sqlglot.dialects.presto
1from __future__ import annotations 2 3import typing as t 4 5from sqlglot import exp, generator, parser, tokens, transforms 6from sqlglot.dialects.dialect import ( 7 Dialect, 8 NormalizationStrategy, 9 binary_from_function, 10 bool_xor_sql, 11 date_trunc_to_time, 12 datestrtodate_sql, 13 encode_decode_sql, 14 build_formatted_time, 15 if_sql, 16 left_to_substring_sql, 17 no_ilike_sql, 18 no_pivot_sql, 19 no_safe_divide_sql, 20 no_timestamp_sql, 21 regexp_extract_sql, 22 rename_func, 23 right_to_substring_sql, 24 struct_extract_sql, 25 str_position_sql, 26 timestamptrunc_sql, 27 timestrtotime_sql, 28 ts_or_ds_add_cast, 29 unit_to_str, 30) 31from sqlglot.dialects.hive import Hive 32from sqlglot.dialects.mysql import MySQL 33from sqlglot.helper import apply_index_offset, seq_get 34from sqlglot.tokens import TokenType 35 36 37def _explode_to_unnest_sql(self: Presto.Generator, expression: exp.Lateral) -> str: 38 if isinstance(expression.this, exp.Explode): 39 return self.sql( 40 exp.Join( 41 this=exp.Unnest( 42 expressions=[expression.this.this], 43 alias=expression.args.get("alias"), 44 offset=isinstance(expression.this, exp.Posexplode), 45 ), 46 kind="cross", 47 ) 48 ) 49 return self.lateral_sql(expression) 50 51 52def _initcap_sql(self: Presto.Generator, expression: exp.Initcap) -> str: 53 regex = r"(\w)(\w*)" 54 return f"REGEXP_REPLACE({self.sql(expression, 'this')}, '{regex}', x -> UPPER(x[1]) || LOWER(x[2]))" 55 56 57def _no_sort_array(self: Presto.Generator, expression: exp.SortArray) -> str: 58 if expression.args.get("asc") == exp.false(): 59 comparator = "(a, b) -> CASE WHEN a < b THEN 1 WHEN a > b THEN -1 ELSE 0 END" 60 else: 61 comparator = None 62 return self.func("ARRAY_SORT", expression.this, comparator) 63 64 65def _schema_sql(self: Presto.Generator, expression: exp.Schema) -> str: 66 if isinstance(expression.parent, exp.Property): 67 columns = ", ".join(f"'{c.name}'" for c in expression.expressions) 68 return f"ARRAY[{columns}]" 69 70 if expression.parent: 71 for schema in expression.parent.find_all(exp.Schema): 72 column_defs = schema.find_all(exp.ColumnDef) 73 if column_defs and isinstance(schema.parent, exp.Property): 74 expression.expressions.extend(column_defs) 75 76 return self.schema_sql(expression) 77 78 79def _quantile_sql(self: Presto.Generator, expression: exp.Quantile) -> str: 80 self.unsupported("Presto does not support exact quantiles") 81 return self.func("APPROX_PERCENTILE", expression.this, expression.args.get("quantile")) 82 83 84def _str_to_time_sql( 85 self: Presto.Generator, expression: exp.StrToDate | exp.StrToTime | exp.TsOrDsToDate 86) -> str: 87 return self.func("DATE_PARSE", expression.this, self.format_time(expression)) 88 89 90def _ts_or_ds_to_date_sql(self: Presto.Generator, expression: exp.TsOrDsToDate) -> str: 91 time_format = self.format_time(expression) 92 if time_format and time_format not in (Presto.TIME_FORMAT, Presto.DATE_FORMAT): 93 return self.sql(exp.cast(_str_to_time_sql(self, expression), "DATE")) 94 return self.sql(exp.cast(exp.cast(expression.this, "TIMESTAMP"), "DATE")) 95 96 97def _ts_or_ds_add_sql(self: Presto.Generator, expression: exp.TsOrDsAdd) -> str: 98 expression = ts_or_ds_add_cast(expression) 99 unit = unit_to_str(expression) 100 return self.func("DATE_ADD", unit, expression.expression, expression.this) 101 102 103def _ts_or_ds_diff_sql(self: Presto.Generator, expression: exp.TsOrDsDiff) -> str: 104 this = exp.cast(expression.this, "TIMESTAMP") 105 expr = exp.cast(expression.expression, "TIMESTAMP") 106 unit = unit_to_str(expression) 107 return self.func("DATE_DIFF", unit, expr, this) 108 109 110def _build_approx_percentile(args: t.List) -> exp.Expression: 111 if len(args) == 4: 112 return exp.ApproxQuantile( 113 this=seq_get(args, 0), 114 weight=seq_get(args, 1), 115 quantile=seq_get(args, 2), 116 accuracy=seq_get(args, 3), 117 ) 118 if len(args) == 3: 119 return exp.ApproxQuantile( 120 this=seq_get(args, 0), quantile=seq_get(args, 1), accuracy=seq_get(args, 2) 121 ) 122 return exp.ApproxQuantile.from_arg_list(args) 123 124 125def _build_from_unixtime(args: t.List) -> exp.Expression: 126 if len(args) == 3: 127 return exp.UnixToTime( 128 this=seq_get(args, 0), 129 hours=seq_get(args, 1), 130 minutes=seq_get(args, 2), 131 ) 132 if len(args) == 2: 133 return exp.UnixToTime(this=seq_get(args, 0), zone=seq_get(args, 1)) 134 135 return exp.UnixToTime.from_arg_list(args) 136 137 138def _unnest_sequence(expression: exp.Expression) -> exp.Expression: 139 if isinstance(expression, exp.Table): 140 if isinstance(expression.this, exp.GenerateSeries): 141 unnest = exp.Unnest(expressions=[expression.this]) 142 143 if expression.alias: 144 return exp.alias_(unnest, alias="_u", table=[expression.alias], copy=False) 145 return unnest 146 return expression 147 148 149def _first_last_sql(self: Presto.Generator, expression: exp.Func) -> str: 150 """ 151 Trino doesn't support FIRST / LAST as functions, but they're valid in the context 152 of MATCH_RECOGNIZE, so we need to preserve them in that case. In all other cases 153 they're converted into an ARBITRARY call. 154 155 Reference: https://trino.io/docs/current/sql/match-recognize.html#logical-navigation-functions 156 """ 157 if isinstance(expression.find_ancestor(exp.MatchRecognize, exp.Select), exp.MatchRecognize): 158 return self.function_fallback_sql(expression) 159 160 return rename_func("ARBITRARY")(self, expression) 161 162 163def _unix_to_time_sql(self: Presto.Generator, expression: exp.UnixToTime) -> str: 164 scale = expression.args.get("scale") 165 timestamp = self.sql(expression, "this") 166 if scale in (None, exp.UnixToTime.SECONDS): 167 return rename_func("FROM_UNIXTIME")(self, expression) 168 169 return f"FROM_UNIXTIME(CAST({timestamp} AS DOUBLE) / POW(10, {scale}))" 170 171 172def _to_int(expression: exp.Expression) -> exp.Expression: 173 if not expression.type: 174 from sqlglot.optimizer.annotate_types import annotate_types 175 176 annotate_types(expression) 177 if expression.type and expression.type.this not in exp.DataType.INTEGER_TYPES: 178 return exp.cast(expression, to=exp.DataType.Type.BIGINT) 179 return expression 180 181 182def _build_to_char(args: t.List) -> exp.TimeToStr: 183 fmt = seq_get(args, 1) 184 if isinstance(fmt, exp.Literal): 185 # We uppercase this to match Teradata's format mapping keys 186 fmt.set("this", fmt.this.upper()) 187 188 # We use "teradata" on purpose here, because the time formats are different in Presto. 189 # See https://prestodb.io/docs/current/functions/teradata.html?highlight=to_char#to_char 190 return build_formatted_time(exp.TimeToStr, "teradata")(args) 191 192 193class Presto(Dialect): 194 INDEX_OFFSET = 1 195 NULL_ORDERING = "nulls_are_last" 196 TIME_FORMAT = MySQL.TIME_FORMAT 197 TIME_MAPPING = MySQL.TIME_MAPPING 198 STRICT_STRING_CONCAT = True 199 SUPPORTS_SEMI_ANTI_JOIN = False 200 TYPED_DIVISION = True 201 TABLESAMPLE_SIZE_IS_PERCENT = True 202 LOG_BASE_FIRST: t.Optional[bool] = None 203 204 # https://github.com/trinodb/trino/issues/17 205 # https://github.com/trinodb/trino/issues/12289 206 # https://github.com/prestodb/presto/issues/2863 207 NORMALIZATION_STRATEGY = NormalizationStrategy.CASE_INSENSITIVE 208 209 class Tokenizer(tokens.Tokenizer): 210 UNICODE_STRINGS = [ 211 (prefix + q, q) 212 for q in t.cast(t.List[str], tokens.Tokenizer.QUOTES) 213 for prefix in ("U&", "u&") 214 ] 215 216 KEYWORDS = { 217 **tokens.Tokenizer.KEYWORDS, 218 "START": TokenType.BEGIN, 219 "MATCH_RECOGNIZE": TokenType.MATCH_RECOGNIZE, 220 "ROW": TokenType.STRUCT, 221 "IPADDRESS": TokenType.IPADDRESS, 222 "IPPREFIX": TokenType.IPPREFIX, 223 } 224 225 class Parser(parser.Parser): 226 VALUES_FOLLOWED_BY_PAREN = False 227 228 FUNCTIONS = { 229 **parser.Parser.FUNCTIONS, 230 "ARBITRARY": exp.AnyValue.from_arg_list, 231 "APPROX_DISTINCT": exp.ApproxDistinct.from_arg_list, 232 "APPROX_PERCENTILE": _build_approx_percentile, 233 "BITWISE_AND": binary_from_function(exp.BitwiseAnd), 234 "BITWISE_NOT": lambda args: exp.BitwiseNot(this=seq_get(args, 0)), 235 "BITWISE_OR": binary_from_function(exp.BitwiseOr), 236 "BITWISE_XOR": binary_from_function(exp.BitwiseXor), 237 "CARDINALITY": exp.ArraySize.from_arg_list, 238 "CONTAINS": exp.ArrayContains.from_arg_list, 239 "DATE_ADD": lambda args: exp.DateAdd( 240 this=seq_get(args, 2), expression=seq_get(args, 1), unit=seq_get(args, 0) 241 ), 242 "DATE_DIFF": lambda args: exp.DateDiff( 243 this=seq_get(args, 2), expression=seq_get(args, 1), unit=seq_get(args, 0) 244 ), 245 "DATE_FORMAT": build_formatted_time(exp.TimeToStr, "presto"), 246 "DATE_PARSE": build_formatted_time(exp.StrToTime, "presto"), 247 "DATE_TRUNC": date_trunc_to_time, 248 "ELEMENT_AT": lambda args: exp.Bracket( 249 this=seq_get(args, 0), expressions=[seq_get(args, 1)], offset=1, safe=True 250 ), 251 "FROM_HEX": exp.Unhex.from_arg_list, 252 "FROM_UNIXTIME": _build_from_unixtime, 253 "FROM_UTF8": lambda args: exp.Decode( 254 this=seq_get(args, 0), replace=seq_get(args, 1), charset=exp.Literal.string("utf-8") 255 ), 256 "NOW": exp.CurrentTimestamp.from_arg_list, 257 "REGEXP_EXTRACT": lambda args: exp.RegexpExtract( 258 this=seq_get(args, 0), expression=seq_get(args, 1), group=seq_get(args, 2) 259 ), 260 "REGEXP_REPLACE": lambda args: exp.RegexpReplace( 261 this=seq_get(args, 0), 262 expression=seq_get(args, 1), 263 replacement=seq_get(args, 2) or exp.Literal.string(""), 264 ), 265 "ROW": exp.Struct.from_arg_list, 266 "SEQUENCE": exp.GenerateSeries.from_arg_list, 267 "SET_AGG": exp.ArrayUniqueAgg.from_arg_list, 268 "SPLIT_TO_MAP": exp.StrToMap.from_arg_list, 269 "STRPOS": lambda args: exp.StrPosition( 270 this=seq_get(args, 0), substr=seq_get(args, 1), instance=seq_get(args, 2) 271 ), 272 "TO_CHAR": _build_to_char, 273 "TO_HEX": exp.Hex.from_arg_list, 274 "TO_UNIXTIME": exp.TimeToUnix.from_arg_list, 275 "TO_UTF8": lambda args: exp.Encode( 276 this=seq_get(args, 0), charset=exp.Literal.string("utf-8") 277 ), 278 } 279 280 FUNCTION_PARSERS = parser.Parser.FUNCTION_PARSERS.copy() 281 FUNCTION_PARSERS.pop("TRIM") 282 283 class Generator(generator.Generator): 284 INTERVAL_ALLOWS_PLURAL_FORM = False 285 JOIN_HINTS = False 286 TABLE_HINTS = False 287 QUERY_HINTS = False 288 IS_BOOL_ALLOWED = False 289 TZ_TO_WITH_TIME_ZONE = True 290 NVL2_SUPPORTED = False 291 STRUCT_DELIMITER = ("(", ")") 292 LIMIT_ONLY_LITERALS = True 293 SUPPORTS_SINGLE_ARG_CONCAT = False 294 LIKE_PROPERTY_INSIDE_SCHEMA = True 295 MULTI_ARG_DISTINCT = False 296 SUPPORTS_TO_NUMBER = False 297 298 PROPERTIES_LOCATION = { 299 **generator.Generator.PROPERTIES_LOCATION, 300 exp.LocationProperty: exp.Properties.Location.UNSUPPORTED, 301 exp.VolatileProperty: exp.Properties.Location.UNSUPPORTED, 302 } 303 304 TYPE_MAPPING = { 305 **generator.Generator.TYPE_MAPPING, 306 exp.DataType.Type.INT: "INTEGER", 307 exp.DataType.Type.FLOAT: "REAL", 308 exp.DataType.Type.BINARY: "VARBINARY", 309 exp.DataType.Type.TEXT: "VARCHAR", 310 exp.DataType.Type.TIMETZ: "TIME", 311 exp.DataType.Type.TIMESTAMPTZ: "TIMESTAMP", 312 exp.DataType.Type.STRUCT: "ROW", 313 exp.DataType.Type.DATETIME: "TIMESTAMP", 314 exp.DataType.Type.DATETIME64: "TIMESTAMP", 315 } 316 317 TRANSFORMS = { 318 **generator.Generator.TRANSFORMS, 319 exp.AnyValue: rename_func("ARBITRARY"), 320 exp.ApproxDistinct: lambda self, e: self.func( 321 "APPROX_DISTINCT", e.this, e.args.get("accuracy") 322 ), 323 exp.ApproxQuantile: rename_func("APPROX_PERCENTILE"), 324 exp.ArgMax: rename_func("MAX_BY"), 325 exp.ArgMin: rename_func("MIN_BY"), 326 exp.Array: lambda self, e: f"ARRAY[{self.expressions(e, flat=True)}]", 327 exp.ArrayAny: rename_func("ANY_MATCH"), 328 exp.ArrayConcat: rename_func("CONCAT"), 329 exp.ArrayContains: rename_func("CONTAINS"), 330 exp.ArraySize: rename_func("CARDINALITY"), 331 exp.ArrayToString: rename_func("ARRAY_JOIN"), 332 exp.ArrayUniqueAgg: rename_func("SET_AGG"), 333 exp.AtTimeZone: rename_func("AT_TIMEZONE"), 334 exp.BitwiseAnd: lambda self, e: self.func("BITWISE_AND", e.this, e.expression), 335 exp.BitwiseLeftShift: lambda self, e: self.func( 336 "BITWISE_ARITHMETIC_SHIFT_LEFT", e.this, e.expression 337 ), 338 exp.BitwiseNot: lambda self, e: self.func("BITWISE_NOT", e.this), 339 exp.BitwiseOr: lambda self, e: self.func("BITWISE_OR", e.this, e.expression), 340 exp.BitwiseRightShift: lambda self, e: self.func( 341 "BITWISE_ARITHMETIC_SHIFT_RIGHT", e.this, e.expression 342 ), 343 exp.BitwiseXor: lambda self, e: self.func("BITWISE_XOR", e.this, e.expression), 344 exp.Cast: transforms.preprocess([transforms.epoch_cast_to_ts]), 345 exp.CurrentTimestamp: lambda *_: "CURRENT_TIMESTAMP", 346 exp.DateAdd: lambda self, e: self.func( 347 "DATE_ADD", 348 unit_to_str(e), 349 _to_int(e.expression), 350 e.this, 351 ), 352 exp.DateDiff: lambda self, e: self.func( 353 "DATE_DIFF", unit_to_str(e), e.expression, e.this 354 ), 355 exp.DateStrToDate: datestrtodate_sql, 356 exp.DateToDi: lambda self, 357 e: f"CAST(DATE_FORMAT({self.sql(e, 'this')}, {Presto.DATEINT_FORMAT}) AS INT)", 358 exp.DateSub: lambda self, e: self.func( 359 "DATE_ADD", 360 unit_to_str(e), 361 _to_int(e.expression * -1), 362 e.this, 363 ), 364 exp.Decode: lambda self, e: encode_decode_sql(self, e, "FROM_UTF8"), 365 exp.DiToDate: lambda self, 366 e: f"CAST(DATE_PARSE(CAST({self.sql(e, 'this')} AS VARCHAR), {Presto.DATEINT_FORMAT}) AS DATE)", 367 exp.Encode: lambda self, e: encode_decode_sql(self, e, "TO_UTF8"), 368 exp.FileFormatProperty: lambda self, e: f"FORMAT='{e.name.upper()}'", 369 exp.First: _first_last_sql, 370 exp.FirstValue: _first_last_sql, 371 exp.FromTimeZone: lambda self, 372 e: f"WITH_TIMEZONE({self.sql(e, 'this')}, {self.sql(e, 'zone')}) AT TIME ZONE 'UTC'", 373 exp.Group: transforms.preprocess([transforms.unalias_group]), 374 exp.GroupConcat: lambda self, e: self.func( 375 "ARRAY_JOIN", self.func("ARRAY_AGG", e.this), e.args.get("separator") 376 ), 377 exp.Hex: rename_func("TO_HEX"), 378 exp.If: if_sql(), 379 exp.ILike: no_ilike_sql, 380 exp.Initcap: _initcap_sql, 381 exp.ParseJSON: rename_func("JSON_PARSE"), 382 exp.Last: _first_last_sql, 383 exp.LastValue: _first_last_sql, 384 exp.LastDay: lambda self, e: self.func("LAST_DAY_OF_MONTH", e.this), 385 exp.Lateral: _explode_to_unnest_sql, 386 exp.Left: left_to_substring_sql, 387 exp.Levenshtein: rename_func("LEVENSHTEIN_DISTANCE"), 388 exp.LogicalAnd: rename_func("BOOL_AND"), 389 exp.LogicalOr: rename_func("BOOL_OR"), 390 exp.Pivot: no_pivot_sql, 391 exp.Quantile: _quantile_sql, 392 exp.RegexpExtract: regexp_extract_sql, 393 exp.Right: right_to_substring_sql, 394 exp.SafeDivide: no_safe_divide_sql, 395 exp.Schema: _schema_sql, 396 exp.SchemaCommentProperty: lambda self, e: self.naked_property(e), 397 exp.Select: transforms.preprocess( 398 [ 399 transforms.eliminate_qualify, 400 transforms.eliminate_distinct_on, 401 transforms.explode_to_unnest(1), 402 transforms.eliminate_semi_and_anti_joins, 403 ] 404 ), 405 exp.SortArray: _no_sort_array, 406 exp.StrPosition: lambda self, e: str_position_sql(self, e, generate_instance=True), 407 exp.StrToDate: lambda self, e: f"CAST({_str_to_time_sql(self, e)} AS DATE)", 408 exp.StrToMap: rename_func("SPLIT_TO_MAP"), 409 exp.StrToTime: _str_to_time_sql, 410 exp.StructExtract: struct_extract_sql, 411 exp.Table: transforms.preprocess([_unnest_sequence]), 412 exp.Timestamp: no_timestamp_sql, 413 exp.TimestampTrunc: timestamptrunc_sql, 414 exp.TimeStrToDate: timestrtotime_sql, 415 exp.TimeStrToTime: timestrtotime_sql, 416 exp.TimeStrToUnix: lambda self, e: self.func( 417 "TO_UNIXTIME", self.func("DATE_PARSE", e.this, Presto.TIME_FORMAT) 418 ), 419 exp.TimeToStr: lambda