sqlglot.tokens
1from __future__ import annotations 2 3import typing as t 4from enum import auto 5 6from sqlglot.helper import AutoName 7from sqlglot.trie import in_trie, new_trie 8 9 10class TokenType(AutoName): 11 L_PAREN = auto() 12 R_PAREN = auto() 13 L_BRACKET = auto() 14 R_BRACKET = auto() 15 L_BRACE = auto() 16 R_BRACE = auto() 17 COMMA = auto() 18 DOT = auto() 19 DASH = auto() 20 PLUS = auto() 21 COLON = auto() 22 DCOLON = auto() 23 SEMICOLON = auto() 24 STAR = auto() 25 BACKSLASH = auto() 26 SLASH = auto() 27 LT = auto() 28 LTE = auto() 29 GT = auto() 30 GTE = auto() 31 NOT = auto() 32 EQ = auto() 33 NEQ = auto() 34 NULLSAFE_EQ = auto() 35 AND = auto() 36 OR = auto() 37 AMP = auto() 38 DPIPE = auto() 39 PIPE = auto() 40 CARET = auto() 41 TILDA = auto() 42 ARROW = auto() 43 DARROW = auto() 44 FARROW = auto() 45 HASH = auto() 46 HASH_ARROW = auto() 47 DHASH_ARROW = auto() 48 LR_ARROW = auto() 49 LT_AT = auto() 50 AT_GT = auto() 51 DOLLAR = auto() 52 PARAMETER = auto() 53 SESSION_PARAMETER = auto() 54 DAMP = auto() 55 56 BLOCK_START = auto() 57 BLOCK_END = auto() 58 59 SPACE = auto() 60 BREAK = auto() 61 62 STRING = auto() 63 NUMBER = auto() 64 IDENTIFIER = auto() 65 DATABASE = auto() 66 COLUMN = auto() 67 COLUMN_DEF = auto() 68 SCHEMA = auto() 69 TABLE = auto() 70 VAR = auto() 71 BIT_STRING = auto() 72 HEX_STRING = auto() 73 BYTE_STRING = auto() 74 NATIONAL_STRING = auto() 75 RAW_STRING = auto() 76 77 # types 78 BIT = auto() 79 BOOLEAN = auto() 80 TINYINT = auto() 81 UTINYINT = auto() 82 SMALLINT = auto() 83 USMALLINT = auto() 84 INT = auto() 85 UINT = auto() 86 BIGINT = auto() 87 UBIGINT = auto() 88 INT128 = auto() 89 UINT128 = auto() 90 INT256 = auto() 91 UINT256 = auto() 92 FLOAT = auto() 93 DOUBLE = auto() 94 DECIMAL = auto() 95 BIGDECIMAL = auto() 96 CHAR = auto() 97 NCHAR = auto() 98 VARCHAR = auto() 99 NVARCHAR = auto() 100 TEXT = auto() 101 MEDIUMTEXT = auto() 102 LONGTEXT = auto() 103 MEDIUMBLOB = auto() 104 LONGBLOB = auto() 105 BINARY = auto() 106 VARBINARY = auto() 107 JSON = auto() 108 JSONB = auto() 109 TIME = auto() 110 TIMESTAMP = auto() 111 TIMESTAMPTZ = auto() 112 TIMESTAMPLTZ = auto() 113 DATETIME = auto() 114 DATETIME64 = auto() 115 DATE = auto() 116 INT4RANGE = auto() 117 INT4MULTIRANGE = auto() 118 INT8RANGE = auto() 119 INT8MULTIRANGE = auto() 120 NUMRANGE = auto() 121 NUMMULTIRANGE = auto() 122 TSRANGE = auto() 123 TSMULTIRANGE = auto() 124 TSTZRANGE = auto() 125 TSTZMULTIRANGE = auto() 126 DATERANGE = auto() 127 DATEMULTIRANGE = auto() 128 UUID = auto() 129 GEOGRAPHY = auto() 130 NULLABLE = auto() 131 GEOMETRY = auto() 132 HLLSKETCH = auto() 133 HSTORE = auto() 134 SUPER = auto() 135 SERIAL = auto() 136 SMALLSERIAL = auto() 137 BIGSERIAL = auto() 138 XML = auto() 139 UNIQUEIDENTIFIER = auto() 140 MONEY = auto() 141 SMALLMONEY = auto() 142 ROWVERSION = auto() 143 IMAGE = auto() 144 VARIANT = auto() 145 OBJECT = auto() 146 INET = auto() 147 148 # keywords 149 ALIAS = auto() 150 ALTER = auto() 151 ALWAYS = auto() 152 ALL = auto() 153 ANTI = auto() 154 ANY = auto() 155 APPLY = auto() 156 ARRAY = auto() 157 ASC = auto() 158 ASOF = auto() 159 AUTO_INCREMENT = auto() 160 BEGIN = auto() 161 BETWEEN = auto() 162 CACHE = auto() 163 CASE = auto() 164 CHARACTER_SET = auto() 165 COLLATE = auto() 166 COMMAND = auto() 167 COMMENT = auto() 168 COMMIT = auto() 169 CONSTRAINT = auto() 170 CREATE = auto() 171 CROSS = auto() 172 CUBE = auto() 173 CURRENT_DATE = auto() 174 CURRENT_DATETIME = auto() 175 CURRENT_TIME = auto() 176 CURRENT_TIMESTAMP = auto() 177 CURRENT_USER = auto() 178 DEFAULT = auto() 179 DELETE = auto() 180 DESC = auto() 181 DESCRIBE = auto() 182 DICTIONARY = auto() 183 DISTINCT = auto() 184 DIV = auto() 185 DROP = auto() 186 ELSE = auto() 187 END = auto() 188 ESCAPE = auto() 189 EXCEPT = auto() 190 EXECUTE = auto() 191 EXISTS = auto() 192 FALSE = auto() 193 FETCH = auto() 194 FILTER = auto() 195 FINAL = auto() 196 FIRST = auto() 197 FOR = auto() 198 FOREIGN_KEY = auto() 199 FORMAT = auto() 200 FROM = auto() 201 FULL = auto() 202 FUNCTION = auto() 203 GLOB = auto() 204 GLOBAL = auto() 205 GROUP_BY = auto() 206 GROUPING_SETS = auto() 207 HAVING = auto() 208 HINT = auto() 209 IF = auto() 210 ILIKE = auto() 211 ILIKE_ANY = auto() 212 IN = auto() 213 INDEX = auto() 214 INNER = auto() 215 INSERT = auto() 216 INTERSECT = auto() 217 INTERVAL = auto() 218 INTO = auto() 219 INTRODUCER = auto() 220 IRLIKE = auto() 221 IS = auto() 222 ISNULL = auto() 223 JOIN = auto() 224 JOIN_MARKER = auto() 225 KEEP = auto() 226 LANGUAGE = auto() 227 LATERAL = auto() 228 LEFT = auto() 229 LIKE = auto() 230 LIKE_ANY = auto() 231 LIMIT = auto() 232 LOAD = auto() 233 LOCK = auto() 234 MAP = auto() 235 MATCH_RECOGNIZE = auto() 236 MERGE = auto() 237 MOD = auto() 238 NATURAL = auto() 239 NEXT = auto() 240 NEXT_VALUE_FOR = auto() 241 NOTNULL = auto() 242 NULL = auto() 243 OFFSET = auto() 244 ON = auto() 245 ORDER_BY = auto() 246 ORDERED = auto() 247 ORDINALITY = auto() 248 OUTER = auto() 249 OVER = auto() 250 OVERLAPS = auto() 251 OVERWRITE = auto() 252 PARTITION = auto() 253 PARTITION_BY = auto() 254 PERCENT = auto() 255 PIVOT = auto() 256 PLACEHOLDER = auto() 257 PRAGMA = auto() 258 PRIMARY_KEY = auto() 259 PROCEDURE = auto() 260 PROPERTIES = auto() 261 PSEUDO_TYPE = auto() 262 QUALIFY = auto() 263 QUOTE = auto() 264 RANGE = auto() 265 RECURSIVE = auto() 266 REPLACE = auto() 267 RETURNING = auto() 268 REFERENCES = auto() 269 RIGHT = auto() 270 RLIKE = auto() 271 ROLLBACK = auto() 272 ROLLUP = auto() 273 ROW = auto() 274 ROWS = auto() 275 SELECT = auto() 276 SEMI = auto() 277 SEPARATOR = auto() 278 SERDE_PROPERTIES = auto() 279 SET = auto() 280 SETTINGS = auto() 281 SHOW = auto() 282 SIMILAR_TO = auto() 283 SOME = auto() 284 STRUCT = auto() 285 TABLE_SAMPLE = auto() 286 TEMPORARY = auto() 287 TOP = auto() 288 THEN = auto() 289 TRUE = auto() 290 UNCACHE = auto() 291 UNION = auto() 292 UNNEST = auto() 293 UNPIVOT = auto() 294 UPDATE = auto() 295 USE = auto() 296 USING = auto() 297 VALUES = auto() 298 VIEW = auto() 299 VOLATILE = auto() 300 WHEN = auto() 301 WHERE = auto() 302 WINDOW = auto() 303 WITH = auto() 304 UNIQUE = auto() 305 306 307class Token: 308 __slots__ = ("token_type", "text", "line", "col", "start", "end", "comments") 309 310 @classmethod 311 def number(cls, number: int) -> Token: 312 """Returns a NUMBER token with `number` as its text.""" 313 return cls(TokenType.NUMBER, str(number)) 314 315 @classmethod 316 def string(cls, string: str) -> Token: 317 """Returns a STRING token with `string` as its text.""" 318 return cls(TokenType.STRING, string) 319 320 @classmethod 321 def identifier(cls, identifier: str) -> Token: 322 """Returns an IDENTIFIER token with `identifier` as its text.""" 323 return cls(TokenType.IDENTIFIER, identifier) 324 325 @classmethod 326 def var(cls, var: str) -> Token: 327 """Returns an VAR token with `var` as its text.""" 328 return cls(TokenType.VAR, var) 329 330 def __init__( 331 self, 332 token_type: TokenType, 333 text: str, 334 line: int = 1, 335 col: int = 1, 336 start: int = 0, 337 end: int = 0, 338 comments: t.List[str] = [], 339 ) -> None: 340 """Token initializer. 341 342 Args: 343 token_type: The TokenType Enum. 344 text: The text of the token. 345 line: The line that the token ends on. 346 col: The column that the token ends on. 347 start: The start index of the token. 348 end: The ending index of the token. 349 comments: The comments to attach to the token. 350 """ 351 self.token_type = token_type 352 self.text = text 353 self.line = line 354 self.col = col 355 self.start = start 356 self.end = end 357 self.comments = comments 358 359 def __repr__(self) -> str: 360 attributes = ", ".join(f"{k}: {getattr(self, k)}" for k in self.__slots__) 361 return f"<Token {attributes}>" 362 363 364class _Tokenizer(type): 365 def __new__(cls, clsname, bases, attrs): 366 klass = super().