sqlglot.dialects.duckdb

   1from __future__ import annotations
   2
   3import typing as t
   4
   5from sqlglot import exp, generator, parser, tokens, transforms
   6
   7from sqlglot.dialects.dialect import (
   8    Dialect,
   9    JSON_EXTRACT_TYPE,
  10    NormalizationStrategy,
  11    Version,
  12    approx_count_distinct_sql,
  13    arrow_json_extract_sql,
  14    binary_from_function,
  15    bool_xor_sql,
  16    build_default_decimal_type,
  17    count_if_to_sum,
  18    date_delta_to_binary_interval_op,
  19    date_trunc_to_time,
  20    datestrtodate_sql,
  21    no_datetime_sql,
  22    encode_decode_sql,
  23    build_formatted_time,
  24    inline_array_unless_query,
  25    no_comment_column_constraint_sql,
  26    no_time_sql,
  27    no_timestamp_sql,
  28    pivot_column_names,
  29    rename_func,
  30    remove_from_array_using_filter,
  31    strposition_sql,
  32    str_to_time_sql,
  33    timestamptrunc_sql,
  34    timestrtotime_sql,
  35    unit_to_str,
  36    sha256_sql,
  37    build_regexp_extract,
  38    explode_to_unnest_sql,
  39    no_make_interval_sql,
  40    groupconcat_sql,
  41)
  42from sqlglot.generator import unsupported_args
  43from sqlglot.helper import seq_get
  44from sqlglot.tokens import TokenType
  45from sqlglot.parser import binary_range_parser
  46
  47
  48# BigQuery -> DuckDB conversion for the DATE function
  49def _date_sql(self: DuckDB.Generator, expression: exp.Date) -> str:
  50    result = f"CAST({self.sql(expression, 'this')} AS DATE)"
  51    zone = self.sql(expression, "zone")
  52
  53    if zone:
  54        date_str = self.func("STRFTIME", result, "'%d/%m/%Y'")
  55        date_str = f"{date_str} || ' ' || {zone}"
  56
  57        # This will create a TIMESTAMP with time zone information
  58        result = self.func("STRPTIME", date_str, "'%d/%m/%Y %Z'")
  59
  60    return result
  61
  62
  63# BigQuery -> DuckDB conversion for the TIME_DIFF function
  64def _timediff_sql(self: DuckDB.Generator, expression: exp.TimeDiff) -> str:
  65    this = exp.cast(expression.this, exp.DataType.Type.TIME)
  66    expr = exp.cast(expression.expression, exp.DataType.Type.TIME)
  67
  68    # Although the 2 dialects share similar signatures, BQ seems to inverse
  69    # the sign of the result so the start/end time operands are flipped
  70    return self.func("DATE_DIFF", unit_to_str(expression), expr, this)
  71
  72
  73@unsupported_args(("expression", "DuckDB's ARRAY_SORT does not support a comparator."))
  74def _array_sort_sql(self: DuckDB.Generator, expression: exp.ArraySort) -> str:
  75    return self.func("ARRAY_SORT", expression.this)
  76
  77
  78def _sort_array_sql(self: DuckDB.Generator, expression: exp.SortArray) -> str:
  79    name = "ARRAY_REVERSE_SORT" if expression.args.get("asc") == exp.false() else "ARRAY_SORT"
  80    return self.func(name, expression.this)
  81
  82
  83def _build_sort_array_desc(args: t.List) -> exp.Expression:
  84    return exp.SortArray(this=seq_get(args, 0), asc=exp.false())
  85
  86
  87def _build_date_diff(args: t.List) -> exp.Expression:
  88    return exp.DateDiff(this=seq_get(args, 2), expression=seq_get(args, 1), unit=seq_get(args, 0))
  89
  90
  91def _build_generate_series(end_exclusive: bool = False) -> t.Callable[[t.List], exp.GenerateSeries]:
  92    def _builder(args: t.List) -> exp.GenerateSeries:
  93        # Check https://duckdb.org/docs/sql/functions/nested.html#range-functions
  94        if len(args) == 1:
  95            # DuckDB uses 0 as a default for the series' start when it's omitted
  96            args.insert(0, exp.Literal.number("0"))
  97
  98        gen_series = exp.GenerateSeries.from_arg_list(args)
  99        gen_series.set("is_end_exclusive", end_exclusive)
 100
 101        return gen_series
 102
 103    return _builder
 104
 105
 106def _build_make_timestamp(args: t.List) -> exp.Expression:
 107    if len(args) == 1:
 108        return exp.UnixToTime(this=seq_get(args, 0), scale=exp.UnixToTime.MICROS)
 109
 110    return exp.TimestampFromParts(
 111        year=seq_get(args, 0),
 112        month=seq_get(args, 1),
 113        day=seq_get(args, 2),
 114        hour=seq_get(args, 3),
 115        min=seq_get(args, 4),
 116        sec=seq_get(args, 5),
 117    )
 118
 119
 120def _show_parser(*args: t.Any, **kwargs: t.Any) -> t.Callable[[DuckDB.Parser], exp.Show]:
 121    def _parse(self: DuckDB.Parser) -> exp.Show:
 122        return self._parse_show_duckdb(*args, **kwargs)
 123
 124    return _parse