self, e: self.func("DATE_FORMAT", e.this, self.format_time(e)), 420 exp.TimeToUnix: rename_func("TO_UNIXTIME"), 421 exp.ToChar: lambda self, e: self.func("DATE_FORMAT", e.this, self.format_time(e)), 422 exp.TryCast: transforms.preprocess([transforms.epoch_cast_to_ts]), 423 exp.TsOrDiToDi: lambda self, 424 e: f"CAST(SUBSTR(REPLACE(CAST({self.sql(e, 'this')} AS VARCHAR), '-', ''), 1, 8) AS INT)", 425 exp.TsOrDsAdd: _ts_or_ds_add_sql, 426 exp.TsOrDsDiff: _ts_or_ds_diff_sql, 427 exp.TsOrDsToDate: _ts_or_ds_to_date_sql, 428 exp.Unhex: rename_func("FROM_HEX"), 429 exp.UnixToStr: lambda self, 430 e: f"DATE_FORMAT(FROM_UNIXTIME({self.sql(e, 'this')}), {self.format_time(e)})", 431 exp.UnixToTime: _unix_to_time_sql, 432 exp.UnixToTimeStr: lambda self, 433 e: f"CAST(FROM_UNIXTIME({self.sql(e, 'this')}) AS VARCHAR)", 434 exp.VariancePop: rename_func("VAR_POP"), 435 exp.With: transforms.preprocess([transforms.add_recursive_cte_column_names]), 436 exp.WithinGroup: transforms.preprocess( 437 [transforms.remove_within_group_for_percentiles] 438 ), 439 exp.Xor: bool_xor_sql, 440 } 441 442 def strtounix_sql(self, expression: exp.StrToUnix) -> str: 443 # Since `TO_UNIXTIME` requires a `TIMESTAMP`, we need to parse the argument into one. 444 # To do this, we first try to `DATE_PARSE` it, but since this can fail when there's a 445 # timezone involved, we wrap it in a `TRY` call and use `PARSE_DATETIME` as a fallback, 446 # which seems to be using the same time mapping as Hive, as per: 447 # https://joda-time.sourceforge.net/apidocs/org/joda/time/format/DateTimeFormat.html 448 value_as_text = exp.cast(expression.this, "text") 449 parse_without_tz = self.func("DATE_PARSE", value_as_text, self.format_time(expression)) 450 parse_with_tz = self.func( 451 "PARSE_DATETIME", 452 value_as_text, 453 self.format_time(expression, Hive.INVERSE_TIME_MAPPING, Hive.INVERSE_TIME_TRIE), 454 ) 455 coalesced = self.func("COALESCE", self.func("TRY", parse_without_tz), parse_with_tz) 456 return self.func("TO_UNIXTIME", coalesced) 457 458 def bracket_sql(self, expression: exp.Bracket) -> str: 459 if expression.args.get("safe"): 460 return self.func( 461 "ELEMENT_AT", 462 expression.this, 463 seq_get( 464 apply_index_offset( 465 expression.this, 466 expression.expressions, 467 1 - expression.args.get("offset", 0), 468 ), 469 0, 470 ), 471 ) 472 return super().bracket_sql(expression) 473 474 def struct_sql(self, expression: exp.Struct) -> str: 475 from sqlglot.optimizer.annotate_types import annotate_types 476 477 expression = annotate_types(expression) 478 values: t.List[str] = [] 479 schema: t.List[str] = [] 480 unknown_type = False 481 482 for e in expression.expressions: 483 if isinstance(e, exp.PropertyEQ): 484 if e.type and e.type.is_type(exp.DataType.Type.UNKNOWN): 485 unknown_type = True 486 else: 487 schema.append(f"{self.sql(e, 'this')} {self.sql(e.type)}") 488 values.append(self.sql(e, "expression")) 489 else: 490 values.append(self.sql(e)) 491 492 size = len(expression.expressions) 493 494 if not size or len(schema) != size: 495 if unknown_type: 496 self.unsupported( 497 "Cannot convert untyped key-value definitions (try annotate_types)." 498 ) 499 return self.func("ROW", *values) 500 return f"CAST(ROW({', '.join(values)}) AS ROW({', '.join(schema)}))" 501 502 def interval_sql(self, expression: exp.Interval) -> str: 503 if expression.this and expression.text("unit").upper().startswith("WEEK"): 504 return f"({expression.this.name} * INTERVAL '7' DAY)" 505 return super().interval_sql(expression) 506 507 def transaction_sql(self, expression: exp.Transaction) -> str: 508 modes = expression.args.get("modes") 509 modes = f" {', '.join(modes)}" if modes else "" 510 return f"START TRANSACTION{modes}" 511 512 def generateseries_sql(self, expression: exp.GenerateSeries) -> str: 513 start = expression.args["start"] 514 end = expression.args["end"] 515 step = expression.args.get("step") 516 517 if isinstance(start, exp.Cast): 518 target_type = start.to 519 elif isinstance(end, exp.Cast): 520 target_type = end.to 521 else: 522 target_type = None 523 524 if target_type and target_type.is_type("timestamp"): 525 if target_type is start.to: 526 end = exp.cast(end, target_type) 527 else: 528 start = exp.cast(start, target_type) 529 530 return self.func("SEQUENCE", start, end, step) 531 532 def offset_limit_modifiers( 533 self, expression: exp.Expression, fetch: bool, limit: t.Optional[exp.Fetch | exp.Limit] 534 ) -> t.List[str]: 535 return [ 536 self.sql(expression, "offset"), 537 self.sql(limit), 538 ] 539 540 def create_sql(self, expression: exp.Create) -> str: 541 """ 542 Presto doesn't support CREATE VIEW with expressions (ex: `CREATE VIEW x (cola)` then `(cola)` is the expression), 543 so we need to remove them 544 """ 545 kind = expression.args["kind"] 546 schema = expression.this 547 if kind == "VIEW" and schema.expressions: 548 expression.this.set("expressions", None) 549 return super().create_sql(expression)
194class Presto(Dialect): 195 INDEX_OFFSET = 1 196 NULL_ORDERING = "nulls_are_last" 197 TIME_FORMAT = MySQL.TIME_FORMAT 198 TIME_MAPPING = MySQL.TIME_MAPPING 199 STRICT_STRING_CONCAT = True 200 SUPPORTS_SEMI_ANTI_JOIN = False 201 TYPED_DIVISION = True 202 TABLESAMPLE_SIZE_IS_PERCENT = True 203 LOG_BASE_FIRST: t.Optional[bool] = None 204 205 # https://github.com/trinodb/trino/issues/17 206 # https://github.com/trinodb/trino/issues/12289 207 # https://github.com/prestodb/presto/issues/2863 208 NORMALIZATION_STRATEGY = NormalizationStrategy.CASE_INSENSITIVE 209 210 class Tokenizer(tokens.Tokenizer): 211 UNICODE_STRINGS = [ 212 (prefix + q, q) 213 for q in t.cast(t.List[str], tokens.Tokenizer.QUOTES) 214 for prefix in ("U&", "u&") 215 ] 216 217 KEYWORDS = { 218 **tokens.Tokenizer.KEYWORDS, 219 "START": TokenType.BEGIN, 220 "MATCH_RECOGNIZE": TokenType.MATCH_RECOGNIZE, 221 "ROW": TokenType.STRUCT, 222 "IPADDRESS": TokenType.IPADDRESS, 223 "IPPREFIX": TokenType.IPPREFIX, 224 } 225 226 class Parser(parser.Parser): 227 VALUES_FOLLOWED_BY_PAREN = False 228 229 FUNCTIONS = { 230 **parser.Parser.FUNCTIONS, 231 "ARBITRARY": exp.AnyValue.from_arg_list, 232 "APPROX_DISTINCT": exp.ApproxDistinct.from_arg_list, 233 "APPROX_PERCENTILE": _build_approx_percentile, 234 "BITWISE_AND": binary_from_function(exp.BitwiseAnd), 235 "BITWISE_NOT": lambda args: exp.BitwiseNot(this=seq_get(args, 0)), 236 "BITWISE_OR": binary_from_function(exp.BitwiseOr), 237 "BITWISE_XOR": binary_from_function(exp.BitwiseXor), 238 "CARDINALITY": exp.ArraySize.from_arg_list, 239 "CONTAINS": exp.ArrayContains.from_arg_list, 240 "DATE_ADD": lambda args: exp.DateAdd( 241 this=seq_get(args, 2), expression=seq_get(args, 1), unit=seq_get(args, 0) 242 ), 243 "DATE_DIFF": lambda args: exp.DateDiff( 244 this=seq_get(args, 2), expression=seq_get(args, 1), unit=seq_get(args, 0) 245 ), 246 "DATE_FORMAT": build_formatted_time(exp.TimeToStr, "presto"), 247 "DATE_PARSE": build_formatted_time(exp.StrToTime, "presto"), 248 "DATE_TRUNC": date_trunc_to_time, 249 "ELEMENT_AT": lambda args: exp.Bracket( 250 this=seq_get(args, 0), expressions=[seq_get(args, 1)], offset=1, safe=True 251 ), 252 "FROM_HEX": exp.Unhex.from_arg_list, 253 "FROM_UNIXTIME": _build_from_unixtime, 254 "FROM_UTF8": lambda args: exp.Decode( 255 this=seq_get(args, 0), replace=seq_get(args, 1), charset=exp.Literal.string("utf-8") 256 ), 257 "NOW": exp.CurrentTimestamp.from_arg_list, 258 "REGEXP_EXTRACT": lambda args: exp.RegexpExtract( 259 this=seq_get(args, 0), expression=seq_get(args, 1), group=seq_get(args, 2) 260 ), 261 "REGEXP_REPLACE": lambda args: exp.RegexpReplace( 262 this=seq_get(args, 0), 263 expression=seq_get(args, 1), 264 replacement=seq_get(args, 2) or exp.Literal.string(""), 265 ), 266 "ROW": exp.Struct.from_arg_list, 267 "SEQUENCE": exp.GenerateSeries.from_arg_list, 