RENAME'>, <TokenType.COMMAND: 'COMMAND'>}
COMMAND_PREFIX_TOKENS =
{<TokenType.SEMICOLON: 'SEMICOLON'>, <TokenType.BEGIN: 'BEGIN'>}
def
reset(self) -> None:
1032 def reset(self) -> None: 1033 self.sql = "" 1034 self.size = 0 1035 self.tokens: t.List[Token] = [] 1036 self._start = 0 1037 self._current = 0 1038 self._line = 1 1039 self._col = 0 1040 self._comments: t.List[str] = [] 1041 1042 self._char = "" 1043 self._end = False 1044 self._peek = "" 1045 self._prev_token_line = -1
1047 def tokenize(self, sql: str) -> t.List[Token]: 1048 """Returns a list of tokens corresponding to the SQL string `sql`.""" 1049 if self.use_rs_tokenizer: 1050 return self.tokenize_rs(sql) 1051 1052 self.reset() 1053 self.sql = sql 1054 self.size = len(sql) 1055 1056 try: 1057 self._scan() 1058 except Exception as e: 1059 start = max(self._current - 50, 0) 1060 end = min(self._current + 50, self.size - 1) 1061 context = self.sql[start:end] 1062 raise TokenError(f"Error tokenizing '{context}'") from e 1063 1064 return self.tokens
Returns a list of tokens corresponding to the SQL string sql
.
1501 def tokenize_rs(self, sql: str) -> t.List[Token]: 1502 if not self._RS_TOKENIZER: 1503 raise SqlglotError("Rust tokenizer is not available") 1504 1505 try: 1506 tokens = self._RS_TOKENIZER.tokenize(sql, self._rs_dialect_settings) 1507 for token in tokens: 1508 token.token_type = _ALL_TOKEN_TYPES[token.token_type_index] 1509 return tokens 1510 except Exception as e: 1511 raise TokenError(str(e))