__new__(cls, clsname, bases, attrs) 367 368 def _convert_quotes(arr: t.List[str | t.Tuple[str, str]]) -> t.Dict[str, str]: 369 return dict( 370 (item, item) if isinstance(item, str) else (item[0], item[1]) for item in arr 371 ) 372 373 def _quotes_to_format( 374 token_type: TokenType, arr: t.List[str | t.Tuple[str, str]] 375 ) -> t.Dict[str, t.Tuple[str, TokenType]]: 376 return {k: (v, token_type) for k, v in _convert_quotes(arr).items()} 377 378 klass._QUOTES = _convert_quotes(klass.QUOTES) 379 klass._IDENTIFIERS = _convert_quotes(klass.IDENTIFIERS) 380 381 klass._FORMAT_STRINGS = { 382 **{ 383 p + s: (e, TokenType.NATIONAL_STRING) 384 for s, e in klass._QUOTES.items() 385 for p in ("n", "N") 386 }, 387 **_quotes_to_format(TokenType.BIT_STRING, klass.BIT_STRINGS), 388 **_quotes_to_format(TokenType.BYTE_STRING, klass.BYTE_STRINGS), 389 **_quotes_to_format(TokenType.HEX_STRING, klass.HEX_STRINGS), 390 **_quotes_to_format(TokenType.RAW_STRING, klass.RAW_STRINGS), 391 } 392 393 klass._STRING_ESCAPES = set(klass.STRING_ESCAPES) 394 klass._IDENTIFIER_ESCAPES = set(klass.IDENTIFIER_ESCAPES) 395 klass._COMMENTS = dict( 396 (comment, None) if isinstance(comment, str) else (comment[0], comment[1]) 397 for comment in klass.COMMENTS 398 ) 399 400 klass._KEYWORD_TRIE = new_trie( 401 key.upper() 402 for key in ( 403 *klass.KEYWORDS, 404 *klass._COMMENTS, 405 *klass._QUOTES, 406 *klass._FORMAT_STRINGS, 407 ) 408 if " " in key or any(single in key for single in klass.SINGLE_TOKENS) 409 ) 410 411 return klass 412 413 414class Tokenizer(metaclass=_Tokenizer): 415 SINGLE_TOKENS = { 416 "(": TokenType.L_PAREN, 417 ")": TokenType.R_PAREN, 418 "[": TokenType.L_BRACKET, 419 "]": TokenType.R_BRACKET, 420 "{": TokenType.L_BRACE, 421 "}": TokenType.R_BRACE, 422 "&": TokenType.AMP, 423 "^": TokenType.CARET, 424 ":": TokenType.COLON, 425 ",": TokenType.COMMA, 426 ".": TokenType.DOT, 427 "-": TokenType.DASH, 428 "=": TokenType.EQ, 429 ">": TokenType.GT, 430 "<": TokenType.LT, 431 "%": TokenType.MOD, 432 "!": TokenType.NOT, 433 "|": TokenType.PIPE, 434 "+": TokenType.PLUS, 435 ";": TokenType.SEMICOLON, 436 "/": TokenType.SLASH, 437 "\\": TokenType.BACKSLASH, 438 "*": TokenType.STAR, 439 "~": TokenType.TILDA, 440 "?": TokenType.PLACEHOLDER, 441 "@": TokenType.PARAMETER, 442 # used for breaking a var like x'y' but nothing else 443 # the token type doesn't matter 444 "'": TokenType.QUOTE, 445 "`": TokenType.IDENTIFIER, 446 '"': TokenType.IDENTIFIER, 447 "#": TokenType.HASH, 448 } 449 450 BIT_STRINGS: t.List[str | t.Tuple[str, str]] = [] 451 BYTE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 452 HEX_STRINGS: t.List[str | t.Tuple[str, str]] = [] 453 RAW_STRINGS: t.List[str | t.Tuple[str, str]] = [] 454 IDENTIFIERS: t.List[str | t.Tuple[str, str]] = ['"'] 455 IDENTIFIER_ESCAPES = ['"'] 456 QUOTES: t.List[t.Tuple[str, str] | str] = ["'"] 457 STRING_ESCAPES = ["'"] 458 VAR_SINGLE_TOKENS: t.Set[str] = set() 459 460 # Autofilled 461 IDENTIFIERS_CAN_START_WITH_DIGIT: bool = False 462 463 _COMMENTS: t.Dict[str, str] = {} 464 _FORMAT_STRINGS: t.Dict[str, t.Tuple[str, TokenType]] = {} 465 _IDENTIFIERS: t.Dict[str, str] = {} 466 _IDENTIFIER_ESCAPES: t.Set[str] = set() 467 _QUOTES: t.Dict[str, str] = {} 468 _STRING_ESCAPES: t.Set[str] = set() 469 _KEYWORD_TRIE: t.Dict = {} 470 471 KEYWORDS: t.Dict[str, TokenType] = { 472 **{f"{{%{postfix}": TokenType.BLOCK_START for postfix in ("", "+", "-")}, 473 **{f"{prefix}%}}": TokenType.BLOCK_END for prefix in ("", "+", "-")}, 474 **{f"{{{{{postfix}": TokenType.BLOCK_START for postfix in ("+", "-")}, 475 **{f"{prefix}}}}}": TokenType.BLOCK_END for prefix in ("+", "-")}, 476 "/*+": TokenType.HINT, 477 "==": TokenType.EQ, 478 "::": TokenType.DCOLON, 479 "||": TokenType.DPIPE, 480 ">=": TokenType.GTE, 481 "<=": TokenType.LTE, 482 "<>": TokenType.NEQ, 483 "!=": TokenType.NEQ, 484 "<=>": TokenType.NULLSAFE_EQ, 485 "->": TokenType.ARROW, 486 "->>": TokenType.DARROW, 487 "=>": TokenType.FARROW, 488 "#>": TokenType.HASH_ARROW, 489 "#>>": TokenType.DHASH_ARROW, 490 "<->": TokenType.LR_ARROW, 491 "&&": TokenType.DAMP, 492 "ALL": TokenType.ALL, 493 "ALWAYS": TokenType.ALWAYS, 494 "AND": TokenType.AND, 495 "ANTI": TokenType.ANTI, 496 "ANY": TokenType.ANY, 497 "ASC": TokenType.ASC, 498 "AS": TokenType.ALIAS, 499 "ASOF": TokenType.ASOF, 500 "AUTOINCREMENT": TokenType.AUTO_INCREMENT, 501 "AUTO_INCREMENT": TokenType.AUTO_INCREMENT, 502 "BEGIN": TokenType.BEGIN, 503 "BETWEEN": TokenType.BETWEEN, 504 "CACHE": TokenType.CACHE, 505 "UNCACHE": TokenType.UNCACHE, 506 "CASE": TokenType.CASE, 507 "CHARACTER SET": TokenType.CHARACTER_SET, 508 "COLLATE": TokenType.COLLATE, 509 "COLUMN": TokenType.COLUMN, 510 "COMMIT": TokenType.COMMIT, 511 "CONSTRAINT": TokenType.CONSTRAINT, 512 "CREATE": TokenType.CREATE, 513 "CROSS": TokenType.CROSS, 514 "CUBE": TokenType.CUBE, 515 "CURRENT_DATE": TokenType.CURRENT_DATE, 516 "CURRENT_TIME": TokenType.CURRENT_TIME, 517 "CURRENT_TIMESTAMP": TokenType.CURRENT_TIMESTAMP, 518 "CURRENT_USER": TokenType.CURRENT_USER, 519 "DATABASE": TokenType.DATABASE, 520 "DEFAULT": TokenType.DEFAULT, 521 "DELETE": TokenType.DELETE, 522 "DESC": TokenType.DESC, 523 "DESCRIBE": TokenType.DESCRIBE, 524 "DISTINCT": TokenType.DISTINCT, 525 "DIV": TokenType.DIV, 526 "DROP": TokenType.DROP, 527 "ELSE": TokenType.ELSE, 528 "END": TokenType.END, 529 "ESCAPE": TokenType.ESCAPE, 530 "EXCEPT": TokenType.EXCEPT, 531 "EXECUTE": TokenType.EXECUTE, 532 "EXISTS": TokenType.EXISTS, 533 "FALSE": TokenType.FALSE, 534 "FETCH": TokenType.FETCH, 535 "FILTER": TokenType.FILTER, 536 "FIRST": TokenType.FIRST, 537 "FULL": TokenType.FULL, 538 "FUNCTION": TokenType.FUNCTION, 539 "FOR": TokenType.FOR, 540 "FOREIGN KEY": TokenType.FOREIGN_KEY, 541 "FORMAT": TokenType.FORMAT, 542 "FROM": TokenType.FROM, 543 "GEOGRAPHY": TokenType.GEOGRAPHY, 544 "GEOMETRY": TokenType.GEOMETRY, 545 "GLOB": TokenType.GLOB, 546 "GROUP BY": TokenType.GROUP_BY, 547 "GROUPING SETS": TokenType.GROUPING_SETS, 548 "HAVING": TokenType.HAVING, 549 "IF": TokenType.IF, 550 "ILIKE": TokenType.ILIKE, 551 "IN": TokenType.IN, 552 "INDEX": TokenType.INDEX, 553 "INET": TokenType.INET, 554 "INNER": TokenType.INNER, 555 "INSERT": TokenType.INSERT, 556 "INTERVAL": TokenType.INTERVAL, 557 "INTERSECT": TokenType.INTERSECT, 558 "INTO": TokenType.INTO, 559 "IS": TokenType.IS, 560 "ISNULL": TokenType.ISNULL, 561 "JOIN": TokenType.JOIN, 562 "KEEP": TokenType.KEEP, 563 "LATERAL": TokenType.LATERAL, 564 "LEFT": TokenType.LEFT, 565 "LIKE": TokenType.LIKE, 566 "LIMIT": TokenType.LIMIT, 567 "LOAD": TokenType.LOAD, 568 "LOCK": TokenType.LOCK, 569 "MERGE": TokenType.MERGE, 570 "NATURAL": TokenType.NATURAL, 571 "NEXT": TokenType.NEXT, 572 "NEXT VALUE FOR": TokenType.NEXT_VALUE_FOR, 573 "NOT": TokenType.NOT, 574 "NOTNULL": TokenType.NOTNULL, 575 "NULL": TokenType.NULL, 576 "OBJECT": TokenType.OBJECT, 577 "OFFSET": TokenType.OFFSET, 578 "ON": TokenType.ON, 579 "OR": TokenType.OR, 580 "ORDER BY": TokenType.ORDER_BY, 581 "ORDINALITY": TokenType.ORDINALITY, 582 "OUTER": TokenType.OUTER, 583 "OVER": TokenType.OVER, 584 "OVERLAPS": TokenType.OVERLAPS, 585 "OVERWRITE": TokenType.OVERWRITE, 586 "PARTITION": TokenType.PARTITION, 587 "PARTITION BY": TokenType.PARTITION_BY, 588 "PARTITIONED BY": TokenType.PARTITION_BY, 589 "PARTITIONED_BY": TokenType.PARTITION_BY, 590 "PERCENT": TokenType.PERCENT, 591 "PIVOT": TokenType.PIVOT, 592 "PRAGMA": TokenType.PRAGMA, 593 "PRIMARY KEY": TokenType.PRIMARY_KEY, 594 "PROCEDURE": TokenType.PROCEDURE, 595 "QUALIFY": TokenType.QUALIFY, 596 "RANGE": TokenType.RANGE, 597 "RECURSIVE": TokenType.RECURSIVE, 598 "REGEXP": TokenType.RLIKE, 599 "REPLACE": TokenType.REPLACE, 600 "REFERENCES": TokenType.REFERENCES, 601 "RIGHT": TokenType.RIGHT, 602 "RLIKE": TokenType.RLIKE, 603 "ROLLBACK": TokenType.ROLLBACK, 604 "ROLLUP": TokenType.ROLLUP, 605 "ROW": TokenType.ROW, 606 "ROWS": TokenType.ROWS, 607 "SCHEMA": TokenType.SCHEMA, 608 "SELECT": TokenType.SELECT, 609 "SEMI": TokenType.SEMI, 610 "SET": TokenType.SET, 611 "SETTINGS": TokenType.SETTINGS, 612 "SHOW": TokenType.SHOW, 613 "SIMILAR TO": TokenType.SIMILAR_TO, 614 "SOME": TokenType.SOME, 615 "TABLE": TokenType.TABLE, 616 "TABLESAMPLE": TokenType.TABLE_SAMPLE, 617 "TEMP": TokenType.TEMPORARY, 618 "TEMPORARY": TokenType.TEMPORARY, 619 "THEN": TokenType.THEN, 620 "TRUE": TokenType.TRUE, 621 "UNION": TokenType.UNION, 622 "UNNEST": TokenType.UNNEST, 623 "UNPIVOT": TokenType.UNPIVOT, 624 "UPDATE": TokenType.UPDATE, 625 "USE": TokenType.USE, 626 "USING": TokenType.USING, 627 "UUID": TokenType.UUID, 628 "VALUES": TokenType.VALUES, 629 "VIEW": TokenType.VIEW, 630 "VOLATILE": TokenType.VOLATILE, 631 "WHEN": TokenType.WHEN, 632 "WHERE": TokenType.WHERE, 633 "WINDOW": TokenType.WINDOW, 634 "WITH": TokenType.WITH, 635 "APPLY": TokenType.APPLY, 636 "ARRAY": TokenType.ARRAY, 637 "BIT": TokenType.BIT, 638 "BOOL": TokenType.BOOLEAN, 639 "BOOLEAN": TokenType.BOOLEAN, 640 "BYTE": TokenType.TINYINT, 641 "TINYINT": TokenType.TINYINT, 642 "SHORT": TokenType.SMALLINT, 643 "SMALLINT": TokenType.SMALLINT, 644 "INT2": TokenType.SMALLINT, 645 "INTEGER": TokenType.INT, 646 "INT": TokenType.INT, 647 "INT4": TokenType.INT, 648 "LONG": TokenType.BIGINT, 649 "BIGINT": TokenType.BIGINT, 650 "INT8": TokenType.BIGINT, 651 "DEC": TokenType.DECIMAL, 652 "DECIMAL": TokenType.DECIMAL, 653 "BIGDECIMAL": TokenType.BIGDECIMAL, 654 "BIGNUMERIC": TokenType.BIGDECIMAL, 655 "MAP": TokenType.MAP, 656 "NULLABLE": TokenType.NULLABLE, 657 "NUMBER": TokenType.DECIMAL, 658 "NUMERIC": TokenType.DECIMAL, 659 "FIXED": TokenType.DECIMAL, 660 "REAL": TokenType.FLOAT, 661 "FLOAT": TokenType.FLOAT, 662 "FLOAT4": TokenType.FLOAT, 663 "FLOAT8": TokenType.DOUBLE, 664 "DOUBLE": TokenType.DOUBLE, 665 "DOUBLE PRECISION": TokenType.DOUBLE, 666 "JSON": TokenType.JSON, 667 "CHAR": TokenType.CHAR, 668 "CHARACTER": TokenType.CHAR, 669 "NCHAR": TokenType.NCHAR, 670 "VARCHAR": TokenType.VARCHAR, 671 "VARCHAR2": TokenType.VARCHAR, 672 "NVARCHAR": TokenType.NVARCHAR, 673 "NVARCHAR2": TokenType.NVARCHAR, 674 "STR": TokenType.TEXT, 675 "STRING": TokenType.TEXT, 676 "TEXT": TokenType.TEXT, 677 "CLOB": TokenType.TEXT, 678 "LONGVARCHAR": TokenType.TEXT, 679 "BINARY": TokenType.BINARY, 680 "BLOB": TokenType.VARBINARY, 681 "BYTEA": TokenType.VARBINARY, 682 "VARBINARY": TokenType.VARBINARY, 683 "TIME": TokenType.TIME, 684 "TIMESTAMP": TokenType.TIMESTAMP, 685 "TIMESTAMPTZ": TokenType.TIMESTAMPTZ, 686 "TIMESTAMPLTZ": TokenType.TIMESTAMPLTZ, 687 "DATE": TokenType.DATE, 688 "DATETIME": TokenType.DATETIME, 689 "INT4RANGE": TokenType.INT4RANGE, 690 "INT4MULTIRANGE": TokenType.INT4MULTIRANGE, 691 "INT8RANGE": TokenType.INT8RANGE, 