 125
 126
 127def _struct_sql(self: DuckDB.Generator, expression: exp.Struct) -> str:
 128    args: t.List[str] = []
 129
 130    # BigQuery allows inline construction such as "STRUCT<a STRING, b INTEGER>('str', 1)" which is
 131    # canonicalized to "ROW('str', 1) AS STRUCT(a TEXT, b INT)" in DuckDB
 132    # The transformation to ROW will take place if:
 133    #  1. The STRUCT itself does not have proper fields (key := value) as a "proper" STRUCT would
 134    #  2. A cast to STRUCT / ARRAY of STRUCTs is found
 135    ancestor_cast = expression.find_ancestor(exp.Cast)
 136    is_bq_inline_struct = (
 137        (expression.find(exp.PropertyEQ) is None)
 138        and ancestor_cast
 139        and any(
 140            casted_type.is_type(exp.DataType.Type.STRUCT)
 141            for casted_type in ancestor_cast.find_all(exp.DataType)
 142        )
 143    )
 144
 145    for i, expr in enumerate(expression.expressions):
 146        is_property_eq = isinstance(expr, exp.PropertyEQ)
 147        value = expr.expression if is_property_eq else expr
 148
 149        if is_bq_inline_struct:
 150            args.append(self.sql(value))
 151        else:
 152            if is_property_eq:
 153                if isinstance(expr.this, exp.Identifier):
 154                    key = self.sql(exp.Literal.string(expr.name))
 155                else:
 156                    key = self.sql(expr.this)
 157            else:
 158                key = self.sql(exp.Literal.string(f"_{i}"))
 159
 160            args.append(f"{key}: {self.sql(value)}")
 161
 162    csv_args = ", ".join(args)
 163
 164    return f"ROW({csv_args})" if is_bq_inline_struct else f"{{{csv_args}}}"
 165
 166
 167def _datatype_sql(self: DuckDB.Generator, expression: exp.DataType) -> str:
 168    if expression.is_type("array"):
 169        return f"{self.expressions(expression, flat=True)}[{self.expressions(expression, key='values', flat=True)}]"
 170
 171    # Modifiers are not supported for TIME, [TIME | TIMESTAMP] WITH TIME ZONE
 172    if expression.is_type(
 173        exp.DataType.Type.TIME, exp.DataType.Type.TIMETZ, exp.DataType.Type.TIMESTAMPTZ
 174    ):
 175        return expression.this.value
 176
 177    return self.datatype_sql(expression)
 178
 179
 180def _json_format_sql(self: DuckDB.Generator, expression: exp.JSONFormat) -> str:
 181    sql = self.func("TO_JSON", expression.this, expression.args.get("options"))
 182    return f"CAST({sql} AS TEXT)"
 183
 184
 185def _unix_to_time_sql(self: DuckDB.Generator, expression: exp.UnixToTime) -> str:
 186    scale = expression.args.get("scale")
 187    timestamp = expression.this
 188
 189    if scale in (None, exp.UnixToTime.SECONDS):
 190        return self.func("TO_TIMESTAMP", timestamp)
 191    if scale == exp.UnixToTime.MILLIS:
 192        return self.func("EPOCH_MS", timestamp)
 193    if scale == exp.UnixToTime.MICROS:
 194        return self.func("MAKE_TIMESTAMP", timestamp)
 195
 196    return self.func("TO_TIMESTAMP", exp.Div(this=timestamp, expression=exp.func("POW", 10, scale)))
 197
 198
 199WRAPPED_JSON_EXTRACT_EXPRESSIONS = (exp.Binary, exp.Bracket, exp.In)
 200
 201
 202def _arrow_json_extract_sql(self: DuckDB.Generator, expression: JSON_EXTRACT_TYPE) -> str:
 203    arrow_sql = arrow_json_extract_sql(self, expression)
 204    if not expression.same_parent and isinstance(
 205        expression.parent, WRAPPED_JSON_EXTRACT_EXPRESSIONS
 206    ):
 207        arrow_sql = self.wrap(arrow_sql)
 208    return arrow_sql
 209
 210
 211def _implicit_datetime_cast(
 212    arg: t.Optional[exp.Expression], type: exp.DataType.Type = exp.DataType.Type.DATE
 213) -> t.Optional[exp.Expression]:
 214    return exp.cast(arg, type) if isinstance(arg, exp.Literal) else arg
 215
 216
 217def _date_diff_sql(self: DuckDB.Generator, expression: exp.DateDiff) -> str:
 218    this = _implicit_datetime_cast(expression.this)
 219    expr = _implicit_datetime_cast(expression.expression)
 220
 221    return self.func("DATE_DIFF", unit_to_str(expression), expr, this)
 222
 223
 224def _generate_datetime_array_sql(
 225    self: DuckDB.Generator, expression: t.Union[exp.GenerateDateArray, exp.GenerateTimestampArray]
 226) -> str:
 227    is_generate_date_array = isinstance(expression, exp.GenerateDateArray)
 228
 229    type = exp.DataType.Type.DATE if is_generate_date_array else exp.DataType.Type.TIMESTAMP
 230    start = _implicit_datetime_cast(expression.args.get("start"), type=type)
 231    end = _implicit_datetime_cast(expression.args.get("end"), type=type)