268 "SET_AGG": exp.ArrayUniqueAgg.from_arg_list, 269 "SPLIT_TO_MAP": exp.StrToMap.from_arg_list, 270 "STRPOS": lambda args: exp.StrPosition( 271 this=seq_get(args, 0), substr=seq_get(args, 1), instance=seq_get(args, 2) 272 ), 273 "TO_CHAR": _build_to_char, 274 "TO_HEX": exp.Hex.from_arg_list, 275 "TO_UNIXTIME": exp.TimeToUnix.from_arg_list, 276 "TO_UTF8": lambda args: exp.Encode( 277 this=seq_get(args, 0), charset=exp.Literal.string("utf-8") 278 ), 279 } 280 281 FUNCTION_PARSERS = parser.Parser.FUNCTION_PARSERS.copy() 282 FUNCTION_PARSERS.pop("TRIM") 283 284 class Generator(generator.Generator): 285 INTERVAL_ALLOWS_PLURAL_FORM = False 286 JOIN_HINTS = False 287 TABLE_HINTS = False 288 QUERY_HINTS = False 289 IS_BOOL_ALLOWED = False 290 TZ_TO_WITH_TIME_ZONE = True 291 NVL2_SUPPORTED = False 292 STRUCT_DELIMITER = ("(", ")") 293 LIMIT_ONLY_LITERALS = True 294 SUPPORTS_SINGLE_ARG_CONCAT = False 295 LIKE_PROPERTY_INSIDE_SCHEMA = True 296 MULTI_ARG_DISTINCT = False 297 SUPPORTS_TO_NUMBER = False 298 299 PROPERTIES_LOCATION = { 300 **generator.Generator.PROPERTIES_LOCATION, 301 exp.LocationProperty: exp.Properties.Location.UNSUPPORTED, 302 exp.VolatileProperty: exp.Properties.Location.UNSUPPORTED, 303 } 304 305 TYPE_MAPPING = { 306 **generator.Generator.TYPE_MAPPING, 307 exp.DataType.Type.INT: "INTEGER", 308 exp.DataType.Type.FLOAT: "REAL", 309 exp.DataType.Type.BINARY: "VARBINARY", 310 exp.DataType.Type.TEXT: "VARCHAR", 311 exp.DataType.Type.TIMETZ: "TIME", 312 exp.DataType.Type.TIMESTAMPTZ: "TIMESTAMP", 313 exp.DataType.Type.STRUCT: "ROW", 314 exp.DataType.Type.DATETIME: "TIMESTAMP", 315 exp.DataType.Type.DATETIME64: "TIMESTAMP", 316 } 317 318 TRANSFORMS = { 319 **generator.Generator.TRANSFORMS, 320 exp.AnyValue: rename_func("ARBITRARY"), 321 exp.ApproxDistinct: lambda self, e: self.func( 322 "APPROX_DISTINCT", e.this, e.args.get("accuracy") 323 ), 324 exp.ApproxQuantile: rename_func("APPROX_PERCENTILE"), 325 exp.ArgMax: rename_func("MAX_BY"), 326 exp.ArgMin: rename_func("MIN_BY"), 327 exp.Array: lambda self, e: f"ARRAY[{self.expressions(e, flat=True)}]", 328 exp.ArrayAny: rename_func("ANY_MATCH"), 329 exp.ArrayConcat: rename_func("CONCAT"), 330 exp.ArrayContains: rename_func("CONTAINS"), 331 exp.ArraySize: rename_func("CARDINALITY"), 332 exp.ArrayToString: rename_func("ARRAY_JOIN"), 333 exp.ArrayUniqueAgg: rename_func("SET_AGG"), 334 exp.AtTimeZone: rename_func("AT_TIMEZONE"), 335 exp.BitwiseAnd: lambda self, e: self.func("BITWISE_AND", e.this, e.expression), 336 exp.BitwiseLeftShift: lambda self, e: self.func( 337 "BITWISE_ARITHMETIC_SHIFT_LEFT", e.this, e.expression 338 ), 339 exp.BitwiseNot: lambda self, e: self.func("BITWISE_NOT", e.this), 340 exp.BitwiseOr: lambda self, e: self.func("BITWISE_OR", e.this, e.expression), 341 exp.BitwiseRightShift: lambda self, e: self.func( 342 "BITWISE_ARITHMETIC_SHIFT_RIGHT", e.this, e.expression 343 ), 344 exp.BitwiseXor: lambda self, e: self.func("BITWISE_XOR", e.this, e.expression), 345 exp.Cast: transforms.preprocess([transforms.epoch_cast_to_ts]), 346 exp.CurrentTimestamp: lambda *_: "CURRENT_TIMESTAMP", 347 exp.DateAdd: lambda self, e: self.func( 348 "DATE_ADD", 349 unit_to_str(e), 350 _to_int(e.expression), 351 e.this, 352 ), 353 exp.DateDiff: lambda self, e: self.func( 354 "DATE_DIFF", unit_to_str(e), e.expression, e.this 355 ), 356 exp.DateStrToDate: datestrtodate_sql, 357 exp.DateToDi: lambda self, 358 e: f"CAST(DATE_FORMAT({self.sql(e, 'this')}, {Presto.DATEINT_FORMAT}) AS INT)", 359 exp.DateSub: lambda self, e: self.func( 360 "DATE_ADD", 361 unit_to_str(e), 362 _to_int(e.expression * -1), 363 e.this, 364 ), 365 exp.Decode: lambda self, e: encode_decode_sql(self, e, "FROM_UTF8"), 366 exp.DiToDate: lambda self, 367 e: f"CAST(DATE_PARSE(CAST({self.sql(e, 'this')} AS VARCHAR), {Presto.DATEINT_FORMAT}) AS DATE)", 368 exp.Encode: lambda self, e: encode_decode_sql(self, e, "TO_UTF8"), 369 exp.FileFormatProperty: lambda self, e: f"FORMAT='{e.name.upper()}'", 370 exp.First: _first_last_sql, 371 exp.FirstValue: _first_last_sql, 372 exp.FromTimeZone: lambda self, 373 e: f"WITH_TIMEZONE({self.sql(e, 'this')}, {self.sql(e, 'zone')}) AT TIME ZONE 'UTC'", 374 exp.Group: transforms.preprocess([transforms.unalias_group]), 375 exp.GroupConcat: lambda self, e: self.func( 376 "ARRAY_JOIN", self.func("ARRAY_AGG", e.this), e.args.get("separator") 377 ), 378 exp.Hex: rename_func("TO_HEX"), 379 exp.If: if_sql(), 380 exp.ILike: no_ilike_sql, 381 exp.Initcap: _initcap_sql, 382 exp.ParseJSON: rename_func("JSON_PARSE"), 383 exp.Last: _first_last_sql, 384 exp.LastValue: _first_last_sql, 385 exp.LastDay: lambda self, e: self.func("LAST_DAY_OF_MONTH", e.this), 386 exp.Lateral: _explode_to_unnest_sql, 387 exp.Left: left_to_substring_sql, 388 exp.Levenshtein: rename_func("LEVENSHTEIN_DISTANCE"), 389 exp.LogicalAnd: rename_func("BOOL_AND"), 390 exp.LogicalOr: rename_func("BOOL_OR"), 391 exp.Pivot: no_pivot_sql, 392 exp.Quantile: _quantile_sql, 393 exp.RegexpExtract: regexp_extract_sql, 394 exp.Right: right_to_substring_sql, 395 exp.SafeDivide: no_safe_divide_sql, 396 exp.Schema: _schema_sql, 397 exp.SchemaCommentProperty: lambda self, e: self.naked_property(e), 398 exp.Select: transforms.preprocess( 399 [ 400 transforms.eliminate_qualify, 401 transforms.eliminate_distinct_on, 402 transforms.explode_to_unnest(1), 403 transforms.eliminate_semi_and_anti_joins, 404 ] 405 ), 406 exp.SortArray: _no_sort_array, 407 exp.StrPosition: lambda self, e: str_position_sql(self, e, generate_instance=True), 408 exp.StrToDate: lambda self, e: f"CAST({_str_to_time_sql(self, e)} AS DATE)", 409 exp.StrToMap: rename_func("SPLIT_TO_MAP"), 410 exp.StrToTime: _str_to_time_sql, 411 exp.StructExtract: struct_extract_sql, 412 exp.Table: transforms.preprocess([_unnest_sequence]), 413 exp.Timestamp: no_timestamp_sql, 414 exp.TimestampTrunc: timestamptrunc_sql, 415 exp.TimeStrToDate: timestrtotime_sql, 416 exp.TimeStrToTime: timestrtotime_sql, 417 exp.TimeStrToUnix: lambda self, e: self.func( 418 "TO_UNIXTIME", self.func("DATE_PARSE", e.this, Presto.TIME_FORMAT) 419 ), 420 exp.TimeToStr: lambda self, e: self.func("DATE_FORMAT", e.this, self.format_time(e)), 421 exp.TimeToUnix: rename_func("TO_UNIXTIME"), 422 exp.ToChar: lambda self, e: self.func("DATE_FORMAT", e.this, self.format_time(e)), 423 exp.TryCast: transforms.preprocess([transforms.epoch_cast_to_ts]), 424 exp.TsOrDiToDi: lambda self, 425 e: f"CAST(SUBSTR(REPLACE(CAST({self.sql(e, 'this')} AS VARCHAR), '-', ''), 1, 8) AS INT)", 426 exp.TsOrDsAdd: _ts_or_ds_add_sql, 427 exp.TsOrDsDiff: _ts_or_ds_diff_sql, 428 exp.TsOrDsToDate: _ts_or_ds_to_date_sql, 429 exp.Unhex: rename_func("FROM_HEX"), 430 exp.UnixToStr: lambda self, 431 e: f"DATE_FORMAT(FROM_UNIXTIME({self.sql(e, 'this')}), {self.format_time(e)})", 432 exp.UnixToTime: _unix_to_time_sql, 433 exp.UnixToTimeStr: lambda self, 434 e: f"CAST(FROM_UNIXTIME({self.sql(e, 'this')}) AS VARCHAR)", 435 exp.VariancePop: rename_func("VAR_POP"), 436 exp.With: transforms.preprocess([transforms.add_recursive_cte_column_names]), 437 exp.WithinGroup: transforms.preprocess( 438 [transforms.remove_within_group_for_percentiles] 439 ), 440 exp.Xor: bool_xor_sql, 441 } 442 443 def strtounix_sql(self, expression: exp.StrToUnix) -> str: 444 # Since `TO_UNIXTIME` requires a `TIMESTAMP`, we need to parse the argument into one. 445 # To do this, we first try to `DATE_PARSE` it, but since this can fail when there's a 446 # timezone involved, we wrap it in a `TRY` call and use `PARSE_DATETIME` as a fallback, 447 # which seems to be using the same time mapping as Hive, as per: 448 # https://joda-time.sourceforge.net/apidocs/org/joda/time/format/DateTimeFormat.html 449 value_as_text = exp.cast(expression.this, "text") 450 parse_without_tz = self.func("DATE_PARSE", value_as_text, self.format_time(expression)) 