692 "INT8MULTIRANGE": TokenType.INT8MULTIRANGE, 693 "NUMRANGE": TokenType.NUMRANGE, 694 "NUMMULTIRANGE": TokenType.NUMMULTIRANGE, 695 "TSRANGE": TokenType.TSRANGE, 696 "TSMULTIRANGE": TokenType.TSMULTIRANGE, 697 "TSTZRANGE": TokenType.TSTZRANGE, 698 "TSTZMULTIRANGE": TokenType.TSTZMULTIRANGE, 699 "DATERANGE": TokenType.DATERANGE, 700 "DATEMULTIRANGE": TokenType.DATEMULTIRANGE, 701 "UNIQUE": TokenType.UNIQUE, 702 "STRUCT": TokenType.STRUCT, 703 "VARIANT": TokenType.VARIANT, 704 "ALTER": TokenType.ALTER, 705 "ANALYZE": TokenType.COMMAND, 706 "CALL": TokenType.COMMAND, 707 "COMMENT": TokenType.COMMENT, 708 "COPY": TokenType.COMMAND, 709 "EXPLAIN": TokenType.COMMAND, 710 "GRANT": TokenType.COMMAND, 711 "OPTIMIZE": TokenType.COMMAND, 712 "PREPARE": TokenType.COMMAND, 713 "TRUNCATE": TokenType.COMMAND, 714 "VACUUM": TokenType.COMMAND, 715 } 716 717 WHITE_SPACE: t.Dict[t.Optional[str], TokenType] = { 718 " ": TokenType.SPACE, 719 "\t": TokenType.SPACE, 720 "\n": TokenType.BREAK, 721 "\r": TokenType.BREAK, 722 "\r\n": TokenType.BREAK, 723 } 724 725 COMMANDS = { 726 TokenType.COMMAND, 727 TokenType.EXECUTE, 728 TokenType.FETCH, 729 TokenType.SHOW, 730 } 731 732 COMMAND_PREFIX_TOKENS = {TokenType.SEMICOLON, TokenType.BEGIN} 733 734 # handle numeric literals like in hive (3L = BIGINT) 735 NUMERIC_LITERALS: t.Dict[str, str] = {} 736 ENCODE: t.Optional[str] = None 737 738 COMMENTS = ["--", ("/*", "*/"), ("{#", "#}")] 739 740 __slots__ = ( 741 "sql", 742 "size", 743 "tokens", 744 "_start", 745 "_current", 746 "_line", 747 "_col", 748 "_comments", 749 "_char", 750 "_end", 751 "_peek", 752 "_prev_token_line", 753 ) 754 755 def __init__(self) -> None: 756 self.reset() 757 758 def reset(self) -> None: 759 self.sql = "" 760 self.size = 0 761 self.tokens: t.List[Token] = [] 762 self._start = 0 763 self._current = 0 764 self._line = 1 765 self._col = 0 766 self._comments: t.List[str] = [] 767 768 self._char = "" 769 self._end = False 770 self._peek = "" 771 self._prev_token_line = -1 772 773 def tokenize(self, sql: str) -> t.List[Token]: 774 """Returns a list of tokens corresponding to the SQL string `sql`.""" 775 self.reset() 776 self.sql = sql 777 self.size = len(sql) 778 779 try: 780 self._scan() 781 except Exception as e: 782 start = max(self._current - 50, 0) 783 end = min(self._current + 50, self.size - 1) 784 context = self.sql[start:end] 785 raise ValueError(f"Error tokenizing '{context}'") from e 786 787 return self.tokens 788 789 def _scan(self, until: t.Optional[t.Callable] = None) -> None: 790 while self.size and not self._end: 791 self._start = self._current 792 self._advance() 793 794 if self._char is None: 795 break 796 797 if self._char not in self.WHITE_SPACE: 798 if self._char.isdigit(): 799 self._scan_number() 800 elif self._char in self._IDENTIFIERS: 801 self._scan_identifier(self._IDENTIFIERS[self._char]) 802 else: 803 self._scan_keywords() 804 805 if until and until(): 806 break 807 808 if self.tokens and self._comments: 809 self.tokens[-1].comments.extend(self._comments) 810 811 def _chars(self, size: int) -> str: 812 if size == 1: 813 return self._char 814 815 start = self._current - 1 816 end = start + size 817 818 return self.sql[start:end] if end <= self.size else "" 819 820 def _advance(self, i: int = 1, alnum: bool = False) -> None: 821 if self.WHITE_SPACE.get(self._char) is TokenType.BREAK: 822 self._col = 1 823 self._line += 1 824 else: 825 self._col += i 826 827 self._current += i 828 self._end = self._current >= self.size 829 self._char = self.sql[self._current - 1] 830 self._peek = "" if self._end else self.sql[self._current] 831 832 if alnum and self._char.isalnum(): 833 # Here we use local variables instead of attributes for better performance 834 _col = self._col 835 _current = self._current 836 _end = self._end 837 _peek = self._peek 838 839 while _peek.isalnum(): 840 _col += 1 841 _current += 1 842 _end = _current >= self.size 843 _peek = "" if _end else self.sql[_current] 844 845 self._col = _col 846 self._current = _current 847 self._end = _end 848 self._peek = _peek 849 self._char = self.sql[_current - 1] 850 851 @property 852 def _text(self) -> str: 853 return self.sql[self._start : self._current] 854 855 def peek(self, i: int = 0) -> str: 856 i = self._current + i 857 if i < self.size: 858 return self.sql[i] 859 return "" 860 861 def _add(self, token_type: TokenType, text: t.Optional[str] = None) -> None: 862 self._prev_token_line = self._line 863 self.tokens.append( 864 Token( 865 token_type, 866 text=self._text if text is None else text, 867 line=self._line, 868 col=self._col, 869 start=self._start, 870 end=self._current - 1, 871 comments=self._comments, 872 ) 873 ) 874 self._comments = [] 875 876 # If we have either a semicolon or a begin token before the command's token, we'll parse 877 # whatever follows the command's token as a string 878 if ( 879 token_type in self.COMMANDS 880 and self._peek != ";" 881 and (len(self.tokens) == 1 or self.tokens[-2].token_type in self.COMMAND_PREFIX_TOKENS) 882 ): 883 start = self._current 884 tokens = len(self.tokens) 885 self._scan(lambda: self._peek == ";") 886 self.tokens = self.tokens[:tokens] 887 text = self.sql[start : self._current].strip() 888 if text: 889 self._add(TokenType.STRING, text) 890 891 def _scan_keywords(self) -> None: 892 size = 0 893 word = None 894 chars = self._text 895 char = chars 896 prev_space = False 897 skip = False 898 trie = self._KEYWORD_TRIE 899 single_token = char in self.SINGLE_TOKENS 900 901 while chars: 902 if skip: 903 result = 1 904 else: 905 result, trie = in_trie(trie, char.upper()) 906 907 if result == 0: 908 break 909 if result == 2: 910 word = chars 911 912 size += 1 913 end = self._current - 1 + size 914 915 if end < self.size: 916 char = self.sql[end] 917 single_token = single_token or char in self.SINGLE_TOKENS 918 is_space = char in self.WHITE_SPACE 919 920 if not is_space or not prev_space: 921 if is_space: 922 char = " " 923 chars += char 924 prev_space = is_space 925 skip = False 926 else: 927 skip = True 928 else: 929 char = "" 930 chars = " " 931 932 word = None if not single_token and chars[-1] not in self.WHITE_SPACE else word 933 934 if not word: 935 if self._char in self.SINGLE_TOKENS: 936 self._add(self.SINGLE_TOKENS[self._char], text=self._char) 937 return 938 self._scan_var() 939 return 940 941 if self._scan_string(word): 942 return 943 if self._scan_comment(word): 944 return 945 946 self._advance(size - 1) 947 word = word.upper() 948 self._add(self.KEYWORDS[word], text=word) 949 950 def _scan_comment(self, comment_start: str) -> bool: 951 if comment_start not in self._COMMENTS: 952 return False 953 954 comment_start_line = self._line 955 comment_start_size = len(comment_start) 956 comment_end = self._COMMENTS[comment_start] 957 958 if comment_end: 959 # Skip the comment's start delimiter 960 self._advance(comment_start_size) 961 962 comment_end_size = len(comment_end) 963 while not self._end and self._chars(comment_end_size) != comment_end: 964 self._advance(alnum=True) 965 966 self._comments.append(self._text[comment_start_size : -comment_end_size + 1]) 967 self._advance(comment_end_size - 1) 968 else: 969 while not self._end and not self.WHITE_SPACE.get(self._peek) is TokenType.BREAK: 970 self._advance(alnum=True) 971 self._comments.append(self._text[comment_start_size:]) 972 973 # Leading comment is attached to the succeeding token, whilst trailing comment to the preceding. 974 # Multiple consecutive comments are preserved by appending them to the current comments list. 975 if comment_start_line == self._prev_token_line: 976 self.tokens[-1].comments.extend(self._comments) 977 self._comments = [] 978 self._prev_token_line = self._line 979 980 return True 981 982 def _scan_number(self) -> None: 983 if self._char == "0": 984 peek = self._peek.upper() 985 if peek == "B": 986 return self._scan_bits() if self.BIT_STRINGS else self._add(TokenType.NUMBER) 987 elif peek == "X": 988 return self._scan_hex() if self.HEX_STRINGS else self._add(TokenType.NUMBER) 989 990 decimal = False 991 scientific = 0 992 993 while True: 994 if self._peek.isdigit(): 995 self._advance() 996 elif self._peek == "." and not decimal: 997 after = self.peek(1) 998 if after.isdigit() or not after.strip(): 999 decimal = True 1000 self._advance() 1001 else: 1002 return self._add(TokenType.VAR) 1003 elif self._peek in ("-", "+") and scientific == 1: 1004 scientific += 1 1005 self._advance() 1006 elif self._peek.upper() == "E" and not scientific: 1007 scientific += 1 1008 self._advance() 1009 elif self._peek.isidentifier(): 1010 number_text = self._text 1011 literal = "" 1012 1013 while self._peek.strip() and self._peek not in self.SINGLE_TOKENS: 1014 literal += self._peek.upper() 1015 self._advance() 1016 1017 token_type = self.KEYWORDS.get(self.NUMERIC_LITERALS.get(literal, "")) 1018 1019 if token_type: 1020 self._add(TokenType.NUMBER, number_text) 1021 self._add(TokenType.DCOLON, "::") 1022 return self._add(token_type, literal) 1023 elif self.IDENTIFIERS_CAN_START_WITH_DIGIT: 1024 return self._add(TokenType.VAR) 1025 1026 self._add(TokenType.NUMBER, number_text) 1027 return self._advance(-len(literal)) 1028 else: 1029 return self._add(TokenType.NUMBER) 