 232
 233    # BQ's GENERATE_DATE_ARRAY & GENERATE_TIMESTAMP_ARRAY are transformed to DuckDB'S GENERATE_SERIES
 234    gen_series: t.Union[exp.GenerateSeries, exp.Cast] = exp.GenerateSeries(
 235        start=start, end=end, step=expression.args.get("step")
 236    )
 237
 238    if is_generate_date_array:
 239        # The GENERATE_SERIES result type is TIMESTAMP array, so to match BQ's semantics for
 240        # GENERATE_DATE_ARRAY we must cast it back to DATE array
 241        gen_series = exp.cast(gen_series, exp.DataType.build("ARRAY<DATE>"))
 242
 243    return self.sql(gen_series)
 244
 245
 246def _json_extract_value_array_sql(
 247    self: DuckDB.Generator, expression: exp.JSONValueArray | exp.JSONExtractArray
 248) -> str:
 249    json_extract = exp.JSONExtract(this=expression.this, expression=expression.expression)
 250    data_type = "ARRAY<STRING>" if isinstance(expression, exp.JSONValueArray) else "ARRAY<JSON>"
 251    return self.sql(exp.cast(json_extract, to=exp.DataType.build(data_type)))
 252
 253
 254class DuckDB(Dialect):
 255    NULL_ORDERING = "nulls_are_last"
 256    SUPPORTS_USER_DEFINED_TYPES = True
 257    SAFE_DIVISION = True
 258    INDEX_OFFSET = 1
 259    CONCAT_COALESCE = True
 260    SUPPORTS_ORDER_BY_ALL = True
 261    SUPPORTS_FIXED_SIZE_ARRAYS = True
 262    STRICT_JSON_PATH_SYNTAX = False
 263    NUMBERS_CAN_BE_UNDERSCORE_SEPARATED = True
 264
 265    # https://duckdb.org/docs/sql/introduction.html#creating-a-new-table
 266    NORMALIZATION_STRATEGY = NormalizationStrategy.CASE_INSENSITIVE
 267
 268    DATE_PART_MAPPING = {
 269        **Dialect.DATE_PART_MAPPING,
 270        "DAYOFWEEKISO": "ISODOW",
 271    }
 272    DATE_PART_MAPPING.pop("WEEKDAY")
 273
 274    def to_json_path(self, path: t.Optional[exp.Expression]) -> t.Optional[exp.Expression]:
 275        if isinstance(path, exp.Literal):
 276            # DuckDB also supports the JSON pointer syntax, where every path starts with a `/`.
 277            # Additionally, it allows accessing the back of lists using the `[#-i]` syntax.
 278            # This check ensures we'll avoid trying to parse these as JSON paths, which can
 279            # either result in a noisy warning or in an invalid representation of the path.
 280            path_text = path.name
 281            if path_text.startswith("/") or "[#" in path_text:
 282                return path
 283
 284        return super().to_json_path(path)
 285
 286    class Tokenizer(tokens.Tokenizer):
 287        BYTE_STRINGS = [("e'", "'"), ("E'", "'")]
 288        HEREDOC_STRINGS = ["$"]
 289
 290        HEREDOC_TAG_IS_IDENTIFIER = True
 291        HEREDOC_STRING_ALTERNATIVE = TokenType.PARAMETER
 292
 293        KEYWORDS = {
 294            **tokens.Tokenizer.KEYWORDS,
 295            "//": TokenType.DIV,
 296            "**": TokenType.DSTAR,
 297            "^@": TokenType.CARET_AT,
 298            "@>": TokenType.AT_GT,
 299            "<@": TokenType.LT_AT,
 300            "ATTACH": TokenType.ATTACH,
 301            "BINARY": TokenType.VARBINARY,
 302            "BITSTRING": TokenType.BIT,
 303            "BPCHAR": TokenType.TEXT,
 304            "CHAR": TokenType.TEXT,
 305            "DATETIME": TokenType.TIMESTAMPNTZ,
 306            "DETACH": TokenType.DETACH,
 307            "EXCLUDE": TokenType.EXCEPT,
 308            "LOGICAL": TokenType.BOOLEAN,
 309            "ONLY": TokenType.ONLY,
 310            "PIVOT_WIDER": TokenType.PIVOT,
 311            "POSITIONAL": TokenType.POSITIONAL,
 312            "RESET": TokenType.COMMAND,
 313            "SIGNED": TokenType.INT,
 314            "STRING": TokenType.TEXT,
 315            "SUMMARIZE": TokenType.SUMMARIZE,
 316            "TIMESTAMP": TokenType.TIMESTAMPNTZ,
 317            "TIMESTAMP_S": TokenType.TIMESTAMP_S,
 318            "TIMESTAMP_MS": TokenType.TIMESTAMP_MS,
 319            "TIMESTAMP_NS": TokenType.TIMESTAMP_NS,
 320            "TIMESTAMP_US": TokenType.TIMESTAMP,
 321            "UBIGINT": TokenType.UBIGINT,
 322            "UINTEGER": TokenType.UINT,
 323            "USMALLINT": TokenType.USMALLINT,
 324            "UTINYINT": TokenType.UTINYINT,
 325            "VARCHAR": TokenType.TEXT,
 326        }
 327        KEYWORDS.pop("/*+")
 328
 329        SINGLE_TOKENS = {
 330            **tokens.Tokenizer.SINGLE_TOKENS,
 331            "$": TokenType.PARAMETER,
 332        }
 333
 334        COMMANDS = tokens.Tokenizer.COMMANDS - {TokenType.SHOW}
 335
 336    class Parser(parser.Parser):