451 parse_with_tz = self.func( 452 "PARSE_DATETIME", 453 value_as_text, 454 self.format_time(expression, Hive.INVERSE_TIME_MAPPING, Hive.INVERSE_TIME_TRIE), 455 ) 456 coalesced = self.func("COALESCE", self.func("TRY", parse_without_tz), parse_with_tz) 457 return self.func("TO_UNIXTIME", coalesced) 458 459 def bracket_sql(self, expression: exp.Bracket) -> str: 460 if expression.args.get("safe"): 461 return self.func( 462 "ELEMENT_AT", 463 expression.this, 464 seq_get( 465 apply_index_offset( 466 expression.this, 467 expression.expressions, 468 1 - expression.args.get("offset", 0), 469 ), 470 0, 471 ), 472 ) 473 return super().bracket_sql(expression) 474 475 def struct_sql(self, expression: exp.Struct) -> str: 476 from sqlglot.optimizer.annotate_types import annotate_types 477 478 expression = annotate_types(expression) 479 values: t.List[str] = [] 480 schema: t.List[str] = [] 481 unknown_type = False 482 483 for e in expression.expressions: 484 if isinstance(e, exp.PropertyEQ): 485 if e.type and e.type.is_type(exp.DataType.Type.UNKNOWN): 486 unknown_type = True 487 else: 488 schema.append(f"{self.sql(e, 'this')} {self.sql(e.type)}") 489 values.append(self.sql(e, "expression")) 490 else: 491 values.append(self.sql(e)) 492 493 size = len(expression.expressions) 494 495 if not size or len(schema) != size: 496 if unknown_type: 497 self.unsupported( 498 "Cannot convert untyped key-value definitions (try annotate_types)." 499 ) 500 return self.func("ROW", *values) 501 return f"CAST(ROW({', '.join(values)}) AS ROW({', '.join(schema)}))" 502 503 def interval_sql(self, expression: exp.Interval) -> str: 504 if expression.this and expression.text("unit").upper().startswith("WEEK"): 505 return f"({expression.this.name} * INTERVAL '7' DAY)" 506 return super().interval_sql(expression) 507 508 def transaction_sql(self, expression: exp.Transaction) -> str: 509 modes = expression.args.get("modes") 510 modes = f" {', '.join(modes)}" if modes else "" 511 return f"START TRANSACTION{modes}" 512 513 def generateseries_sql(self, expression: exp.GenerateSeries) -> str: 514 start = expression.args["start"] 515 end = expression.args["end"] 516 step = expression.args.get("step") 517 518 if isinstance(start, exp.Cast): 519 target_type = start.to 520 elif isinstance(end, exp.Cast): 521 target_type = end.to 522 else: 523 target_type = None 524 525 if target_type and target_type.is_type("timestamp"): 526 if target_type is start.to: 527 end = exp.cast(end, target_type) 528 else: 529 start = exp.cast(start, target_type) 530 531 return self.func("SEQUENCE", start, end, step) 532 533 def offset_limit_modifiers( 534 self, expression: exp.Expression, fetch: bool, limit: t.Optional[exp.Fetch | exp.Limit] 535 ) -> t.List[str]: 536 return [ 537 self.sql(expression, "offset"), 538 self.sql(limit), 539 ] 540 541 def create_sql(self, expression: exp.Create) -> str: 542 """ 543 Presto doesn't support CREATE VIEW with expressions (ex: `CREATE VIEW x (cola)` then `(cola)` is the expression), 544 so we need to remove them 545 """ 546 kind = expression.args["kind"] 547 schema = expression.this 548 if kind == "VIEW" and schema.expressions: 549 expression.this.set("expressions", None) 550 return super().create_sql(expression)
Default NULL
ordering method to use if not explicitly set.
Possible values: "nulls_are_small"
, "nulls_are_large"
, "nulls_are_last"
Associates this dialect's time formats with their equivalent Python strftime
formats.
Whether the behavior of a / b
depends on the types of a
and b
.
False means a / b
is always float division.
True means a / b
is integer division if both a
and b
are integers.
Whether the base comes first in the LOG
function.
Possible values: True
, False
, None
(two arguments are not supported by LOG
)
Specifies the strategy according to which identifiers should be normalized.
Inherited Members
- sqlglot.dialects.dialect.Dialect
- Dialect
- WEEK_OFFSET
- UNNEST_COLUMN_ONLY
- ALIAS_POST_TABLESAMPLE
- IDENTIFIERS_CAN_START_WITH_DIGIT
- DPIPE_IS_STRING_CONCAT
- SUPPORTS_USER_DEFINED_TYPES
- NORMALIZE_FUNCTIONS
- SAFE_DIVISION
- CONCAT_COALESCE
- DATE_FORMAT
- DATEINT_FORMAT
- FORMAT_MAPPING
- UNESCAPED_SEQUENCES
- PSEUDOCOLUMNS
- PREFER_CTE_ALIAS_COLUMN
- get_or_raise
- format_time
- normalize_identifier
- case_sensitive
- can_identify
- quote_identifier
- to_json_path
- parse
- parse_into
- generate
- transpile
- tokenize
- tokenizer
- parser
- generator
210 class Tokenizer(tokens.Tokenizer): 211 UNICODE_STRINGS = [ 212 (prefix + q, q) 213 for q in t.cast(t.List[str], tokens.Tokenizer.QUOTES) 214 for prefix in ("U&", "u&") 215 ] 216 217 KEYWORDS = { 218 **tokens.Tokenizer.KEYWORDS, 219 "START": TokenType.BEGIN, 220 "MATCH_RECOGNIZE": TokenType.MATCH_RECOGNIZE, 221 "ROW": TokenType.STRUCT, 222 "IPADDRESS": TokenType.IPADDRESS, 223 "IPPREFIX": TokenType.IPPREFIX, 224 }
Inherited Members
- sqlglot.tokens.Tokenizer
- Tokenizer
- SINGLE_TOKENS
- BIT_STRINGS
- BYTE_STRINGS
- HEX_STRINGS
- RAW_STRINGS
- HEREDOC_STRINGS
- IDENTIFIERS
- IDENTIFIER_ESCAPES
- QUOTES
- STRING_ESCAPES
- VAR_SINGLE_TOKENS
- HEREDOC_TAG_IS_IDENTIFIER
- HEREDOC_STRING_ALTERNATIVE
- WHITE_SPACE
- COMMANDS
- COMMAND_PREFIX_TOKENS
- NUMERIC_LITERALS
- COMMENTS
- dialect
- reset
- tokenize
- tokenize_rs
- size
- sql
- tokens
226 class Parser(parser.Parser): 227 VALUES_FOLLOWED_BY_PAREN = False 228 229 FUNCTIONS = { 230 **parser.Parser.FUNCTIONS, 231 "ARBITRARY": exp.AnyValue.from_arg_list, 232 "APPROX_DISTINCT": exp.ApproxDistinct.from_arg_list, 233 "APPROX_PERCENTILE": _build_approx_percentile, 234 "BITWISE_AND": binary_from_function(exp.BitwiseAnd), 235 "BITWISE_NOT": lambda args: exp.BitwiseNot(this=seq_get(args, 0)), 236 "BITWISE_OR": binary_from_function(exp.BitwiseOr), 237 "BITWISE_XOR": binary_from_function(exp.BitwiseXor), 238 "CARDINALITY": exp.ArraySize.from_arg_list, 239 "CONTAINS": exp.ArrayContains.from_arg_list, 240 "DATE_ADD": lambda args: exp.DateAdd( 241 this=seq_get(args, 2), expression=seq_get(args, 1), unit=seq_get(args, 0) 242 ), 243 "DATE_DIFF": lambda args: exp.DateDiff( 244 this=seq_get(args, 2), expression=seq_get(args, 1), unit=seq_get(args, 0) 245 ), 246 "DATE_FORMAT": build_formatted_time(exp.TimeToStr, "presto"), 247 "DATE_PARSE": build_formatted_time(exp.StrToTime, "presto"), 248 "DATE_TRUNC": date_trunc_to_time, 249 "ELEMENT_AT": lambda args: exp.Bracket( 250 this=seq_get(args, 0), expressions=[seq_get(args, 1)], offset=1, safe=True 251 ), 252 "FROM_HEX": exp.Unhex.from_arg_list, 253 "FROM_UNIXTIME": _build_from_unixtime, 254 "FROM_UTF8": lambda args: exp.Decode( 255 this=seq_get(args, 0), replace=seq_get(args, 1), charset=exp.Literal.string("utf-8") 256 ), 257 "NOW": exp.CurrentTimestamp.from_arg_list, 258 "REGEXP_EXTRACT": lambda args: exp.RegexpExtract( 259 this=seq_get(args, 0), expression=seq_get(args, 1), group=seq_get(args, 2) 260 ), 261 "REGEXP_REPLACE": lambda args: exp.RegexpReplace( 262 this=seq_get(args, 0), 263 expression=seq_get(args, 1), 264 replacement=seq_get(args, 2) or exp.Literal.string(""), 265 ), 266 "ROW": exp.Struct.from_arg_list, 267 "SEQUENCE": exp.GenerateSeries.from_arg_list, 268 "SET_AGG": exp.ArrayUniqueAgg.from_arg_list, 269 "SPLIT_TO_MAP": exp.StrToMap.from_arg_list, 270 "STRPOS": lambda args: exp.StrPosition( 271 this=seq_get(args, 0), substr=seq_get(args, 1), instance=seq_get(args, 2) 272 ), 273 "TO_CHAR": _build_to_char, 274 "TO_HEX": exp.Hex.from_arg_list, 275 "TO_UNIXTIME": exp.TimeToUnix.from_arg_list, 276 "TO_UTF8": lambda args: exp.Encode( 277 this=seq_get(args, 0), charset=exp.Literal.string("utf-8") 278 ), 279 } 280 281 FUNCTION_PARSERS = parser.Parser.FUNCTION_PARSERS.copy() 282 FUNCTION_PARSERS.pop("TRIM")
Parser consumes a list of tokens produced by the Tokenizer and produces a parsed syntax tree.