1030 1031 def _scan_bits(self) -> None: 1032 self._advance() 1033 value = self._extract_value() 1034 try: 1035 # If `value` can't be converted to a binary, fallback to tokenizing it as an identifier 1036 int(value, 2) 1037 self._add(TokenType.BIT_STRING, value[2:]) # Drop the 0b 1038 except ValueError: 1039 self._add(TokenType.IDENTIFIER) 1040 1041 def _scan_hex(self) -> None: 1042 self._advance() 1043 value = self._extract_value() 1044 try: 1045 # If `value` can't be converted to a hex, fallback to tokenizing it as an identifier 1046 int(value, 16) 1047 self._add(TokenType.HEX_STRING, value[2:]) # Drop the 0x 1048 except ValueError: 1049 self._add(TokenType.IDENTIFIER) 1050 1051 def _extract_value(self) -> str: 1052 while True: 1053 char = self._peek.strip() 1054 if char and char not in self.SINGLE_TOKENS: 1055 self._advance(alnum=True) 1056 else: 1057 break 1058 1059 return self._text 1060 1061 def _scan_string(self, start: str) -> bool: 1062 base = None 1063 token_type = TokenType.STRING 1064 1065 if start in self._QUOTES: 1066 end = self._QUOTES[start] 1067 elif start in self._FORMAT_STRINGS: 1068 end, token_type = self._FORMAT_STRINGS[start] 1069 1070 if token_type == TokenType.HEX_STRING: 1071 base = 16 1072 elif token_type == TokenType.BIT_STRING: 1073 base = 2 1074 else: 1075 return False 1076 1077 self._advance(len(start)) 1078 text = self._extract_string(end) 1079 1080 if base: 1081 try: 1082 int(text, base) 1083 except: 1084 raise RuntimeError( 1085 f"Numeric string contains invalid characters from {self._line}:{self._start}" 1086 ) 1087 else: 1088 text = text.encode(self.ENCODE).decode(self.ENCODE) if self.ENCODE else text 1089 1090 self._add(token_type, text) 1091 return True 1092 1093 def _scan_identifier(self, identifier_end: str) -> None: 1094 self._advance() 1095 text = self._extract_string(identifier_end, self._IDENTIFIER_ESCAPES) 1096 self._add(TokenType.IDENTIFIER, text) 1097 1098 def _scan_var(self) -> None: 1099 while True: 1100 char = self._peek.strip() 1101 if char and (char in self.VAR_SINGLE_TOKENS or char not in self.SINGLE_TOKENS): 1102 self._advance(alnum=True) 1103 else: 1104 break 1105 1106 self._add( 1107 TokenType.VAR 1108 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER 1109 else self.KEYWORDS.get(self._text.upper(), TokenType.VAR) 1110 ) 1111 1112 def _extract_string(self, delimiter: str, escapes=None) -> str: 1113 text = "" 1114 delim_size = len(delimiter) 1115 escapes = self._STRING_ESCAPES if escapes is None else escapes 1116 1117 while True: 1118 if self._char in escapes and (self._peek == delimiter or self._peek in escapes): 1119 if self._peek == delimiter: 1120 text += self._peek 1121 else: 1122 text += self._char + self._peek 1123 1124 if self._current + 1 < self.size: 1125 self._advance(2) 1126 else: 1127 raise RuntimeError(f"Missing {delimiter} from {self._line}:{self._current}") 1128 else: 1129 if self._chars(delim_size) == delimiter: 1130 if delim_size > 1: 1131 self._advance(delim_size - 1) 1132 break 1133 1134 if self._end: 1135 raise RuntimeError(f"Missing {delimiter} from {self._line}:{self._start}") 1136 1137 current = self._current - 1 1138 self._advance(alnum=True) 1139 text += self.sql[current : self._current - 1] 1140 1141 return text
11class TokenType(AutoName): 12 L_PAREN = auto() 13 R_PAREN = auto() 14 L_BRACKET = auto() 15 R_BRACKET = auto() 16 L_BRACE = auto() 17 R_BRACE = auto() 18 COMMA = auto() 19 DOT = auto() 20 DASH = auto() 21 PLUS = auto() 22 COLON = auto() 23 DCOLON = auto() 24 SEMICOLON = auto() 25 STAR = auto() 26 BACKSLASH = auto() 27 SLASH = auto() 28 LT = auto() 29 LTE = auto() 30 GT = auto() 31 GTE = auto() 32 NOT = auto() 33 EQ = auto() 34 NEQ = auto() 35 NULLSAFE_EQ = auto() 36 AND = auto() 37 OR = auto() 38 AMP = auto() 39 DPIPE = auto() 40 PIPE = auto() 41 CARET = auto() 42 TILDA = auto() 43 ARROW = auto() 44 DARROW = auto() 45 FARROW = auto() 46 HASH = auto() 47 HASH_ARROW = auto() 48 DHASH_ARROW = auto() 49 LR_ARROW = auto() 50 LT_AT = auto() 51 AT_GT = auto() 52 DOLLAR = auto() 53 PARAMETER = auto() 54 SESSION_PARAMETER = auto() 55 DAMP = auto() 56 57 BLOCK_START = auto() 58 BLOCK_END = auto() 59 60 SPACE = auto() 61 BREAK = auto() 62 63 STRING = auto() 64 NUMBER = auto() 65 IDENTIFIER = auto() 66 DATABASE = auto() 67 COLUMN = auto() 68 COLUMN_DEF = auto() 69 SCHEMA = auto() 70 TABLE = auto() 71 VAR = auto() 72 BIT_STRING = auto() 73 HEX_STRING = auto() 74 BYTE_STRING = auto() 75 NATIONAL_STRING = auto() 76 RAW_STRING = auto() 77 78 # types 79 BIT = auto() 80 BOOLEAN = auto() 81 TINYINT = auto() 82 UTINYINT = auto() 83 SMALLINT = auto() 84 USMALLINT = auto() 85 INT = auto() 86 UINT = auto() 87 BIGINT = auto() 88 UBIGINT = auto() 89 INT128 = auto() 90 UINT128 = auto() 91 INT256 = auto() 92 UINT256 = auto() 93 FLOAT = auto() 94 DOUBLE = auto() 95 DECIMAL = auto() 96 BIGDECIMAL = auto() 97 CHAR = auto() 98 NCHAR = auto() 99 VARCHAR = auto() 100 NVARCHAR = auto() 101 TEXT = auto() 102 MEDIUMTEXT = auto() 103 LONGTEXT = auto() 104 MEDIUMBLOB = auto() 105 LONGBLOB = auto() 106 BINARY = auto() 107 VARBINARY = auto() 108 JSON = auto() 109 JSONB = auto() 110 TIME = auto() 111 TIMESTAMP = auto() 112 TIMESTAMPTZ = auto() 113 TIMESTAMPLTZ = auto() 114 DATETIME = auto() 115 DATETIME64 = auto() 116 DATE = auto() 117 INT4RANGE = auto() 118 INT4MULTIRANGE = auto() 119 INT8RANGE = auto() 120 INT8MULTIRANGE = auto() 121 NUMRANGE = auto() 122 NUMMULTIRANGE = auto() 123 TSRANGE = auto() 124 TSMULTIRANGE = auto() 125 TSTZRANGE = auto() 126 TSTZMULTIRANGE = auto() 127 DATERANGE = auto() 128 DATEMULTIRANGE = auto() 129 UUID = auto() 130 GEOGRAPHY = auto() 131 NULLABLE = auto() 132 GEOMETRY = auto() 133 HLLSKETCH = auto() 134 HSTORE = auto() 135 SUPER = auto() 136 SERIAL = auto() 137 SMALLSERIAL = auto() 138 BIGSERIAL = auto() 139 XML = auto() 140 UNIQUEIDENTIFIER = auto() 141 MONEY = auto() 142 SMALLMONEY = auto() 143 ROWVERSION = auto() 144 IMAGE = auto() 145 VARIANT = auto() 146 OBJECT = auto() 147 INET = auto() 148 149 # keywords 150 ALIAS = auto() 151 ALTER = auto() 152 ALWAYS = auto() 153 ALL = auto() 154 ANTI = auto() 155 ANY = auto() 156 APPLY = auto() 157 ARRAY = auto() 158 ASC = auto() 159 ASOF = auto() 160 AUTO_INCREMENT = auto() 161 BEGIN = auto() 162 BETWEEN = auto() 163 CACHE = auto() 164 CASE = auto() 165 CHARACTER_SET = auto() 166 COLLATE = auto() 167 COMMAND = auto() 168 COMMENT = auto() 169 COMMIT = auto() 170 CONSTRAINT = auto() 171 CREATE = auto() 172 CROSS = auto() 173 CUBE = auto() 174 CURRENT_DATE = auto() 175 CURRENT_DATETIME = auto() 176 CURRENT_TIME = auto() 177 CURRENT_TIMESTAMP = auto() 178 CURRENT_USER = auto() 179 DEFAULT = auto() 180 DELETE = auto() 181 DESC = auto() 182 DESCRIBE = auto() 183 DICTIONARY = auto() 184 DISTINCT = auto() 185 DIV = auto() 186 DROP = auto() 187 ELSE = auto() 188 END = auto() 189 ESCAPE = auto() 190 EXCEPT = auto() 191 EXECUTE = auto() 192 EXISTS = auto() 193 FALSE = auto() 194 FETCH = auto() 195 FILTER = auto() 196 FINAL = auto() 197 FIRST = auto() 198 FOR = auto() 199 FOREIGN_KEY = auto() 200 FORMAT = auto() 201 FROM = auto() 202 FULL = auto() 203 FUNCTION = auto() 204 GLOB = auto() 205 GLOBAL = auto() 206 GROUP_BY = auto() 207 GROUPING_SETS = auto() 208 HAVING = auto() 209 HINT = auto() 210 IF = auto() 211 ILIKE = auto() 212 ILIKE_ANY = auto() 213 IN = auto() 214 INDEX = auto() 215 INNER = auto() 216 INSERT = auto() 217 INTERSECT = auto() 218 INTERVAL = auto() 219 INTO = auto() 220 INTRODUCER = auto() 221 IRLIKE = auto() 222 IS = auto() 223 ISNULL = auto() 224 JOIN = auto() 225 JOIN_MARKER = auto() 226 KEEP = auto() 227 LANGUAGE = auto() 228 LATERAL = auto() 229 LEFT = auto() 230 LIKE = auto() 231 LIKE_ANY = auto() 232 LIMIT = auto() 233 LOAD = auto() 234 LOCK = auto() 235 MAP = auto() 236 MATCH_RECOGNIZE = auto() 237 MERGE = auto() 238 MOD = auto() 239 NATURAL = auto() 240 NEXT = auto() 241 NEXT_VALUE_FOR = auto() 242 NOTNULL = auto() 243 NULL = auto() 244 OFFSET = auto() 245 ON = auto() 246 ORDER_BY = auto() 247 ORDERED = auto() 248 ORDINALITY = auto() 249 OUTER = auto() 250 OVER = auto() 251 OVERLAPS = auto() 252 OVERWRITE = auto() 253 PARTITION = auto() 254 PARTITION_BY = auto() 255 PERCENT = auto() 256 PIVOT = auto() 257 PLACEHOLDER = auto() 258 PRAGMA = auto() 259 PRIMARY_KEY = auto() 260 PROCEDURE = auto() 261 PROPERTIES = auto() 262 PSEUDO_TYPE = auto() 263 QUALIFY = auto() 264 QUOTE = auto() 265 RANGE = auto() 266 RECURSIVE = auto() 267 REPLACE = auto() 268 RETURNING = auto() 269 REFERENCES = auto() 270 RIGHT = auto() 271 RLIKE = auto() 272 ROLLBACK = auto() 273 ROLLUP = auto() 274 ROW = auto() 275 ROWS = auto() 276 SELECT = auto() 277 SEMI = auto() 278 SEPARATOR = auto() 279 SERDE_PROPERTIES = auto() 280 SET = auto() 281 SETTINGS = auto() 282 SHOW = auto() 283 SIMILAR_TO = auto() 284 SOME = auto() 285 STRUCT = auto() 286 TABLE_SAMPLE = auto() 287 TEMPORARY = auto() 288 TOP = auto() 289 THEN = auto() 290 TRUE = auto() 291 UNCACHE = auto() 292 UNION = auto() 293 UNNEST = auto() 294 UNPIVOT = auto() 295 UPDATE = auto() 296 USE = auto() 297 USING = auto() 298 VALUES = auto() 299 VIEW = auto() 300 VOLATILE = auto() 301 WHEN = auto() 302 WHERE = auto() 303 WINDOW = auto() 304 WITH = auto() 305 UNIQUE = auto()
An enumeration.