 337        MAP_KEYS_ARE_ARBITRARY_EXPRESSIONS = True
 338
 339        BITWISE = {
 340            **parser.Parser.BITWISE,
 341            TokenType.TILDA: exp.RegexpLike,
 342        }
 343        BITWISE.pop(TokenType.CARET)
 344
 345        RANGE_PARSERS = {
 346            **parser.Parser.RANGE_PARSERS,
 347            TokenType.DAMP: binary_range_parser(exp.ArrayOverlaps),
 348            TokenType.CARET_AT: binary_range_parser(exp.StartsWith),
 349        }
 350
 351        EXPONENT = {
 352            **parser.Parser.EXPONENT,
 353            TokenType.CARET: exp.Pow,
 354            TokenType.DSTAR: exp.Pow,
 355        }
 356
 357        FUNCTIONS_WITH_ALIASED_ARGS = {*parser.Parser.FUNCTIONS_WITH_ALIASED_ARGS, "STRUCT_PACK"}
 358
 359        SHOW_PARSERS = {
 360            "TABLES": _show_parser("TABLES"),
 361            "ALL TABLES": _show_parser("ALL TABLES"),
 362        }
 363
 364        FUNCTIONS = {
 365            **parser.Parser.FUNCTIONS,
 366            "ANY_VALUE": lambda args: exp.IgnoreNulls(this=exp.AnyValue.from_arg_list(args)),
 367            "ARRAY_REVERSE_SORT": _build_sort_array_desc,
 368            "ARRAY_SORT": exp.SortArray.from_arg_list,
 369            "DATEDIFF": _build_date_diff,
 370            "DATE_DIFF": _build_date_diff,
 371            "DATE_TRUNC": date_trunc_to_time,
 372            "DATETRUNC": date_trunc_to_time,
 373            "DECODE": lambda args: exp.Decode(
 374                this=seq_get(args, 0), charset=exp.Literal.string("utf-8")
 375            ),
 376            "EDITDIST3": exp.Levenshtein.from_arg_list,
 377            "ENCODE": lambda args: exp.Encode(
 378                this=seq_get(args, 0), charset=exp.Literal.string("utf-8")
 379            ),
 380            "EPOCH": exp.TimeToUnix.from_arg_list,
 381            "EPOCH_MS": lambda args: exp.UnixToTime(
 382                this=seq_get(args, 0), scale=exp.UnixToTime.MILLIS
 383            ),
 384            "GENERATE_SERIES": _build_generate_series(),
 385            "JSON": exp.ParseJSON.from_arg_list,
 386            "JSON_EXTRACT_PATH": parser.build_extract_json_with_path(exp.JSONExtract),
 387            "JSON_EXTRACT_STRING": parser.build_extract_json_with_path(exp.JSONExtractScalar),
 388            "LIST_CONTAINS": exp.ArrayContains.from_arg_list,
 389            "LIST_FILTER": exp.ArrayFilter.from_arg_list,
 390            "LIST_HAS": exp.ArrayContains.from_arg_list,
 391            "LIST_HAS_ANY": exp.ArrayOverlaps.from_arg_list,
 392            "LIST_REVERSE_SORT": _build_sort_array_desc,
 393            "LIST_SORT": exp.SortArray.from_arg_list,
 394            "LIST_TRANSFORM": exp.Transform.from_arg_list,
 395            "LIST_VALUE": lambda args: exp.Array(expressions=args),
 396            "MAKE_TIME": exp.TimeFromParts.from_arg_list,
 397            "MAKE_TIMESTAMP": _build_make_timestamp,
 398            "QUANTILE_CONT": exp.PercentileCont.from_arg_list,
 399            "QUANTILE_DISC": exp.PercentileDisc.from_arg_list,
 400            "RANGE": _build_generate_series(end_exclusive=True),
 401            "REGEXP_EXTRACT": build_regexp_extract(exp.RegexpExtract),
 402            "REGEXP_EXTRACT_ALL": build_regexp_extract(exp.RegexpExtractAll),
 403            "REGEXP_MATCHES": exp.RegexpLike.from_arg_list,
 404            "REGEXP_REPLACE": lambda args: exp.RegexpReplace(
 405                this=seq_get(args, 0),
 406                expression=seq_get(args, 1),
 407                replacement=seq_get(args, 2),
 408                modifiers=seq_get(args, 3),
 409            ),
 410            "SHA256": lambda args: exp.SHA2(this=seq_get(args, 0), length=exp.Literal.number(256)),
 411            "STRFTIME": build_formatted_time(exp.TimeToStr, "duckdb"),
 412            "STRING_SPLIT": exp.Split.from_arg_list,
 413            "STRING_SPLIT_REGEX": exp.RegexpSplit.from_arg_list,
 414            "STRING_TO_ARRAY": exp.Split.from_arg_list,
 415            "STRPTIME": build_formatted_time(exp.StrToTime, "duckdb"),
 416            "STRUCT_PACK": exp.Struct.from_arg_list,
 417            "STR_SPLIT": exp.Split.from_arg_list,
 418            "STR_SPLIT_REGEX": exp.RegexpSplit.from_arg_list,
 419            "TIME_BUCKET": exp.DateBin.from_arg_list,
 420            "TO_TIMESTAMP": exp.UnixToTime.from_arg_list,
 421            "UNNEST": exp.Explode.from_arg_list,
 422            "XOR": binary_from_function(exp.BitwiseXor),
 423        }
 424
 425        FUNCTIONS.pop("DATE_SUB")
 426        FUNCTIONS.pop("GLOB")
 427