Arguments:
- error_level: The desired error level. Default: ErrorLevel.IMMEDIATE
- error_message_context: The amount of context to capture from a query string when displaying the error message (in number of characters). Default: 100
- max_errors: Maximum number of error messages to include in a raised ParseError. This is only relevant if error_level is ErrorLevel.RAISE. Default: 3
Inherited Members
- sqlglot.parser.Parser
- Parser
- NO_PAREN_FUNCTIONS
- STRUCT_TYPE_TOKENS
- NESTED_TYPE_TOKENS
- ENUM_TYPE_TOKENS
- AGGREGATE_TYPE_TOKENS
- TYPE_TOKENS
- SIGNED_TO_UNSIGNED_TYPE_TOKEN
- SUBQUERY_PREDICATES
- RESERVED_TOKENS
- DB_CREATABLES
- CREATABLES
- ID_VAR_TOKENS
- INTERVAL_VARS
- ALIAS_TOKENS
- COMMENT_TABLE_ALIAS_TOKENS
- UPDATE_ALIAS_TOKENS
- TRIM_TYPES
- FUNC_TOKENS
- CONJUNCTION
- EQUALITY
- COMPARISON
- BITWISE
- TERM
- FACTOR
- EXPONENT
- TIMES
- TIMESTAMPS
- SET_OPERATIONS
- JOIN_METHODS
- JOIN_SIDES
- JOIN_KINDS
- JOIN_HINTS
- LAMBDAS
- COLUMN_OPERATORS
- EXPRESSION_PARSERS
- STATEMENT_PARSERS
- UNARY_PARSERS
- STRING_PARSERS
- NUMERIC_PARSERS
- PRIMARY_PARSERS
- PLACEHOLDER_PARSERS
- RANGE_PARSERS
- PROPERTY_PARSERS
- CONSTRAINT_PARSERS
- ALTER_PARSERS
- SCHEMA_UNNAMED_CONSTRAINTS
- NO_PAREN_FUNCTION_PARSERS
- INVALID_FUNC_NAME_TOKENS
- FUNCTIONS_WITH_ALIASED_ARGS
- KEY_VALUE_DEFINITIONS
- QUERY_MODIFIER_PARSERS
- SET_PARSERS
- SHOW_PARSERS
- TYPE_LITERAL_PARSERS
- DDL_SELECT_TOKENS
- PRE_VOLATILE_TOKENS
- TRANSACTION_KIND
- TRANSACTION_CHARACTERISTICS
- CONFLICT_ACTIONS
- CREATE_SEQUENCE
- ISOLATED_LOADING_OPTIONS
- USABLES
- CAST_ACTIONS
- INSERT_ALTERNATIVES
- CLONE_KEYWORDS
- HISTORICAL_DATA_KIND
- OPCLASS_FOLLOW_KEYWORDS
- OPTYPE_FOLLOW_TOKENS
- TABLE_INDEX_HINT_TOKENS
- VIEW_ATTRIBUTES
- WINDOW_ALIAS_TOKENS
- WINDOW_BEFORE_PAREN_TOKENS
- WINDOW_SIDES
- JSON_KEY_VALUE_SEPARATOR_TOKENS
- FETCH_TOKENS
- ADD_CONSTRAINT_TOKENS
- DISTINCT_TOKENS
- NULL_TOKENS
- UNNEST_OFFSET_ALIAS_TOKENS
- SELECT_START_TOKENS
- STRICT_CAST
- PREFIXED_PIVOT_COLUMNS
- IDENTIFY_PIVOT_STRINGS
- LOG_DEFAULTS_TO_LN
- ALTER_TABLE_ADD_REQUIRED_FOR_EACH_COLUMN
- TABLESAMPLE_CSV
- SET_REQUIRES_ASSIGNMENT_DELIMITER
- TRIM_PATTERN_FIRST
- STRING_ALIASES
- MODIFIERS_ATTACHED_TO_UNION
- UNION_MODIFIERS
- NO_PAREN_IF_COMMANDS
- JSON_ARROWS_REQUIRE_JSON_TYPE
- SUPPORTS_IMPLICIT_UNNEST
- INTERVAL_SPANS
- error_level
- error_message_context
- max_errors
- dialect
- reset
- parse
- parse_into
- check_errors
- raise_error
- expression
- validate_expression
- errors
- sql
284 class Generator(generator.Generator): 285 INTERVAL_ALLOWS_PLURAL_FORM = False 286 JOIN_HINTS = False 287 TABLE_HINTS = False 288 QUERY_HINTS = False 289 IS_BOOL_ALLOWED = False 290 TZ_TO_WITH_TIME_ZONE = True 291 NVL2_SUPPORTED = False 292 STRUCT_DELIMITER = ("(", ")") 293 LIMIT_ONLY_LITERALS = True 294 SUPPORTS_SINGLE_ARG_CONCAT = False 295 LIKE_PROPERTY_INSIDE_SCHEMA = True 296 MULTI_ARG_DISTINCT = False 297 SUPPORTS_TO_NUMBER = False 298 299 PROPERTIES_LOCATION = { 300 **generator.Generator.PROPERTIES_LOCATION, 301 exp.LocationProperty: exp.Properties.Location.UNSUPPORTED, 302 exp.VolatileProperty: exp.Properties.Location.UNSUPPORTED, 303 } 304 305 TYPE_MAPPING = { 306 **generator.Generator.TYPE_MAPPING, 307 exp.DataType.Type.INT: "INTEGER", 308 exp.DataType.Type.FLOAT: "REAL", 309 exp.DataType.Type.BINARY: "VARBINARY", 310 exp.DataType.Type.TEXT: "VARCHAR", 311 exp.DataType.Type.TIMETZ: "TIME", 312 exp.DataType.Type.TIMESTAMPTZ: "TIMESTAMP", 313 exp.DataType.Type.STRUCT: "ROW", 314 exp.DataType.Type.DATETIME: "TIMESTAMP", 315 exp.DataType.Type.DATETIME64: "TIMESTAMP", 316 } 317 318 TRANSFORMS = { 319 **generator.Generator.TRANSFORMS, 320 exp.AnyValue: rename_func("ARBITRARY"), 321 exp.ApproxDistinct: lambda self, e: self.func( 322 "APPROX_DISTINCT", e.this, e.args.get("accuracy") 323 ), 324 exp.ApproxQuantile: rename_func("APPROX_PERCENTILE"), 325 exp.ArgMax: rename_func("MAX_BY"), 326 exp.ArgMin: rename_func("MIN_BY"), 327 exp.Array: lambda self, e: f"ARRAY[{self.expressions(e, flat=True)}]", 328 exp.ArrayAny: rename_func("ANY_MATCH"), 329 exp.ArrayConcat: rename_func("CONCAT"), 330 exp.ArrayContains: rename_func("CONTAINS"), 331 exp.ArraySize: rename_func("CARDINALITY"), 332 exp.ArrayToString: rename_func("ARRAY_JOIN"), 333 exp.ArrayUniqueAgg: rename_func("SET_AGG"), 334 exp.AtTimeZone: rename_func("AT_TIMEZONE"), 335 exp.BitwiseAnd: lambda self, e: self.func("BITWISE_AND", e.this, e.expression), 336 exp.BitwiseLeftShift: lambda self, e: self.func( 337 "BITWISE_ARITHMETIC_SHIFT_LEFT", e.this, e.expression 338 ), 339 exp.BitwiseNot: lambda self, e: self.func("BITWISE_NOT", e.this), 340 exp.BitwiseOr: lambda self, e: self.func("BITWISE_OR", e.this, e.expression), 341 exp.BitwiseRightShift: lambda self, e: self.func( 342 "BITWISE_ARITHMETIC_SHIFT_RIGHT", e.this, e.expression 343 ), 344 exp.BitwiseXor: lambda self, e: self.func("BITWISE_XOR", e.this, e.expression), 345 exp.Cast: transforms.preprocess([transforms.epoch_cast_to_ts]), 346 exp.CurrentTimestamp: lambda *_: "CURRENT_TIMESTAMP", 347 exp.DateAdd: lambda self, e: self.func( 348 "DATE_ADD", 349 unit_to_str(e), 350 _to_int(e.expression), 351 e.this, 352 ), 353 exp.DateDiff: lambda self, e: self.func( 354 "DATE_DIFF", unit_to_str(e), e.expression, e.this 355 ), 356 exp.DateStrToDate: datestrtodate_sql, 357 exp.DateToDi: lambda self, 358 e: f"CAST(DATE_FORMAT({self.sql(e, 'this')}, {Presto.DATEINT_FORMAT}) AS INT)", 359 exp.DateSub: lambda self, e: self.func( 360 "DATE_ADD", 361 unit_to_str(e), 362 _to_int(e.expression * -1), 363 e.this, 364 ), 365 exp.Decode: lambda self, e: encode_decode_sql(self, e, "FROM_UTF8"), 366 exp.DiToDate: lambda self, 367 e: f"CAST(DATE_PARSE(CAST({self.sql(e, 'this')} AS VARCHAR), {Presto.DATEINT_FORMAT}) AS DATE)", 368 exp.Encode: lambda self, e: encode_decode_sql(self, e, "TO_UTF8"), 369 exp.FileFormatProperty: lambda self, e: f"FORMAT='{e.name.upper()}'", 370 exp.First: _first_last_sql, 371 exp.FirstValue: _first_last_sql, 372 exp.FromTimeZone: lambda self, 373 e: f"WITH_TIMEZONE({self.sql(e, 'this')}, {self.sql(e, 'zone')}) AT TIME ZONE 'UTC'", 374 exp.Group: transforms.preprocess([transforms.unalias_group]), 375 exp.GroupConcat: lambda self, e: self.func( 376 "ARRAY_JOIN", self.func("ARRAY_AGG", e.this), e.args.get("separator") 377 ), 378 exp.Hex: rename_func("TO_HEX"), 379 exp.If: if_sql(), 380 exp.ILike: no_ilike_sql, 381 exp.Initcap: _initcap_sql, 382 exp.ParseJSON: rename_func("JSON_PARSE"), 383 exp.Last: _first_last_sql, 384 exp.LastValue: _first_last_sql, 385 exp.LastDay: lambda self, e: self.func("LAST_DAY_OF_MONTH", e.this), 386 exp.Lateral: _explode_to_unnest_sql, 387 exp.Left: left_to_substring_sql, 388 exp.Levenshtein: rename_func("LEVENSHTEIN_DISTANCE"), 389 exp.LogicalAnd: rename_func("BOOL_AND"), 390 exp.LogicalOr: rename_func("BOOL_OR"), 391 exp.Pivot: no_pivot_sql, 392 exp.Quantile: _quantile_sql, 393 exp.RegexpExtract: regexp_extract_sql, 394 exp.Right: right_to_substring_sql, 395 exp.SafeDivide: no_safe_divide_sql, 396 exp.Schema: _schema_sql, 397 exp.SchemaCommentProperty: lambda self, e: self.naked_property(e), 398 exp.Select: transforms.preprocess( 399 [ 400 transforms.eliminate_qualify, 401 transforms.eliminate_distinct_on, 402 transforms.explode_to_unnest(1), 403 transforms.eliminate_semi_and_anti_joins, 404 ] 405 ), 406 exp.SortArray: _no_sort_array, 407 exp.StrPosition: lambda self, e: str_position_sql(self, e, generate_instance=True), 408 exp.StrToDate: lambda self, e: f"CAST({_str_to_time_sql(self, e)} AS DATE)", 409 exp.StrToMap: rename_func("SPLIT_TO_MAP"), 410 exp.StrToTime: _str_to_time_sql, 411 exp.StructExtract: struct_extract_sql, 412 exp.Table: transforms.preprocess([_unnest_sequence]), 413 exp.Timestamp: no_timestamp_sql, 414 exp.TimestampTrunc: timestamptrunc_sql, 415 exp.TimeStrToDate: timestrtotime_sql, 416 exp.TimeStrToTime: timestrtotime_sql, 417 exp.TimeStrToUnix: lambda self, e: self.func( 418 "TO_UNIXTIME", self.func("DATE_PARSE", e.this, Presto.TIME_FORMAT) 419 ), 420 exp.TimeToStr: lambda self, e: self.func("DATE_FORMAT", e.this, self.format_time(e)), 421 exp.TimeToUnix: rename_func("TO_UNIXTIME"), 422 exp.ToChar: lambda self, e: self.func("DATE_FORMAT", e.this, self.format_time(e)), 423 exp.TryCast: transforms.preprocess([transforms.epoch_cast_to_ts]), 424 exp.TsOrDiToDi: lambda self, 425 e: f"CAST(SUBSTR(REPLACE(CAST({self.sql(e, 'this')} AS VARCHAR), '-', ''), 1, 8) AS INT)", 426 exp.TsOrDsAdd: _ts_or_ds_add_sql, 427 exp.TsOrDsDiff: _ts_or_ds_diff_sql, 428 exp.TsOrDsToDate: _ts_or_ds_to_date_sql, 429 exp.Unhex: rename_func("FROM_HEX"), 430 exp.UnixToStr: lambda self, 431 e: f"DATE_FORMAT(FROM_UNIXTIME({self.sql(e, 'this')}), {self.format_time(e)})", 432 exp.UnixToTime: _unix_to_time_sql, 433 exp.UnixToTimeStr: lambda self, 434 e: f"CAST(FROM_UNIXTIME({self.sql(e, 'this')}) AS VARCHAR)", 435 exp.VariancePop: rename_func("VAR_POP"), 436 exp.With: transforms.preprocess([transforms.add_recursive_cte_column_names]), 437 exp.WithinGroup: transforms.preprocess( 438 [transforms.remove_within_group_for_percentiles] 439 ), 440 exp.Xor: bool_xor_sql, 441 } 442 443 def strtounix_sql(self, expression: exp.StrToUnix) -> str: 444 # Since `TO_UNIXTIME` requires a `TIMESTAMP`, we need to parse the argument into one. 445 # To do this, we first try to `DATE_PARSE` it, but since this can fail when there's a 446 # timezone involved, we wrap it in a `TRY` call and use `PARSE_DATETIME` as a fallback, 447 # which seems to be using the same time mapping as Hive, as per: 448 # https://joda-time.sourceforge.net/apidocs/org/joda/time/format/DateTimeFormat.html 449 value_as_text = exp.cast(expression.this, "text") 450 parse_without_tz = self.func("DATE_PARSE", value_as_text, self.format_time(expression)) 451 parse_with_tz = self.func( 452 "PARSE_DATETIME", 453 value_as_text, 454 self.format_time(expression, Hive.INVERSE_TIME_MAPPING, Hive.INVERSE_TIME_TRIE), 455 ) 456 coalesced = self.func("COALESCE", self.func("TRY", parse_without_tz), parse_with_tz) 457 return self.func("TO_UNIXTIME", coalesced) 458 459 def bracket_sql(self, expression: exp.Bracket) -> str: 460 if expression.args.get("safe"): 461 return self.func( 462 "ELEMENT_AT", 463 expression.this, 464 seq_get( 465 apply_index_offset( 466 expression.this, 467 expression.expressions, 468 1 - expression.args.get("offset", 0), 469 ), 470 0, 471 ), 472 ) 473 return super().bracket_sql(expression) 474 475 def struct_sql(self, expression: exp.Struct) -> str: 476 from sqlglot.optimizer.annotate_types import annotate_types 477 478 expression = annotate_types(expression) 479 values: t.List[str] = [] 480 schema: t.List[str] = [] 481 unknown_type = False 482 483 for e in expression.expressions: 484 if isinstance(e, exp.PropertyEQ): 485 if e.type and e.type.is_type(exp.DataType.Type.UNKNOWN): 486 unknown_type = True 487 else: 488 schema.append(f"{self.sql(e, 'this')} {self.sql(e.type)}") 489 values.append(self.sql(e, "expression")) 490 else: 491 values.append(self.sql(e)) 492 493 size = len(expression.expressions) 494 495 if not size or len(schema) != size: 496 if unknown_type: 497 self.unsupported( 498 "Cannot convert untyped key-value definitions (try annotate_types)." 499 ) 500 return self.func("ROW", *values) 501 return f"CAST(ROW({', '.join(values)}) AS ROW({', '.join(schema)}))" 502 503 def interval_sql(self, expression: exp.Interval) -> str: 504 if expression.this and expression.text("unit").upper().startswith("WEEK"): 505 return f"({expression.this.name} * INTERVAL '7' DAY)" 506 return super().interval_sql(expression) 507 508 def transaction_sql(self, expression: exp.Transaction) -> str: 509 modes = expression.args.get("modes") 510 modes = f" {', '.join(modes)}" if modes else "" 511 return f"START TRANSACTION{modes}" 512 513 def generateseries_sql(self, expression: exp.GenerateSeries) -> str: 514 start = expression.args["start"] 515 end = expression.args["end"] 516 step = expression.args.get("step") 517 518 if isinstance(start, exp.Cast): 519 target_type = start.to 520 elif isinstance(end, exp.Cast): 521 target_type = end.to 522 else: 523 target_type = None 524 525 if target_type and target_type.is_type("timestamp"): 526 if target_type is start.to: 527 end = exp.cast(end, target_type) 528 else: 529 start = exp.cast(start, target_type) 530 531 return self.func("SEQUENCE", start, end, step) 532 533 def offset_limit_modifiers( 534 self, expression: exp.Expression, fetch: bool, limit: t.Optional[exp.Fetch | exp.Limit] 535 ) -> t.List[str]: 536 return [ 537 self.sql(expression, "offset"), 538 self.sql(limit), 539 ] 540 541 def create_sql(self, expression: exp.Create) -> str: 542 """ 543 Presto doesn't support CREATE VIEW with expressions (ex: `CREATE VIEW x (cola)` then `(cola)` is the expression), 544 so we need to remove them 545 """ 546 kind = expression.args["kind"] 547 schema = expression.this 548 if kind == "VIEW" and schema.expressions: 549 expression.this.set("expressions", None) 550 return super().create_sql(expression)
Generator converts a given syntax tree to the corresponding SQL string.
Arguments:
- pretty: Whether to format the produced SQL string. Default: False.
- identify: Determines when an identifier should be quoted. Possible values are: False (default): Never quote, except in cases where it's mandatory by the dialect. True or 'always': Always quote. 'safe': Only quote identifiers that are case insensitive.
- normalize: Whether to normalize identifiers to lowercase. Default: False.
- pad: The pad size in a formatted string. For example, this affects the indentation of a projection in a query, relative to its nesting level. Default: 2.
- indent: The indentation size in a formatted string. For example, this affects the
indentation of subqueries and filters under a
WHERE
clause. Default: 2. - normalize_functions: How to normalize function names. Possible values are: "upper" or True (default): Convert names to uppercase. "lower": Convert names to lowercase. False: Disables function name normalization.
- unsupported_level: Determines the generator's behavior when it encounters unsupported expressions. Default ErrorLevel.WARN.