L_PAREN =
<TokenType.L_PAREN: 'L_PAREN'>
R_PAREN =
<TokenType.R_PAREN: 'R_PAREN'>
L_BRACKET =
<TokenType.L_BRACKET: 'L_BRACKET'>
R_BRACKET =
<TokenType.R_BRACKET: 'R_BRACKET'>
L_BRACE =
<TokenType.L_BRACE: 'L_BRACE'>
R_BRACE =
<TokenType.R_BRACE: 'R_BRACE'>
COMMA =
<TokenType.COMMA: 'COMMA'>
DOT =
<TokenType.DOT: 'DOT'>
DASH =
<TokenType.DASH: 'DASH'>
PLUS =
<TokenType.PLUS: 'PLUS'>
COLON =
<TokenType.COLON: 'COLON'>
DCOLON =
<TokenType.DCOLON: 'DCOLON'>
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'>
AND =
<TokenType.AND: 'AND'>
OR =
<TokenType.OR: 'OR'>
AMP =
<TokenType.AMP: 'AMP'>
DPIPE =
<TokenType.DPIPE: 'DPIPE'>
PIPE =
<TokenType.PIPE: 'PIPE'>
CARET =
<TokenType.CARET: 'CARET'>
TILDA =
<TokenType.TILDA: 'TILDA'>
ARROW =
<TokenType.ARROW: 'ARROW'>
DARROW =
<TokenType.DARROW: 'DARROW'>
FARROW =
<TokenType.FARROW: 'FARROW'>
HASH =
<TokenType.HASH: 'HASH'>
HASH_ARROW =
<TokenType.HASH_ARROW: 'HASH_ARROW'>
DHASH_ARROW =
<TokenType.DHASH_ARROW: 'DHASH_ARROW'>
LR_ARROW =
<TokenType.LR_ARROW: 'LR_ARROW'>
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'>
BLOCK_START =
<TokenType.BLOCK_START: 'BLOCK_START'>
BLOCK_END =
<TokenType.BLOCK_END: 'BLOCK_END'>
SPACE =
<TokenType.SPACE: 'SPACE'>
BREAK =
<TokenType.BREAK: 'BREAK'>
STRING =
<TokenType.STRING: 'STRING'>
NUMBER =
<TokenType.NUMBER: 'NUMBER'>
IDENTIFIER =
<TokenType.IDENTIFIER: 'IDENTIFIER'>
DATABASE =
<TokenType.DATABASE: 'DATABASE'>
COLUMN =
<TokenType.COLUMN: 'COLUMN'>
COLUMN_DEF =
<TokenType.COLUMN_DEF: 'COLUMN_DEF'>
SCHEMA =
<TokenType.SCHEMA: 'SCHEMA'>
TABLE =
<TokenType.TABLE: 'TABLE'>
VAR =
<TokenType.VAR: 'VAR'>
BIT_STRING =
<TokenType.BIT_STRING: 'BIT_STRING'>
HEX_STRING =
<TokenType.HEX_STRING: 'HEX_STRING'>
BYTE_STRING =
<TokenType.BYTE_STRING: 'BYTE_STRING'>
NATIONAL_STRING =
<TokenType.NATIONAL_STRING: 'NATIONAL_STRING'>
RAW_STRING =
<TokenType.RAW_STRING: 'RAW_STRING'>
BIT =
<TokenType.BIT: 'BIT'>
BOOLEAN =
<TokenType.BOOLEAN: 'BOOLEAN'>
TINYINT =
<TokenType.TINYINT: 'TINYINT'>
UTINYINT =
<TokenType.UTINYINT: 'UTINYINT'>
SMALLINT =
<TokenType.SMALLINT: 'SMALLINT'>
USMALLINT =
<TokenType.USMALLINT: 'USMALLINT'>
INT =
<TokenType.INT: 'INT'>
UINT =
<TokenType.UINT: 'UINT'>
BIGINT =
<TokenType.BIGINT: 'BIGINT'>
UBIGINT =
<TokenType.UBIGINT: 'UBIGINT'>
INT128 =
<TokenType.INT128: 'INT128'>
UINT128 =
<TokenType.UINT128: 'UINT128'>
INT256 =
<TokenType.INT256: 'INT256'>
UINT256 =
<TokenType.UINT256: 'UINT256'>
FLOAT =
<TokenType.FLOAT: 'FLOAT'>
DOUBLE =
<TokenType.DOUBLE: 'DOUBLE'>
DECIMAL =
<TokenType.DECIMAL: 'DECIMAL'>
BIGDECIMAL =
<TokenType.BIGDECIMAL: 'BIGDECIMAL'>
CHAR =
<TokenType.CHAR: 'CHAR'>
NCHAR =
<TokenType.NCHAR: 'NCHAR'>
VARCHAR =
<TokenType.VARCHAR: 'VARCHAR'>
NVARCHAR =
<TokenType.NVARCHAR: 'NVARCHAR'>
TEXT =
<TokenType.TEXT: 'TEXT'>
MEDIUMTEXT =
<TokenType.MEDIUMTEXT: 'MEDIUMTEXT'>
LONGTEXT =
<TokenType.LONGTEXT: 'LONGTEXT'>
MEDIUMBLOB =
<TokenType.MEDIUMBLOB: 'MEDIUMBLOB'>
LONGBLOB =
<TokenType.LONGBLOB: 'LONGBLOB'>
BINARY =
<TokenType.BINARY: 'BINARY'>
VARBINARY =
<TokenType.VARBINARY: 'VARBINARY'>
JSON =
<TokenType.JSON: 'JSON'>
JSONB =
<TokenType.JSONB: 'JSONB'>
TIME =
<TokenType.TIME: 'TIME'>
TIMESTAMP =
<TokenType.TIMESTAMP: 'TIMESTAMP'>
TIMESTAMPTZ =
<TokenType.TIMESTAMPTZ: 'TIMESTAMPTZ'>
TIMESTAMPLTZ =
<TokenType.TIMESTAMPLTZ: 'TIMESTAMPLTZ'>
DATETIME =
<TokenType.DATETIME: 'DATETIME'>
DATETIME64 =
<TokenType.DATETIME64: 'DATETIME64'>
DATE =
<TokenType.DATE: 'DATE'>
INT4RANGE =
<TokenType.INT4RANGE: 'INT4RANGE'>
INT4MULTIRANGE =
<TokenType.INT4MULTIRANGE: 'INT4MULTIRANGE'>
INT8RANGE =
<TokenType.INT8RANGE: 'INT8RANGE'>
INT8MULTIRANGE =
<TokenType.INT8MULTIRANGE: 'INT8MULTIRANGE'>
NUMRANGE =
<TokenType.NUMRANGE: 'NUMRANGE'>
NUMMULTIRANGE =
<TokenType.NUMMULTIRANGE: 'NUMMULTIRANGE'>
TSRANGE =
<TokenType.TSRANGE: 'TSRANGE'>
TSMULTIRANGE =
<TokenType.TSMULTIRANGE: 'TSMULTIRANGE'>
TSTZRANGE =
<TokenType.TSTZRANGE: 'TSTZRANGE'>
TSTZMULTIRANGE =
<TokenType.TSTZMULTIRANGE: 'TSTZMULTIRANGE'>
DATERANGE =
<TokenType.DATERANGE: 'DATERANGE'>
DATEMULTIRANGE =
<TokenType.DATEMULTIRANGE: 'DATEMULTIRANGE'>
UUID =
<TokenType.UUID: 'UUID'>
GEOGRAPHY =
<TokenType.GEOGRAPHY: 'GEOGRAPHY'>
NULLABLE =
<TokenType.NULLABLE: 'NULLABLE'>
GEOMETRY =
<TokenType.GEOMETRY: 'GEOMETRY'>
HLLSKETCH =
<TokenType.HLLSKETCH: 'HLLSKETCH'>
HSTORE =
<TokenType.HSTORE: 'HSTORE'>
SUPER =
<TokenType.SUPER: 'SUPER'>
SERIAL =
<TokenType.SERIAL: 'SERIAL'>
SMALLSERIAL =
<TokenType.SMALLSERIAL: 'SMALLSERIAL'>
BIGSERIAL =
<TokenType.BIGSERIAL: 'BIGSERIAL'>
XML =
<TokenType.XML: 'XML'>
UNIQUEIDENTIFIER =
<TokenType.UNIQUEIDENTIFIER: 'UNIQUEIDENTIFIER'>
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'>
ALIAS =
<TokenType.ALIAS: 'ALIAS'>
ALTER =
<TokenType.ALTER: 'ALTER'>
ALWAYS =
<TokenType.ALWAYS: 'ALWAYS'>
ALL =
<TokenType.ALL: 'ALL'>
ANTI =
<TokenType.ANTI: 'ANTI'>
ANY =
<TokenType.ANY: 'ANY'>
APPLY =
<TokenType.APPLY: 'APPLY'>
ARRAY =
<TokenType.ARRAY: 'ARRAY'>
ASC =
<TokenType.ASC: 'ASC'>
ASOF =
<TokenType.ASOF: 'ASOF'>
AUTO_INCREMENT =
<TokenType.AUTO_INCREMENT: 'AUTO_INCREMENT'>
BEGIN =
<TokenType.BEGIN: 'BEGIN'>
BETWEEN =
<TokenType.BETWEEN: 'BETWEEN'>
CACHE =
<TokenType.CACHE: 'CACHE'>
CASE =
<TokenType.CASE: 'CASE'>
CHARACTER_SET =
<TokenType.CHARACTER_SET: 'CHARACTER_SET'>
COLLATE =
<TokenType.COLLATE: 'COLLATE'>
COMMAND =
<TokenType.COMMAND: 'COMMAND'>
COMMENT =
<TokenType.COMMENT: 'COMMENT'>
COMMIT =
<TokenType.COMMIT: 'COMMIT'>
CONSTRAINT =
<TokenType.CONSTRAINT: 'CONSTRAINT'>
CREATE =
<TokenType.CREATE: 'CREATE'>
CROSS =
<TokenType.CROSS: 'CROSS'>
CUBE =
<TokenType.CUBE: 'CUBE'>
CURRENT_DATE =
<TokenType.CURRENT_DATE: 'CURRENT_DATE'>
CURRENT_DATETIME =
<TokenType.CURRENT_DATETIME: 'CURRENT_DATETIME'>
CURRENT_TIME =
<TokenType.CURRENT_TIME: 'CURRENT_TIME'>
CURRENT_TIMESTAMP =
<TokenType.CURRENT_TIMESTAMP: 'CURRENT_TIMESTAMP'>
CURRENT_USER =
<TokenType.CURRENT_USER: 'CURRENT_USER'>
DEFAULT =
<TokenType.DEFAULT: 'DEFAULT'>
DELETE =
<TokenType.DELETE: 'DELETE'>
DESC =
<TokenType.DESC: 'DESC'>
DESCRIBE =
<TokenType.DESCRIBE: 'DESCRIBE'>
DICTIONARY =
<TokenType.DICTIONARY: 'DICTIONARY'>
DISTINCT =
<TokenType.DISTINCT: 'DISTINCT'>
DIV =
<TokenType.DIV: 'DIV'>
DROP =
<TokenType.DROP: 'DROP'>
ELSE =
<TokenType.ELSE: 'ELSE'>
END =
<TokenType.END: 'END'>
ESCAPE =
<TokenType.ESCAPE: 'ESCAPE'>
EXCEPT =
<TokenType.EXCEPT: 'EXCEPT'>
EXECUTE =
<TokenType.EXECUTE: 'EXECUTE'>
EXISTS =
<TokenType.EXISTS: 'EXISTS'>
FALSE =
<TokenType.FALSE: 'FALSE'>
FETCH =
<TokenType.FETCH: 'FETCH'>
FILTER =
<TokenType.FILTER: 'FILTER'>
FINAL =
<TokenType.FINAL: 'FINAL'>
FIRST =
<TokenType.FIRST: 'FIRST'>
FOR =
<TokenType.FOR: 'FOR'>
FOREIGN_KEY =
<TokenType.FOREIGN_KEY: 'FOREIGN_KEY'>
FORMAT =
<TokenType.FORMAT: 'FORMAT'>
FROM =
<TokenType.FROM: 'FROM'>
FULL =
<TokenType.FULL: 'FULL'>
FUNCTION =
<TokenType.FUNCTION: 'FUNCTION'>
GLOB =
<TokenType.GLOB: 'GLOB'>
GLOBAL =
<TokenType.GLOBAL: 'GLOBAL'>
GROUP_BY =
<TokenType.GROUP_BY: 'GROUP_BY'>
GROUPING_SETS =
<TokenType.GROUPING_SETS: 'GROUPING_SETS'>
HAVING =
<TokenType.HAVING: 'HAVING'>
HINT =
<TokenType.HINT: 'HINT'>
IF =
<TokenType.IF: 'IF'>
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'>
LANGUAGE =
<TokenType.LANGUAGE: 'LANGUAGE'>
LATERAL =
<TokenType.LATERAL: 'LATERAL'>
LEFT =
<TokenType.LEFT: 'LEFT'>
LIKE =
<TokenType.LIKE: 'LIKE'>
LIKE_ANY =
<TokenType.LIKE_ANY: 'LIKE_ANY'>
LIMIT =
<TokenType.LIMIT: 'LIMIT'>
LOAD =
<TokenType.LOAD: 'LOAD'>
LOCK =
<TokenType.LOCK: 'LOCK'>
MAP =
<TokenType.MAP: 'MAP'>
MATCH_RECOGNIZE =
<TokenType.MATCH_RECOGNIZE: 'MATCH_RECOGNIZE'>
MERGE =
<TokenType.MERGE: 'MERGE'>
MOD =
<TokenType.MOD: 'MOD'>
NATURAL =
<TokenType.NATURAL: 'NATURAL'>
NEXT =
<TokenType.NEXT: 'NEXT'>
NEXT_VALUE_FOR =
<TokenType.NEXT_VALUE_FOR: 'NEXT_VALUE_FOR'>
NOTNULL =
<TokenType.NOTNULL: 'NOTNULL'>
NULL =
<TokenType.NULL: 'NULL'>
OFFSET =
<TokenType.OFFSET: 'OFFSET'>
ON =
<TokenType.ON: 'ON'>
ORDER_BY =
<TokenType.ORDER_BY: 'ORDER_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'>
PRAGMA =
<TokenType.PRAGMA: 'PRAGMA'>
PRIMARY_KEY =
<TokenType.PRIMARY_KEY: 'PRIMARY_KEY'>
PROCEDURE =
<TokenType.PROCEDURE: 'PROCEDURE'>
PROPERTIES =
<TokenType.PROPERTIES: 'PROPERTIES'>
PSEUDO_TYPE =
<TokenType.PSEUDO_TYPE: 'PSEUDO_TYPE'>
QUALIFY =
<TokenType.QUALIFY: 'QUALIFY'>
QUOTE =
<TokenType.QUOTE: 'QUOTE'>
RANGE =
<TokenType.RANGE: 'RANGE'>
RECURSIVE =
<TokenType.RECURSIVE: 'RECURSIVE'>
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'>
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'>
STRUCT =
<TokenType.STRUCT: 'STRUCT'>
TABLE_SAMPLE =
<TokenType.TABLE_SAMPLE: 'TABLE_SAMPLE'>
TEMPORARY =
<TokenType.TEMPORARY: 'TEMPORARY'>
TOP =