 428        FUNCTION_PARSERS = {
 429            **parser.Parser.FUNCTION_PARSERS,
 430            **dict.fromkeys(
 431                ("GROUP_CONCAT", "LISTAGG", "STRINGAGG"), lambda self: self._parse_string_agg()
 432            ),
 433        }
 434        FUNCTION_PARSERS.pop("DECODE")
 435
 436        NO_PAREN_FUNCTION_PARSERS = {
 437            **parser.Parser.NO_PAREN_FUNCTION_PARSERS,
 438            "MAP": lambda self: self._parse_map(),
 439            "@": lambda self: exp.Abs(this=self._parse_bitwise()),
 440        }
 441
 442        TABLE_ALIAS_TOKENS = parser.Parser.TABLE_ALIAS_TOKENS - {
 443            TokenType.SEMI,
 444            TokenType.ANTI,
 445        }
 446
 447        PLACEHOLDER_PARSERS = {
 448            **parser.Parser.PLACEHOLDER_PARSERS,
 449            TokenType.PARAMETER: lambda self: (
 450                self.expression(exp.Placeholder, this=self._prev.text)
 451                if self._match(TokenType.NUMBER) or self._match_set(self.ID_VAR_TOKENS)
 452                else None
 453            ),
 454        }
 455
 456        TYPE_CONVERTERS = {
 457            # https://duckdb.org/docs/sql/data_types/numeric
 458            exp.DataType.Type.DECIMAL: build_default_decimal_type(precision=18, scale=3),
 459            # https://duckdb.org/docs/sql/data_types/text
 460            exp.DataType.Type.TEXT: lambda dtype: exp.DataType.build("TEXT"),
 461        }
 462
 463        STATEMENT_PARSERS = {
 464            **parser.Parser.STATEMENT_PARSERS,
 465            TokenType.ATTACH: lambda self: self._parse_attach_detach(),
 466            TokenType.DETACH: lambda self: self._parse_attach_detach(is_attach=False),
 467            TokenType.SHOW: lambda self: self._parse_show(),
 468        }
 469
 470        SET_PARSERS = {
 471            **parser.Parser.SET_PARSERS,
 472            "VARIABLE": lambda self: self._parse_set_item_assignment("VARIABLE"),
 473        }
 474
 475        def _parse_lambda(self, alias: bool = False) -> t.Optional[exp.Expression]:
 476            index = self._index
 477            if not self._match_text_seq("LAMBDA"):
 478                return super()._parse_lambda(alias=alias)
 479
 480            expressions = self._parse_csv(self._parse_lambda_arg)
 481            if not self._match(TokenType.COLON):
 482                self._retreat(index)
 483                return None
 484
 485            this = self._replace_lambda(self._parse_assignment(), expressions)
 486            return self.expression(exp.Lambda, this=this, expressions=expressions, colon=True)
 487
 488        def _parse_expression(self) -> t.Optional[exp.Expression]:
 489            # DuckDB supports prefix aliases, e.g. foo: 1
 490            if self._next and self._next.token_type == TokenType.COLON:
 491                alias = self._parse_id_var(tokens=self.ALIAS_TOKENS)
 492                self._match(TokenType.COLON)
 493                comments = self._prev_comments or []
 494
 495                this = self._parse_assignment()
 496                if isinstance(this, exp.Expression):
 497                    # Moves the comment next to the alias in `alias: expr /* comment */`
 498                    comments += this.pop_comments() or []
 499
 500                return self.expression(exp.Alias, comments=comments, this=this, alias=alias)
 501
 502            return super()._parse_expression()
 503
 504        def _parse_table(
 505            self,
 506            schema: bool = False,
 507            joins: bool = False,
 508            alias_tokens: t.Optional[t.Collection[TokenType]] = None,
 509            parse_bracket: bool = False,
 510            is_db_reference: bool = False,
 511            parse_partition: bool = False,
 512            consume_pipe: bool = False,
 513        ) -> t.Optional[exp.Expression]:
 514            # DuckDB supports prefix aliases, e.g. FROM foo: bar
 515            if self._next and self._next.token_type == TokenType.COLON:
 516                alias = self._parse_table_alias(
 517                    alias_tokens=alias_tokens or self.TABLE_ALIAS_TOKENS
 518                )
 519                self._match(TokenType.COLON)
 520                comments = self._prev_comments or []
 521            else:
 522                alias = None
 523                comments = []
 524
 525            table = super()._parse_table(
 526                schema=schema,
 527                joins=joins,
 528                alias_tokens=alias_tokens,
 529                parse_bracket=parse_bracket,