- max_unsupported: Maximum number of unsupported messages to include in a raised UnsupportedError. This is only relevant if unsupported_level is ErrorLevel.RAISE. Default: 3
- leading_comma: Whether the comma is leading or trailing in select expressions. This is only relevant when generating in pretty mode. Default: False
- max_text_width: The max number of characters in a segment before creating new lines in pretty mode. The default is on the smaller end because the length only represents a segment and not the true line length. Default: 80
- comments: Whether to preserve comments in the output SQL code. Default: True
443 def strtounix_sql(self, expression: exp.StrToUnix) -> str: 444 # Since `TO_UNIXTIME` requires a `TIMESTAMP`, we need to parse the argument into one. 445 # To do this, we first try to `DATE_PARSE` it, but since this can fail when there's a 446 # timezone involved, we wrap it in a `TRY` call and use `PARSE_DATETIME` as a fallback, 447 # which seems to be using the same time mapping as Hive, as per: 448 # https://joda-time.sourceforge.net/apidocs/org/joda/time/format/DateTimeFormat.html 449 value_as_text = exp.cast(expression.this, "text") 450 parse_without_tz = self.func("DATE_PARSE", value_as_text, self.format_time(expression)) 451 parse_with_tz = self.func( 452 "PARSE_DATETIME", 453 value_as_text, 454 self.format_time(expression, Hive.INVERSE_TIME_MAPPING, Hive.INVERSE_TIME_TRIE), 455 ) 456 coalesced = self.func("COALESCE", self.func("TRY", parse_without_tz), parse_with_tz) 457 return self.func("TO_UNIXTIME", coalesced)
459 def bracket_sql(self, expression: exp.Bracket) -> str: 460 if expression.args.get("safe"): 461 return self.func( 462 "ELEMENT_AT", 463 expression.this, 464 seq_get( 465 apply_index_offset( 466 expression.this, 467 expression.expressions, 468 1 - expression.args.get("offset", 0), 469 ), 470 0, 471 ), 472 ) 473 return super().bracket_sql(expression)
475 def struct_sql(self, expression: exp.Struct) -> str: 476 from sqlglot.optimizer.annotate_types import annotate_types 477 478 expression = annotate_types(expression) 479 values: t.List[str] = [] 480 schema: t.List[str] = [] 481 unknown_type = False 482 483 for e in expression.expressions: 484 if isinstance(e, exp.PropertyEQ): 485 if e.type and e.type.is_type(exp.DataType.Type.UNKNOWN): 486 unknown_type = True 487 else: 488 schema.append(f"{self.sql(e, 'this')} {self.sql(e.type)}") 489 values.append(self.sql(e, "expression")) 490 else: 491 values.append(self.sql(e)) 492 493 size = len(expression.expressions) 494 495 if not size or len(schema) != size: 496 if unknown_type: 497 self.unsupported( 498 "Cannot convert untyped key-value definitions (try annotate_types)." 499 ) 500 return self.func("ROW", *values) 501 return f"CAST(ROW({', '.join(values)}) AS ROW({', '.join(schema)}))"
513 def generateseries_sql(self, expression: exp.GenerateSeries) -> str: 514 start = expression.args["start"] 515 end = expression.args["end"] 516 step = expression.args.get("step") 517 518 if isinstance(start, exp.Cast): 519 target_type = start.to 520 elif isinstance(end, exp.Cast): 521 target_type = end.to 522 else: 523 target_type = None 524 525 if target_type and target_type.is_type("timestamp"): 526 if target_type is start.to: 527 end = exp.cast(end, target_type) 528 else: 529 start = exp.cast(start, target_type) 530 531 return self.func("SEQUENCE", start, end, step)
541 def create_sql(self, expression: exp.Create) -> str: 542 """ 543 Presto doesn't support CREATE VIEW with expressions (ex: `CREATE VIEW x (cola)` then `(cola)` is the expression), 544 so we need to remove them 545 """ 546 kind = expression.args["kind"] 547 schema = expression.this 548 if kind == "VIEW" and schema.expressions: 549 expression.this.set("expressions", None) 550 return super().create_sql(expression)
Presto doesn't support CREATE VIEW with expressions (ex: CREATE VIEW x (cola)
then (cola)
is the expression),
so we need to remove them
Inherited Members
- sqlglot.generator.Generator
- Generator
- NULL_ORDERING_SUPPORTED
- IGNORE_NULLS_IN_FUNC
- LOCKING_READS_SUPPORTED
- EXPLICIT_UNION
- WRAP_DERIVED_VALUES
- CREATE_FUNCTION_RETURN_AS
- MATCHED_BY_SOURCE
- SINGLE_STRING_INTERVAL
- LIMIT_FETCH
- RENAME_TABLE_WITH_DB
- GROUPINGS_SEP
- INDEX_ON
- QUERY_HINT_SEP
- DUPLICATE_KEY_UPDATE_WITH_SET
- LIMIT_IS_TOP
- RETURNING_END
- COLUMN_JOIN_MARKS_SUPPORTED
- EXTRACT_ALLOWS_QUOTES
- VALUES_AS_TABLE
- ALTER_TABLE_INCLUDE_COLUMN_KEYWORD
- UNNEST_WITH_ORDINALITY
- AGGREGATE_FILTER_SUPPORTED
- SEMI_ANTI_JOIN_WITH_SIDE
- COMPUTED_COLUMN_WITH_TYPE
- SUPPORTS_TABLE_COPY
- TABLESAMPLE_REQUIRES_PARENS
- TABLESAMPLE_SIZE_IS_ROWS
- TABLESAMPLE_KEYWORDS
- TABLESAMPLE_WITH_METHOD
- TABLESAMPLE_SEED_KEYWORD
- COLLATE_IS_FUNC
- DATA_TYPE_SPECIFIERS_ALLOWED
- ENSURE_BOOLS
- CTE_RECURSIVE_KEYWORD_REQUIRED
- LAST_DAY_SUPPORTS_DATE_PART
- SUPPORTS_TABLE_ALIAS_COLUMNS
- UNPIVOT_ALIASES_ARE_IDENTIFIERS
- JSON_KEY_VALUE_PAIR_SEP
- INSERT_OVERWRITE
- SUPPORTS_SELECT_INTO
- SUPPORTS_UNLOGGED_TABLES
- SUPPORTS_CREATE_TABLE_LIKE
- JSON_TYPE_REQUIRED_FOR_EXTRACTION
- JSON_PATH_BRACKETED_KEY_SUPPORTED
- JSON_PATH_SINGLE_QUOTE_ESCAPE
- SUPPORTED_JSON_PATH_PARTS
- CAN_IMPLEMENT_ARRAY_ANY
- STAR_MAPPING
- TIME_PART_SINGULARS
- TOKEN_MAPPING
- PARAMETER_TOKEN
- NAMED_PLACEHOLDER_TOKEN
- RESERVED_KEYWORDS
- WITH_SEPARATED_COMMENTS
- EXCLUDE_COMMENTS
- UNWRAPPED_INTERVAL_VALUES
- PARAMETERIZABLE_TEXT_TYPES
- EXPRESSIONS_WITHOUT_NESTED_CTES
- SENTINEL_LINE_BREAK
- pretty
- identify
- normalize
- pad
- unsupported_level
- max_unsupported
- leading_comma
- max_text_width
- comments
- dialect
- normalize_functions
- unsupported_messages
- generate
- preprocess
- unsupported
- sep
- seg
- pad_comment
- maybe_comment
- wrap
- no_identify
- normalize_func
- indent
- sql
- uncache_sql
- cache_sql
- characterset_sql
- column_sql
- columnposition_sql
- columndef_sql
- columnconstraint_sql
- computedcolumnconstraint_sql
- autoincrementcolumnconstraint_sql
- compresscolumnconstraint_sql
- generatedasidentitycolumnconstraint_sql
- generatedasrowcolumnconstraint_sql
- periodforsystemtimeconstraint_sql
- notnullcolumnconstraint_sql
- transformcolumnconstraint_sql
- primarykeycolumnconstraint_sql
- uniquecolumnconstraint_sql
- createable_sql
- sequenceproperties_sql
- clone_sql
- describe_sql
- heredoc_sql
- prepend_ctes
- with_sql
- cte_sql
- tablealias_sql
- bitstring_sql
- hexstring_sql
- bytestring_sql
- unicodestring_sql
- rawstring_sql
- datatypeparam_sql
- datatype_sql
- directory_sql
- delete_sql
- drop_sql
- except_sql
- except_op
- fetch_sql
- filter_sql
- hint_sql
- indexparameters_sql
- index_sql
- identifier_sql
- inputoutputformat_sql
- national_sql
- partition_sql
- properties_sql
- root_properties
- properties
- with_properties
- locate_properties
- property_name
- property_sql
- likeproperty_sql
- fallbackproperty_sql
- journalproperty_sql
- freespaceproperty_sql
- checksumproperty_sql
- mergeblockratioproperty_sql
- datablocksizeproperty_sql
- blockcompressionproperty_sql
- isolatedloadingproperty_sql
- partitionboundspec_sql
- partitionedofproperty_sql
- lockingproperty_sql
- withdataproperty_sql
- withsystemversioningproperty_sql
- insert_sql
- intersect_sql
- intersect_op
- introducer_sql
- kill_sql
- pseudotype_sql
- objectidentifier_sql
- onconflict_sql
- returning_sql
- rowformatdelimitedproperty_sql
- withtablehint_sql
- indextablehint_sql
- historicaldata_sql
- table_parts
- table_sql
- tablesample_sql
- pivot_sql
- version_sql
- tuple_sql
- update_sql
- values_sql
- var_sql
- into_sql
- from_sql
- group_sql
- having_sql
- connect_sql
- prior_sql
- join_sql
- lambda_sql
- lateral_op
- lateral_sql
- limit_sql
- offset_sql
- setitem_sql
- set_sql
- pragma_sql
- lock_sql
- literal_sql
- escape_str
- loaddata_sql
- null_sql
- boolean_sql
- order_sql
- withfill_sql
- cluster_sql
- distribute_sql
- sort_sql
- ordered_sql
- matchrecognize_sql
- query_modifiers
- queryoption_sql
- after_limit_modifiers
- select_sql
- schema_sql
- schema_columns_sql
- star_sql
- parameter_sql
- sessionparameter_sql
- placeholder_sql
- subquery_sql
- qualify_sql
- set_operations
- union_sql
- union_op
- unnest_sql
- prewhere_sql
- where_sql
- window_sql
- partition_by_sql
- windowspec_sql
- withingroup_sql
- between_sql
- all_sql
- any_sql
- exists_sql
- case_sql
- constraint_sql
- nextvaluefor_sql
- extract_sql
- trim_sql
- convert_concat_args
- concat_sql
- concatws_sql
- check_sql
- foreignkey_sql
- primarykey_sql
- if_sql
- matchagainst_sql
- jsonkeyvalue_sql
- jsonpath_sql
- json_path_part
- formatjson_sql
- jsonobject_sql
- jsonobjectagg_sql
- jsonarray_sql
- jsonarrayagg_sql
- jsoncolumndef_sql
- jsonschema_sql
- jsontable_sql
- openjsoncolumndef_sql
- openjson_sql
- in_sql
- in_unnest_op
- return_sql
- reference_sql
- anonymous_sql
- paren_sql
- neg_sql
- not_sql
- alias_sql
- pivotalias_sql
- aliases_sql
- atindex_sql
- attimezone_sql
- fromtimezone_sql
- add_sql
- and_sql
- or_sql
- xor_sql
- connector_sql
- bitwiseand_sql
- bitwiseleftshift_sql
- bitwisenot_sql
- bitwiseor_sql
- bitwiserightshift_sql
- bitwisexor_sql
- cast_sql
- currentdate_sql
- currenttimestamp_sql
- collate_sql
- command_sql
- comment_sql
- mergetreettlaction_sql
- mergetreettl_sql
- commit_sql
- rollback_sql
- altercolumn_sql
- renametable_sql
- renamecolumn_sql
- altertable_sql
- add_column_sql
- droppartition_sql
- addconstraint_sql
- distinct_sql
- ignorenulls_sql
- respectnulls_sql
- havingmax_sql
- intdiv_sql
- dpipe_sql
- div_sql
- overlaps_sql
- distance_sql
- dot_sql
- eq_sql
- propertyeq_sql
- escape_sql
- glob_sql
- gt_sql
- gte_sql
- ilike_sql
- ilikeany_sql
- is_sql
- like_sql
- likeany_sql
- similarto_sql
- lt_sql
- lte_sql
- mod_sql
- mul_sql
- neq_sql
- nullsafeeq_sql
- nullsafeneq_sql
- slice_sql
- sub_sql
- trycast_sql
- log_sql
- use_sql
- binary
- function_fallback_sql
- func
- format_args
- text_width
- format_time
- expressions
- op_expressions
- naked_property
- tag_sql
- token_sql
- userdefinedfunction_sql
- joinhint_sql
- kwarg_sql
- when_sql
- merge_sql
- tochar_sql
- tonumber_sql
- dictproperty_sql
- dictrange_sql
- dictsubproperty_sql
- oncluster_sql
- clusteredbyproperty_sql
- anyvalue_sql
- querytransform_sql
- indexconstraintoption_sql
- checkcolumnconstraint_sql
- indexcolumnconstraint_sql
- nvl2_sql
- comprehension_sql
- columnprefix_sql
- opclass_sql
- predict_sql
- forin_sql
- refresh_sql
- operator_sql
- toarray_sql
- tsordstotime_sql
- tsordstotimestamp_sql
- tsordstodate_sql
- unixdate_sql
- lastday_sql
- dateadd_sql
- arrayany_sql
- partitionrange_sql
- truncatetable_sql
- convert_sql