<TokenType.TOP: 'TOP'>
THEN =
<TokenType.THEN: 'THEN'>
TRUE =
<TokenType.TRUE: 'TRUE'>
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'>
Inherited Members
- enum.Enum
- name
- value
class
Token:
308class Token: 309 __slots__ = ("token_type", "text", "line", "col", "start", "end", "comments") 310 311 @classmethod 312 def number(cls, number: int) -> Token: 313 """Returns a NUMBER token with `number` as its text.""" 314 return cls(TokenType.NUMBER, str(number)) 315 316 @classmethod 317 def string(cls, string: str) -> Token: 318 """Returns a STRING token with `string` as its text.""" 319 return cls(TokenType.STRING, string) 320 321 @classmethod 322 def identifier(cls, identifier: str) -> Token: 323 """Returns an IDENTIFIER token with `identifier` as its text.""" 324 return cls(TokenType.IDENTIFIER, identifier) 325 326 @classmethod 327 def var(cls, var: str) -> Token: 328 """Returns an VAR token with `var` as its text.""" 329 return cls(TokenType.VAR, var) 330 331 def __init__( 332 self, 333 token_type: TokenType, 334 text: str, 335 line: int = 1, 336 col: int = 1, 337 start: int = 0, 338 end: int = 0, 339 comments: t.List[str] = [], 340 ) -> None: 341 """Token initializer. 342 343 Args: 344 token_type: The TokenType Enum. 345 text: The text of the token. 346 line: The line that the token ends on. 347 col: The column that the token ends on. 348 start: The start index of the token. 349 end: The ending index of the token. 350 comments: The comments to attach to the token. 351 """ 352 self.token_type = token_type 353 self.text = text 354 self.line = line 355 self.col = col 356 self.start = start 357 self.end = end 358 self.comments = comments 359 360 def __repr__(self) -> str: 361 attributes = ", ".join(f"{k}: {getattr(self, k)}" for k in self.__slots__) 362 return f"<Token {attributes}>"
Token( token_type: sqlglot.tokens.TokenType, text: str, line: int = 1, col: int = 1, start: int = 0, end: int = 0, comments: List[str] = [])
331 def __init__( 332 self, 333 token_type: TokenType, 334 text: str, 335 line: int = 1, 336 col: int = 1, 337 start: int = 0, 338 end: int = 0, 339 comments: t.List[str] = [], 340 ) -> None: 341 """Token initializer. 342 343 Args: 344 token_type: The TokenType Enum. 345 text: The text of the token. 346 line: The line that the token ends on. 347 col: The column that the token ends on. 348 start: The start index of the token. 349 end: The ending index of the token. 350 comments: The comments to attach to the token. 351 """ 352 self.token_type = token_type 353 self.text = text 354 self.line = line 355 self.col = col 356 self.start = start 357 self.end = end 358 self.comments = 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.
311 @classmethod 312 def number(cls, number: int) -> Token: 313 """Returns a NUMBER token with `number` as its text.""" 314 return cls(TokenType.NUMBER, str(number))
Returns a NUMBER token with number
as its text.
316 @classmethod 317 def string(cls, string: str) -> Token: 318 """Returns a STRING token with `string` as its text.""" 319 return cls(TokenType.STRING, string)
Returns a STRING token with string
as its text.
321 @classmethod 322 def identifier(cls, identifier: str) -> Token: 323 """Returns an IDENTIFIER token with `identifier` as its text.""" 324 return cls(TokenType.IDENTIFIER, identifier)
Returns an IDENTIFIER token with identifier
as its text.
class
Tokenizer:
415class Tokenizer(metaclass=_Tokenizer): 416 SINGLE_TOKENS = { 417 "(": TokenType.L_PAREN, 418 ")": TokenType.R_PAREN, 419 "[": TokenType.L_BRACKET, 420 "]": TokenType.R_BRACKET, 421 "{": TokenType.L_BRACE, 422 "}": TokenType.R_BRACE, 423 "&": TokenType.AMP, 424 "^": TokenType.CARET, 425 ":": TokenType.COLON, 426 ",": TokenType.COMMA, 427 ".": TokenType.DOT, 428 "-": TokenType.DASH, 429 "=": TokenType.EQ, 430 ">": TokenType.GT, 431 "<": TokenType.LT, 432 "%": TokenType.MOD, 433 "!": TokenType.NOT, 434 "|": TokenType.PIPE, 435 "+": TokenType.PLUS, 436 ";": TokenType.SEMICOLON, 437 "/": TokenType.SLASH, 438 "\\": TokenType.BACKSLASH, 439 "*": TokenType.STAR, 440 "~": TokenType.TILDA, 441 "?": TokenType.PLACEHOLDER, 442 "@": TokenType.PARAMETER, 443 # used for breaking a var like x'y' but nothing else 444 # the token type doesn't matter 445 "'": TokenType.QUOTE, 446 "`": TokenType.IDENTIFIER, 447 '"': TokenType.IDENTIFIER, 448 "#": TokenType.HASH, 449 } 450 451 BIT_STRINGS: t.List[str | t.Tuple[str, str]] = [] 452 BYTE_STRINGS: t.List[str | t.Tuple[str, str]] = [] 453 HEX_STRINGS: t.List[str | t.Tuple[str, str]] = [] 454 RAW_STRINGS: t.List[str | t.Tuple[str, str]] = [] 455 IDENTIFIERS: t.List[str | t.Tuple[str, str]] = ['"'] 456 IDENTIFIER_ESCAPES = ['"'] 457 QUOTES: t.List[t.Tuple[str, str] | str] = ["'"] 458 STRING_ESCAPES = ["'"] 459 VAR_SINGLE_TOKENS: t.Set[str] = set() 460 461 # Autofilled 462 IDENTIFIERS_CAN_START_WITH_DIGIT: bool = False 463 464 _COMMENTS: t.Dict[str, str] = {} 465 _FORMAT_STRINGS: t.Dict[str, t.Tuple[str, TokenType]] = {} 466 _IDENTIFIERS: t.Dict[str, str] = {} 467 _IDENTIFIER_ESCAPES: t.Set[str] = set() 468 _QUOTES: t.Dict[str, str] = {} 469 _STRING_ESCAPES: t.Set[str] = set() 470 _KEYWORD_TRIE: t.Dict = {} 471 472 KEYWORDS: t.Dict[str, TokenType] = { 473 **{f"{{%{postfix}": TokenType.BLOCK_START for postfix in ("", "+", "-")}, 474 **{f"{prefix}%}}": TokenType.BLOCK_END for prefix in ("", "+", "-")}, 475 **{f"{{{{{postfix}": TokenType.BLOCK_START for postfix in ("+", "-")}, 476 **{f"{prefix}}}}}": TokenType.BLOCK_END for prefix in ("+", "-")}, 477 "/*+": TokenType.HINT, 478 "==": TokenType.EQ, 479 "::": TokenType.DCOLON, 480 "||": TokenType.DPIPE, 481 ">=": TokenType.GTE, 482 "<=": TokenType.LTE, 483 "<>": TokenType.NEQ, 484 "!=": TokenType.NEQ, 485 "<=>": TokenType.NULLSAFE_EQ, 486 "->": TokenType.ARROW, 487 "->>": TokenType.DARROW, 488 "=>": TokenType.FARROW, 489 "#>": TokenType.HASH_ARROW, 490 "#>>": TokenType.DHASH_ARROW, 491 "<->": TokenType.LR_ARROW, 492 "&&": TokenType.DAMP, 493 "ALL": TokenType.ALL, 494 "ALWAYS": TokenType.ALWAYS, 495 "AND": TokenType.AND, 496 "ANTI": TokenType.ANTI, 497 "ANY": TokenType.ANY, 498 "ASC": TokenType.ASC, 499 "AS": TokenType.ALIAS, 500 "ASOF": TokenType.ASOF, 501 "AUTOINCREMENT": TokenType.AUTO_INCREMENT, 502 "AUTO_INCREMENT": TokenType.AUTO_INCREMENT, 503 "BEGIN": TokenType.BEGIN, 504 "BETWEEN": TokenType.BETWEEN, 505 "CACHE": TokenType.CACHE, 506 "UNCACHE": TokenType.UNCACHE, 507 "CASE": TokenType.CASE, 508 "CHARACTER SET": TokenType.CHARACTER_SET, 509 "COLLATE": TokenType.COLLATE, 510 "COLUMN": TokenType.COLUMN, 511 "COMMIT": TokenType.COMMIT, 512 "CONSTRAINT": TokenType.CONSTRAINT, 513 "CREATE": TokenType.CREATE, 514 "CROSS": TokenType.CROSS, 515 "CUBE": TokenType.CUBE, 516 "CURRENT_DATE": TokenType.CURRENT_DATE, 517 "CURRENT_TIME": TokenType.CURRENT_TIME, 518 "CURRENT_TIMESTAMP": TokenType.CURRENT_TIMESTAMP, 519 "CURRENT_USER": TokenType.CURRENT_USER, 520 "DATABASE": TokenType.DATABASE, 521 "DEFAULT": TokenType.DEFAULT, 522 "DELETE": TokenType.DELETE, 523 "DESC": TokenType.DESC, 524 "DESCRIBE": TokenType.DESCRIBE, 525 "DISTINCT": TokenType.DISTINCT, 526 "DIV": TokenType.DIV, 527 "DROP": TokenType.DROP, 528 "ELSE": TokenType.ELSE, 529 "END": TokenType.END, 530 "ESCAPE": TokenType.ESCAPE, 531 "EXCEPT": TokenType.EXCEPT, 532 "EXECUTE": TokenType.EXECUTE, 533 "EXISTS": TokenType.EXISTS, 534 "FALSE": TokenType.FALSE, 535 "FETCH": TokenType.FETCH, 536 "FILTER": TokenType.FILTER, 537 "FIRST": TokenType.FIRST, 538 "FULL": TokenType.FULL, 539 "FUNCTION": TokenType.FUNCTION, 540 "FOR": TokenType.FOR, 541 "FOREIGN KEY": TokenType.FOREIGN_KEY, 542 "FORMAT": TokenType.FORMAT, 543 "FROM": TokenType.FROM, 544 "GEOGRAPHY": TokenType.GEOGRAPHY, 545 "GEOMETRY": TokenType.GEOMETRY, 546 "GLOB": TokenType.GLOB, 547 "GROUP BY": TokenType.GROUP_BY, 548 "GROUPING SETS": TokenType.GROUPING_SETS, 549 "HAVING": TokenType.HAVING, 550 "IF": TokenType.IF, 551 "ILIKE": TokenType.ILIKE, 552 "IN": TokenType.IN, 553 "INDEX": TokenType.INDEX, 554 "INET": TokenType.INET, 555 "INNER": TokenType.INNER, 556 "INSERT": TokenType.INSERT, 557 "INTERVAL": TokenType.INTERVAL, 558 "INTERSECT": TokenType.INTERSECT, 559 "INTO": TokenType.INTO, 560 "IS": TokenType.IS, 561 "ISNULL": TokenType.ISNULL, 562 "JOIN": TokenType.JOIN, 563 "KEEP": TokenType.KEEP, 564 "LATERAL": TokenType.LATERAL, 565 "LEFT": TokenType.LEFT, 566 "LIKE": TokenType.LIKE, 567 "LIMIT": TokenType.LIMIT, 568 "LOAD": TokenType.LOAD, 569 "LOCK": TokenType.LOCK, 570 "MERGE": TokenType.MERGE, 571 "NATURAL": TokenType.NATURAL, 572 "NEXT": TokenType.NEXT, 573 "NEXT VALUE FOR": TokenType.NEXT_VALUE_FOR, 574 "NOT": TokenType.NOT, 575 "NOTNULL": TokenType.NOTNULL, 