 530                is_db_reference=is_db_reference,
 531                parse_partition=parse_partition,
 532            )
 533            if isinstance(table, exp.Expression) and isinstance(alias, exp.TableAlias):
 534                # Moves the comment next to the alias in `alias: table /* comment */`
 535                comments += table.pop_comments() or []
 536                alias.comments = alias.pop_comments() + comments
 537                table.set("alias", alias)
 538
 539            return table
 540
 541        def _parse_table_sample(self, as_modifier: bool = False) -> t.Optional[exp.TableSample]:
 542            # https://duckdb.org/docs/sql/samples.html
 543            sample = super()._parse_table_sample(as_modifier=as_modifier)
 544            if sample and not sample.args.get("method"):
 545                if sample.args.get("size"):
 546                    sample.set("method", exp.var("RESERVOIR"))
 547                else:
 548                    sample.set("method", exp.var("SYSTEM"))
 549
 550            return sample
 551
 552        def _parse_bracket(
 553            self, this: t.Optional[exp.Expression] = None
 554        ) -> t.Optional[exp.Expression]:
 555            bracket = super()._parse_bracket(this)
 556
 557            if self.dialect.version < Version("1.2.0") and isinstance(bracket, exp.Bracket):
 558                # https://duckdb.org/2025/02/05/announcing-duckdb-120.html#breaking-changes
 559                bracket.set("returns_list_for_maps", True)
 560
 561            return bracket
 562
 563        def _parse_map(self) -> exp.ToMap | exp.Map:
 564            if self._match(TokenType.L_BRACE, advance=False):
 565                return self.expression(exp.ToMap, this=self._parse_bracket())
 566
 567            args = self._parse_wrapped_csv(self._parse_assignment)
 568            return self.expression(exp.Map, keys=seq_get(args, 0), values=seq_get(args, 1))
 569
 570        def _parse_struct_types(self, type_required: bool = False) -> t.Optional[exp.Expression]:
 571            return self._parse_field_def()
 572
 573        def _pivot_column_names(self, aggregations: t.List[exp.Expression]) -> t.List[str]:
 574            if len(aggregations) == 1:
 575                return super()._pivot_column_names(aggregations)
 576            return pivot_column_names(aggregations, dialect="duckdb")
 577
 578        def _parse_attach_detach(self, is_attach=True) -> exp.Attach | exp.Detach:
 579            def _parse_attach_option() -> exp.AttachOption:
 580                return self.expression(
 581                    exp.AttachOption,
 582                    this=self._parse_var(any_token=True),
 583                    expression=self._parse_field(any_token=True),
 584                )
 585
 586            self._match(TokenType.DATABASE)
 587            exists = self._parse_exists(not_=is_attach)
 588            this = self._parse_alias(self._parse_primary_or_var(), explicit=True)
 589
 590            if self._match(TokenType.L_PAREN, advance=False):
 591                expressions = self._parse_wrapped_csv(_parse_attach_option)
 592            else:
 593                expressions = None
 594
 595            return (
 596                self.expression(exp.Attach, this=this, exists=exists, expressions=expressions)
 597                if is_attach
 598                else self.expression(exp.Detach, this=this, exists=exists)
 599            )
 600
 601        def _parse_show_duckdb(self, this: str) -> exp.Show:
 602            return self.expression(exp.Show, this=this)
 603
 604        def _parse_primary(self) -> t.Optional[exp.Expression]:
 605            if self._match_pair(TokenType.HASH, TokenType.NUMBER):
 606                return exp.PositionalColumn(this=exp.Literal.number(self._prev.text))
 607
 608            return super()._parse_primary()
 609
 610    class Generator(generator.Generator):
 611        PARAMETER_TOKEN = "$"
 612        NAMED_PLACEHOLDER_TOKEN = "$"
 613        JOIN_HINTS = False
 614        TABLE_HINTS = False
 615        QUERY_HINTS = False
 616        LIMIT_FETCH = "LIMIT"
 617        STRUCT_DELIMITER = ("(", ")")
 618        RENAME_TABLE_WITH_DB = False
 619        NVL2_SUPPORTED = False
 620        SEMI_ANTI_JOIN_WITH_SIDE = False
 621        TABLESAMPLE_KEYWORDS = "USING SAMPLE"
 622        TABLESAMPLE_SEED_KEYWORD = "REPEATABLE"
 623        LAST_DAY_SUPPORTS_DATE_PART = False
 624        JSON_KEY_VALUE_PAIR_SEP = ","
 625        IGNORE_NULLS_IN_FUNC = True
 626        JSON_PATH_BRACKETED_KEY_SUPPORTED = False
 627        SUPPORTS_CREATE_TABLE_LIKE = False
 628        MULTI_ARG_DISTINCT = False
 629        CAN_IMPLEMENT_ARRAY_ANY = True
 630        SUPPORTS_TO_NUMBER = False
 631        SUPPORTS_WINDOW_EXCLUDE = True
 632        COPY_HAS_INTO_KEYWORD = False
 633        STAR_EXCEPT = "EXCLUDE"
 634        PAD_FILL_PATTERN_IS_REQUIRED = True
 635        ARRAY_CONCAT_IS_VAR_LEN = False
 636        ARRAY_SIZE_DIM_REQUIRED = False
 637        NORMALIZE_EXTRACT_DATE_PARTS = True
 638        SUPPORTS_LIKE_QUANTIFIERS = False
 639
 640        TRANSFORMS = {
 641            **generator.Generator.TRANSFORMS,
 642            exp.ApproxDistinct: approx_count_distinct_sql,
 643            exp.Array: inline_array_unless_query,
 644            exp.ArrayFilter: rename_func("LIST_FILTER"),
 645            exp.ArrayRemove: remove_from_array_using_filter,
 646            exp.ArraySort: _array_sort_sql,
 647            exp.ArraySum: rename_func("LIST_SUM"),
 648            exp.ArrayUniqueAgg: lambda self, e: self.func(
 649                "LIST", exp.Distinct(expressions=[e.this])
 650            ),
 651            exp.BitwiseXor: rename_func("XOR"),
 652            exp.CommentColumnConstraint: no_comment_column_constraint_sql,
 653            exp.CurrentDate: lambda *_: "CURRENT_DATE",
 654            exp.CurrentTime: lambda *_: "CURRENT_TIME",
 655            exp.CurrentTimestamp: lambda *_: "CURRENT_TIMESTAMP",
 656            exp.DayOfMonth: rename_func("DAYOFMONTH"),
 657            exp.DayOfWeek: rename_func("DAYOFWEEK"),
 658            exp.DayOfWeekIso: rename_func("ISODOW"),
 659            exp.DayOfYear: rename_func("DAYOFYEAR"),
 660            exp.DataType: _datatype_sql,
 661            exp.Date: _date_sql,
 662            exp.DateAdd: date_delta_to_binary_interval_op(),
 663            exp.DateFromParts: rename_func("MAKE_DATE"),
 664            exp.DateSub: date_delta_to_binary_interval_op(),
 665            exp.DateDiff: _date_diff_sql,
 666            exp.DateStrToDate: datestrtodate_sql,
 667            exp.Datetime: no_datetime_sql,
 668            exp.DatetimeSub: date_delta_to_binary_interval_op(),
 669            exp.DatetimeAdd: date_delta_to_binary_interval_op(),
 670            exp.DateToDi: lambda self,
 671            e: f"CAST(STRFTIME({self.sql(e, 'this')}, {DuckDB.DATEINT_FORMAT}) AS INT)",
 672            exp.Decode: lambda self, e: encode_decode_sql(self, e, "DECODE", replace=False),
 673            exp.DiToDate: lambda self,
 674            e: f"CAST(STRPTIME(CAST({self.sql(e, 'this')} AS TEXT), {DuckDB.DATEINT_FORMAT}) AS DATE)",
 675            exp.Encode: lambda self, e: encode_decode_sql(self, e, "ENCODE", replace=False),
 676            exp.GenerateDateArray: _generate_datetime_array_sql,
 677            exp.GenerateTimestampArray: _generate_datetime_array_sql,
 678            exp.GroupConcat: lambda self, e: groupconcat_sql(self, e, within_group=False),
 679            exp.HexString: lambda self, e: self.hexstring_sql(e, binary_function_repr="FROM_HEX"),
 680            exp.Explode: rename_func("UNNEST"),
 681            exp.IntDiv: lambda self, e: self.binary(e, "//"),
 682            exp.IsInf: rename_func("ISINF"),
 683            exp.IsNan: rename_func("ISNAN"),
 684            exp.JSONBExists: rename_func("JSON_EXISTS"),
 685            exp.JSONExtract: _arrow_json_extract_sql,
 686            exp.JSONExtractArray: _json_extract_value_array_sql,
 687            exp.JSONExtractScalar: _arrow_json_extract_sql,
 688            exp.JSONFormat: _json_format_sql,
 689            exp.JSONValueArray: _json_extract_value_array_sql,
 690            exp.Lateral: explode_to_unnest_sql,
 691            exp.LogicalOr: rename_func("BOOL_OR"),
 692            exp.LogicalAnd: rename_func("BOOL_AND"),
 693            exp.MakeInterval: lambda self, e: no_make_interval_sql(self, e, sep=" "),
 694            exp.MD5Digest: lambda self, e: self.func("UNHEX", self.func("MD5", e.this)),
 695            exp.MonthsBetween: lambda self, e: self.func(
 696                "DATEDIFF",
 697                "'month'",
 698                exp.cast(e.expression, exp.DataType.Type.TIMESTAMP, copy=True),
 699                exp.cast(e.this, exp.DataType.Type.TIMESTAMP, copy=True),
 700            ),
 701            exp.PercentileCont: rename_func("QUANTILE_CONT"),
 702            exp.PercentileDisc: rename_func("QUANTILE_DISC"),
 703            # DuckDB doesn't allow qualified columns inside of PIVOT expressions.