576 "NULL": TokenType.NULL, 577 "OBJECT": TokenType.OBJECT, 578 "OFFSET": TokenType.OFFSET, 579 "ON": TokenType.ON, 580 "OR": TokenType.OR, 581 "ORDER BY": TokenType.ORDER_BY, 582 "ORDINALITY": TokenType.ORDINALITY, 583 "OUTER": TokenType.OUTER, 584 "OVER": TokenType.OVER, 585 "OVERLAPS": TokenType.OVERLAPS, 586 "OVERWRITE": TokenType.OVERWRITE, 587 "PARTITION": TokenType.PARTITION, 588 "PARTITION BY": TokenType.PARTITION_BY, 589 "PARTITIONED BY": TokenType.PARTITION_BY, 590 "PARTITIONED_BY": TokenType.PARTITION_BY, 591 "PERCENT": TokenType.PERCENT, 592 "PIVOT": TokenType.PIVOT, 593 "PRAGMA": TokenType.PRAGMA, 594 "PRIMARY KEY": TokenType.PRIMARY_KEY, 595 "PROCEDURE": TokenType.PROCEDURE, 596 "QUALIFY": TokenType.QUALIFY, 597 "RANGE": TokenType.RANGE, 598 "RECURSIVE": TokenType.RECURSIVE, 599 "REGEXP": TokenType.RLIKE, 600 "REPLACE": TokenType.REPLACE, 601 "REFERENCES": TokenType.REFERENCES, 602 "RIGHT": TokenType.RIGHT, 603 "RLIKE": TokenType.RLIKE, 604 "ROLLBACK": TokenType.ROLLBACK, 605 "ROLLUP": TokenType.ROLLUP, 606 "ROW": TokenType.ROW, 607 "ROWS": TokenType.ROWS, 608 "SCHEMA": TokenType.SCHEMA, 609 "SELECT": TokenType.SELECT, 610 "SEMI": TokenType.SEMI, 611 "SET": TokenType.SET, 612 "SETTINGS": TokenType.SETTINGS, 613 "SHOW": TokenType.SHOW, 614 "SIMILAR TO": TokenType.SIMILAR_TO, 615 "SOME": TokenType.SOME, 616 "TABLE": TokenType.TABLE, 617 "TABLESAMPLE": TokenType.TABLE_SAMPLE, 618 "TEMP": TokenType.TEMPORARY, 619 "TEMPORARY": TokenType.TEMPORARY, 620 "THEN": TokenType.THEN, 621 "TRUE": TokenType.TRUE, 622 "UNION": TokenType.UNION, 623 "UNNEST": TokenType.UNNEST, 624 "UNPIVOT": TokenType.UNPIVOT, 625 "UPDATE": TokenType.UPDATE, 626 "USE": TokenType.USE, 627 "USING": TokenType.USING, 628 "UUID": TokenType.UUID, 629 "VALUES": TokenType.VALUES, 630 "VIEW": TokenType.VIEW, 631 "VOLATILE": TokenType.VOLATILE, 632 "WHEN": TokenType.WHEN, 633 "WHERE": TokenType.WHERE, 634 "WINDOW": TokenType.WINDOW, 635 "WITH": TokenType.WITH, 636 "APPLY": TokenType.APPLY, 637 "ARRAY": TokenType.ARRAY, 638 "BIT": TokenType.BIT, 639 "BOOL": TokenType.BOOLEAN, 640 "BOOLEAN": TokenType.BOOLEAN, 641 "BYTE": TokenType.TINYINT, 642 "TINYINT": TokenType.TINYINT, 643 "SHORT": TokenType.SMALLINT, 644 "SMALLINT": TokenType.SMALLINT, 645 "INT2": TokenType.SMALLINT, 646 "INTEGER": TokenType.INT, 647 "INT": TokenType.INT, 648 "INT4": TokenType.INT, 649 "LONG": TokenType.BIGINT, 650 "BIGINT": TokenType.BIGINT, 651 "INT8": TokenType.BIGINT, 652 "DEC": TokenType.DECIMAL, 653 "DECIMAL": TokenType.DECIMAL, 654 "BIGDECIMAL": TokenType.BIGDECIMAL, 655 "BIGNUMERIC": TokenType.BIGDECIMAL, 656 "MAP": TokenType.MAP, 657 "NULLABLE": TokenType.NULLABLE, 658 "NUMBER": TokenType.DECIMAL, 659 "NUMERIC": TokenType.DECIMAL, 660 "FIXED": TokenType.DECIMAL, 661 "REAL": TokenType.FLOAT, 662 "FLOAT": TokenType.FLOAT, 663 "FLOAT4": TokenType.FLOAT, 664 "FLOAT8": TokenType.DOUBLE, 665 "DOUBLE": TokenType.DOUBLE, 666 "DOUBLE PRECISION": TokenType.DOUBLE, 667 "JSON": TokenType.JSON, 668 "CHAR": TokenType.CHAR, 669 "CHARACTER": TokenType.CHAR, 670 "NCHAR": TokenType.NCHAR, 671 "VARCHAR": TokenType.VARCHAR, 672 "VARCHAR2": TokenType.VARCHAR, 673 "NVARCHAR": TokenType.NVARCHAR, 674 "NVARCHAR2": TokenType.NVARCHAR, 675 "STR": TokenType.TEXT, 676 "STRING": TokenType.TEXT, 677 "TEXT": TokenType.TEXT, 678 "CLOB": TokenType.TEXT, 679 "LONGVARCHAR": TokenType.TEXT, 680 "BINARY": TokenType.BINARY, 681 "BLOB": TokenType.VARBINARY, 682 "BYTEA": TokenType.VARBINARY, 683 "VARBINARY": TokenType.VARBINARY, 684 "TIME": TokenType.TIME, 685 "TIMESTAMP": TokenType.TIMESTAMP, 686 "TIMESTAMPTZ": TokenType.TIMESTAMPTZ, 687 "TIMESTAMPLTZ": TokenType.TIMESTAMPLTZ, 688 "DATE": TokenType.DATE, 689 "DATETIME": TokenType.DATETIME, 690 "INT4RANGE": TokenType.INT4RANGE, 691 "INT4MULTIRANGE": TokenType.INT4MULTIRANGE, 692 "INT8RANGE": TokenType.INT8RANGE, 693 "INT8MULTIRANGE": TokenType.INT8MULTIRANGE, 694 "NUMRANGE": TokenType.NUMRANGE, 695 "NUMMULTIRANGE": TokenType.NUMMULTIRANGE, 696 "TSRANGE": TokenType.TSRANGE, 697 "TSMULTIRANGE": TokenType.TSMULTIRANGE, 698 "TSTZRANGE": TokenType.TSTZRANGE, 699 "TSTZMULTIRANGE": TokenType.TSTZMULTIRANGE, 700 "DATERANGE": TokenType.DATERANGE, 701 "DATEMULTIRANGE": TokenType.DATEMULTIRANGE, 702 "UNIQUE": TokenType.UNIQUE, 703 "STRUCT": TokenType.STRUCT, 704 "VARIANT": TokenType.VARIANT, 705 "ALTER": TokenType.ALTER, 706 "ANALYZE": TokenType.COMMAND, 707 "CALL": TokenType.COMMAND, 708 "COMMENT": TokenType.COMMENT, 709 "COPY": TokenType.COMMAND, 710 "EXPLAIN": TokenType.COMMAND, 711 "GRANT": TokenType.COMMAND, 712 "OPTIMIZE": TokenType.COMMAND, 713 "PREPARE": TokenType.COMMAND, 714 "TRUNCATE": TokenType.COMMAND, 715 "VACUUM": TokenType.COMMAND, 716 } 717 718 WHITE_SPACE: t.Dict[t.Optional[str], TokenType] = { 719 " ": TokenType.SPACE, 720 "\t": TokenType.SPACE, 721 "\n": TokenType.BREAK, 722 "\r": TokenType.BREAK, 723 "\r\n": TokenType.BREAK, 724 } 725 726 COMMANDS = { 727 TokenType.COMMAND, 728 TokenType.EXECUTE, 729 TokenType.FETCH, 730 TokenType.SHOW, 731 } 732 733 COMMAND_PREFIX_TOKENS = {TokenType.SEMICOLON, TokenType.BEGIN} 734 735 # handle numeric literals like in hive (3L = BIGINT) 736 NUMERIC_LITERALS: t.Dict[str, str] = {} 737 ENCODE: t.Optional[str] = None 738 739 COMMENTS = ["--", ("/*", "*/"), ("{#", "#}")] 740 741 __slots__ = ( 742 "sql", 743 "size", 744 "tokens", 745 "_start", 746 "_current", 747 "_line", 748 "_col", 749 "_comments", 750 "_char", 751 "_end", 752 "_peek", 753 "_prev_token_line", 754 ) 755 756 def __init__(self) -> None: 757 self.reset() 758 759 def reset(self) -> None: 760 self.sql = "" 761 self.size = 0 762 self.tokens: t.List[Token] = [] 763 self._start = 0 764 self._current = 0 765 self._line = 1 766 self._col = 0 767 self._comments: t.List[str] = [] 768 769 self._char = "" 770 self._end = False 771 self._peek = "" 772 self._prev_token_line = -1 773 774 def tokenize(self, sql: str) -> t.List[Token]: 775 """Returns a list of tokens corresponding to the SQL string `sql`.""" 776 self.reset() 777 self.sql = sql 778 self.size = len(sql) 779 780 try: 781 self._scan() 782 except Exception as e: 783 start = max(self._current - 50, 0) 784 end = min(self._current + 50, self.size - 1) 785 context = self.sql[start:end] 786 raise ValueError(f"Error tokenizing '{context}'") from e 787 788 return self.tokens 789 790 def _scan(self, until: t.Optional[t.Callable] = None) -> None: 791 while self.size and not self._end: 792 self._start = self._current 793 self._advance() 794 795 if self._char is None: 796 break 797 798 if self._char not in self.WHITE_SPACE: 799 if self._char.isdigit(): 800 self._scan_number() 801 elif self._char in self._IDENTIFIERS: 802 self._scan_identifier(self._IDENTIFIERS[self._char]) 803 else: 804 self._scan_keywords() 805 806 if until and until(): 807 break 808 809 if self.tokens and self._comments: 810 self.tokens[-1].comments.extend(self._comments) 811 812 def _chars(self, size: int) -> str: 813 if size == 1: 814 return self._char 815 816 start = self._current - 1 817 end = start + size 818 819 return self.sql[start:end] if end <= self.size else "" 820 821 def _advance(self, i: int = 1, alnum: bool = False) -> None: 822 if self.WHITE_SPACE.get(self._char) is TokenType.BREAK: 823 self._col = 1 824 self._line += 1 825 else: 826 self._col += i 827 828 self._current += i 829 self._end = self._current >= self.size 830 self._char = self.sql[self._current - 1] 831 self._peek = "" if self._end else self.sql[self._current] 832 833 if alnum and self._char.isalnum(): 834 # Here we use local variables instead of attributes for better performance 835 _col = self._col 836 _current = self._current 837 _end = self._end 838 _peek = self._peek 839 840 while _peek.isalnum(): 841 _col += 1 842 _current += 1 843 _end = _current >= self.size 844 _peek = "" if _end else self.sql[_current] 845 846 self._col = _col 847 self._current = _current 848 self._end = _end 849 self._peek = _peek 850 self._char = self.sql[_current - 1] 851 852 @property 853 def _text(self) -> str: 854 return self.sql[self._start : self._current] 855 856 def peek(self, i: int = 0) -> str: 857 i = self._current + i 858 if i < self.size: 859 return self.sql[i] 860 return "" 861 862 def _add(self, token_type: TokenType, text: t.Optional[str] = None) -> None: 863 self._prev_token_line = self._line 864 self.tokens.append( 865 Token( 866 token_type, 867 text=self._text if text is None else text, 868 line=self._line, 869 col=self._col, 870 start=self._start, 871 end=self._current - 1, 872 comments=self._comments, 873 ) 874 ) 875 self._comments = [] 876 877 # If we have either a semicolon or a begin token before the command's token, we'll parse 878 # whatever follows the command's token as a string 879 if ( 880 token_type in self.COMMANDS 881 and self._peek != ";" 882 and (len(self.tokens) == 1 or self.tokens[-2].token_type in self.COMMAND_PREFIX_TOKENS) 883 ): 884 start = self._current 885 tokens = len(self.tokens) 886 self._scan(lambda: self._peek == ";") 887 self.tokens = self.tokens[:tokens] 888 text = self.sql[start : self._current].strip() 889 if text: 890 self._add(TokenType.STRING, text) 891 892 def _scan_keywords(self) -> None: 893 size = 0 894 word = None 895 chars = self._text 896 char = chars 897 prev_space = False 898 skip = False 899 trie = self._KEYWORD_TRIE 900 single_token = char in self.SINGLE_TOKENS 901 902 while chars: 903 if skip: 904 result = 1 905 else: 906 result, trie = in_trie(trie, char.upper()) 907 908 if result == 0: 909 break 910 if result == 2: 911 word = chars 912 913 size += 1 914 end = self._current - 1 + size 915 916 if end < self.size: 917 char = self.sql[end] 918 single_token = single_token or char in self.SINGLE_TOKENS 919 is_space = char in self.WHITE_SPACE 920 921 if not is_space or not prev_space: 922 if is_space: 923 char = " " 924 chars += char 925 prev_space = is_space 926 skip = False 927 else: 928 skip = True 929 else: 930 char = "" 931 chars = " " 932 933 word = None if not single_token and chars[-1] not in self.WHITE_SPACE else word 934 935 if not word: 936 if self._char in self.SINGLE_TOKENS: 937 self._add(self.SINGLE_TOKENS[self._char], text=self._char) 938 return 939 self._scan_var() 940 return 941 942 if self._scan_string(word): 943 return 944 if self._scan_comment(word): 945 return 946 947 self._advance(size - 1) 948 word = word.upper() 949 self._add(self.KEYWORDS[word], text=word) 950 951 def _scan_comment(self, comment_start: str) -> bool: 952 if comment_start not in self._COMMENTS: 953 return False 954 955 comment_start_line = self._line 956 comment_start_size = len(comment_start) 957 comment_end = self._COMMENTS[comment_start] 958 959 if comment_end: 960 # Skip the comment's start delimiter 961 self._advance(comment_start_size) 962 963 comment_end_size = len(comment_end) 964 while not self._end and self._chars(comment_end_size) != comment_end: 965 self._advance(alnum=True) 966 967 self._comments.append(self._text[comment_start_size : -comment_end_size + 1]) 968 self._advance(comment_end_size - 1) 969 else: 970 while not self._end and not self.WHITE_SPACE.get(self._peek) is TokenType.BREAK: 971 self._advance(alnum=True) 972 self._comments.append(self._text[comment_start_size:]) 973 974 # Leading comment is attached to the succeeding token, whilst trailing comment to the preceding. 975 # Multiple consecutive comments are preserved by appending them to the current comments list. 976 if comment_start_line == self._prev_token_line: 977 self.tokens[-1].comments.extend(self._comments) 978 self._comments = [] 979 self._prev_token_line = self._line 980 981 return True 982 983 def _scan_number(self) -> None: 984 if self._char == "0": 985 peek = self._peek.upper() 986 if peek == "B": 987 return self._scan_bits() if self.BIT_STRINGS else self._add(TokenType.NUMBER) 988 elif peek == "X": 989 return self._scan_hex() if self.HEX_STRINGS else self._add(TokenType.NUMBER) 990 991 decimal = False 992 scientific = 0 993 994 while True: 995 if self._peek.isdigit(): 996 self._advance() 997 elif self._peek == "." and not decimal: 998 after = self.peek(1) 999 if after.isdigit() or not after.strip(): 1000 decimal = True 1001 self._advance() 1002 else: 1003 return self._add(TokenType.VAR) 1004 elif self._peek in ("-", "+") and scientific == 1: 1005 scientific += 1 1006 self._advance() 1007 elif self._peek.upper() == "E" and not scientific: 1008 scientific += 1 1009 self._advance() 1010 elif self._peek.isidentifier(): 1011 number_text = self._text 1012 literal = "" 1013 1014 while self._peek.strip() and self._peek not in self.SINGLE_TOKENS: 1015 literal += self._peek.upper() 1016 self._advance() 1017 1018 token_type = self.KEYWORDS.get(self.NUMERIC_LITERALS.get(literal, "")) 1019 1020 if token_type: 1021 self._add(TokenType.NUMBER, number_text) 1022 self._add(TokenType.DCOLON, "::") 1023 return self._add(token_type, literal) 1024 elif self.IDENTIFIERS_CAN_START_WITH_DIGIT: 1025 return self._add(TokenType.VAR) 1026 1027 self._add(TokenType.NUMBER, number_text) 1028 return self._advance(-len(literal)) 1029 else: 1030 return self._add(TokenType.NUMBER) 1031 1032 def _scan_bits(self) -> None: 1033 self._advance() 1034 value = self._extract_value() 1035 try: 1036 # If `value` can't be converted to a binary, fallback to tokenizing it as an identifier 1037 int(value, 2) 1038 self._add(TokenType.BIT_STRING, value[2:]) # Drop the 0b 1039 except ValueError: 1040 self._add(TokenType.IDENTIFIER) 1041 1042 def _scan_hex(self) -> None: 1043 self._advance() 1044 value = self._extract_value() 1045 try: 1046 # If `value` can't be converted to a hex, fallback to tokenizing it as an identifier 1047 int(value, 16) 1048 self._add(TokenType.HEX_STRING, value[2:]) # Drop the 0x 1049 except ValueError: 1050 self._add(TokenType.IDENTIFIER) 1051 1052 def _extract_value(self) -> str: 1053 while True: 1054 char = self._peek.strip() 1055 if char and char not in self.SINGLE_TOKENS: 1056 self._advance(alnum=True) 1057 else: 1058 break 1059 1060 return self._text 1061 1062 def _scan_string(self, start: str) -> bool: 1063 base = None 1064 token_type = TokenType.STRING 1065 1066 if start in self._QUOTES: 1067 end = self._QUOTES[start] 1068 elif start in self._FORMAT_STRINGS: 1069 end, token_type = self._FORMAT_STRINGS[start] 1070 1071 if token_type == TokenType.HEX_STRING: 1072 base = 16 1073 elif token_type == TokenType.BIT_STRING: 1074 base = 2 1075 else: 1076 return False 1077 1078 self._advance(len(start)) 1079 text = self._extract_string(end) 1080 1081 if base: 1082 try: 1083 int(text, base) 1084 except: 1085 raise RuntimeError( 1086 f"Numeric string contains invalid characters from {self._line}:{self._start}" 1087 ) 1088 else: 1089 text = text.encode(self.ENCODE).decode(self.ENCODE) if self.ENCODE else text 1090 1091 self._add(token_type, text) 1092 return True 1093 1094 def _scan_identifier(self, identifier_end: str) -> None: 1095 self._advance() 1096 text = self._extract_string(identifier_end, self._IDENTIFIER_ESCAPES) 1097 self._add(TokenType.IDENTIFIER, text) 1098 1099 def _scan_var(self) -> None: 1100 while True: 1101 char = self._peek.strip() 1102 if char and (char in self.VAR_SINGLE_TOKENS or char not in self.SINGLE_TOKENS): 1103 self._advance(alnum=True) 1104 else: 1105 break 1106 1107 self._add( 1108 TokenType.VAR 1109 if self.tokens and self.tokens[-1].token_type == TokenType.PARAMETER 1110 else self.KEYWORDS.get(self._text.upper(), TokenType.VAR) 1111 ) 1112 1113 def _extract_string(self, delimiter: str, escapes=None) -> str: 1114 text = "" 1115 delim_size = len(delimiter) 1116 escapes = self._STRING_ESCAPES if escapes is None else escapes 1117 1118 while True: 1119 if self._char in escapes and (self._peek == delimiter or self._peek in escapes): 1120 if self._peek == delimiter: 1121 text += self._peek 1122 else: 1123 text += self._char + self._peek 1124 1125 if self._current + 1 < self.size: 1126 self._advance(2) 1127 else: 1128 raise RuntimeError(f"Missing {delimiter} from {self._line}:{self._current}") 1129 else: 1130 if self._chars(delim_size) == delimiter: 1131 if delim_size > 1: 1132 self._advance(delim_size - 1) 1133 break 1134 1135 if self._end: 1136 raise RuntimeError(f"Missing {delimiter} from {self._line}:{self._start}") 1137 1138 current = self._current - 1 1139 self._advance(alnum=True) 1140 text += self.sql[current : self._current - 1] 1141 1142 return text
def
reset(self) -> None:
759 def reset(self) -> None: 760 self.sql = "" 761 self.size = 0 762 self.tokens: t.List[Token] = [] 763 self._start = 0 764 self._current = 0 765 self._line = 1 766 self._col = 0 767 self._comments: t.List[str] = [] 768 769 self._char = "" 770 self._end = False 771 self._peek = "" 772 self._prev_token_line = -1
774 def tokenize(self, sql: str) -> t.List[Token]: 775 """Returns a list of tokens corresponding to the SQL string `sql`.""" 776 self.reset() 777 self.sql = sql 778 self.size = len(sql) 779 780 try: 781 self._scan() 782 except Exception as e: 783 start = max(self._current - 50, 0) 784 end = min(self._current + 50, self.size - 1) 785 context = self.sql[start:end] 786 raise ValueError(f"Error tokenizing '{context}'") from e 787 788 return self.tokens
Returns a list of tokens corresponding to the SQL string sql
.