 704            # See: https://github.com/duckdb/duckdb/blob/671faf92411182f81dce42ac43de8bfb05d9909e/src/planner/binder/tableref/bind_pivot.cpp#L61-L62
 705            exp.Pivot: transforms.preprocess([transforms.unqualify_columns]),
 706            exp.RegexpReplace: lambda self, e: self.func(
 707                "REGEXP_REPLACE",
 708                e.this,
 709                e.expression,
 710                e.args.get("replacement"),
 711                e.args.get("modifiers"),
 712            ),
 713            exp.RegexpLike: rename_func("REGEXP_MATCHES"),
 714            exp.RegexpILike: lambda self, e: self.func(
 715                "REGEXP_MATCHES", e.this, e.expression, exp.Literal.string("i")
 716            ),
 717            exp.RegexpSplit: rename_func("STR_SPLIT_REGEX"),
 718            exp.Return: lambda self, e: self.sql(e, "this"),
 719            exp.ReturnsProperty: lambda self, e: "TABLE" if isinstance(e.this, exp.Schema) else "",
 720            exp.Rand: rename_func("RANDOM"),
 721            exp.SHA: rename_func("SHA1"),
 722            exp.SHA2: sha256_sql,
 723            exp.Split: rename_func("STR_SPLIT"),
 724            exp.SortArray: _sort_array_sql,
 725            exp.StrPosition: strposition_sql,
 726            exp.StrToUnix: lambda self, e: self.func(
 727                "EPOCH", self.func("STRPTIME", e.this, self.format_time(e))
 728            ),
 729            exp.Struct: _struct_sql,
 730            exp.Transform: rename_func("LIST_TRANSFORM"),
 731            exp.TimeAdd: date_delta_to_binary_interval_op(),
 732            exp.Time: no_time_sql,
 733            exp.TimeDiff: _timediff_sql,
 734            exp.Timestamp: no_timestamp_sql,
 735            exp.TimestampDiff: lambda self, e: self.func(
 736                "DATE_DIFF", exp.Literal.string(e.unit), e.expression, e.this
 737            ),
 738            exp.TimestampTrunc: timestamptrunc_sql(),
 739            exp.TimeStrToDate: lambda self, e: self.sql(exp.cast(e.this, exp.DataType.Type.DATE)),
 740            exp.TimeStrToTime: timestrtotime_sql,
 741            exp.TimeStrToUnix: lambda self, e: self.func(
 742                "EPOCH", exp.cast(e.this, exp.DataType.Type.TIMESTAMP)
 743            ),
 744            exp.TimeToStr: lambda self, e: self.func("STRFTIME", e.this, self.format_time(e)),
 745            exp.TimeToUnix: rename_func("EPOCH"),
 746            exp.TsOrDiToDi: lambda self,
 747            e: f"CAST(SUBSTR(REPLACE(CAST({self.sql(e, 'this')} AS TEXT), '-', ''), 1, 8) AS INT)",
 748            exp.TsOrDsAdd: date_delta_to_binary_interval_op(),
 749            exp.TsOrDsDiff: lambda self, e: self.func(
 750                "DATE_DIFF",
 751                f"'{e.args.get('unit') or 'DAY'}'",
 752                exp.cast(e.expression, exp.DataType.Type.TIMESTAMP),
 753                exp.cast(e.this, exp.DataType.Type.TIMESTAMP),
 754            ),
 755            exp.UnixToStr: lambda self, e: self.func(
 756                "STRFTIME", self.func("TO_TIMESTAMP", e.this), self.format_time(e)
 757            ),
 758            exp.DatetimeTrunc: lambda self, e: self.func(
 759                "DATE_TRUNC", unit_to_str(e), exp.cast(e.this, exp.DataType.Type.DATETIME)
 760            ),
 761            exp.UnixToTime: _unix_to_time_sql,
 762            exp.UnixToTimeStr: lambda self, e: f"CAST(TO_TIMESTAMP({self.sql(e, 'this')}) AS TEXT)",
 763            exp.VariancePop: rename_func("VAR_POP"),
 764            exp.WeekOfYear: rename_func("WEEKOFYEAR"),
 765            exp.Xor: bool_xor_sql,
 766            exp.Levenshtein: unsupported_args("ins_cost", "del_cost", "sub_cost", "max_dist")(
 767                rename_func("LEVENSHTEIN")
 768            ),
 769            exp.JSONObjectAgg: rename_func("JSON_GROUP_OBJECT"),
 770            exp.JSONBObjectAgg: rename_func("JSON_GROUP_OBJECT"),
 771            exp.DateBin: rename_func("TIME_BUCKET"),
 772        }
 773
 774        SUPPORTED_JSON_PATH_PARTS = {
 775            exp.JSONPathKey,
 776            exp.JSONPathRoot,
 777            exp.JSONPathSubscript,
 778            exp.JSONPathWildcard,
 779        }
 780
 781        TYPE_MAPPING = {
 782            **generator.Generator.TYPE_MAPPING,
 783            exp.DataType.Type.BINARY: "BLOB",
 784            exp.DataType.Type.BPCHAR: "TEXT",
 785            exp.DataType.Type.CHAR: "TEXT",
 786            exp.DataType.Type.DATETIME: "TIMESTAMP",
 787            exp.DataType.Type.FLOAT: "REAL",
 788            exp.DataType.Type.JSONB: "JSON",
 789            exp.DataType.Type.NCHAR: "TEXT",
 790            exp.DataType.Type.NVARCHAR: "TEXT",
 791            exp.DataType.Type.UINT: "UINTEGER",
 792            exp.DataType.Type.VARBINARY: "BLOB",
 793            exp.DataType.Type.ROWVERSION: "BLOB",
 794            exp.DataType.Type.VARCHAR: "TEXT",
 795            exp.DataType.Type.TIMESTAMPNTZ: "TIMESTAMP",
 796            exp.DataType.Type.TIMESTAMP_S: "TIMESTAMP_S",
 797            exp.DataType.Type.TIMESTAMP_MS: "TIMESTAMP_MS",
 798            exp.DataType.Type.TIMESTAMP_NS: "TIMESTAMP_NS",
 799        }
 800
 801        # https://github.com/duckdb/duckdb/blob/ff7f24fd8e3128d94371827523dae85ebaf58713/third_party/libpg_query/grammar/keywords/reserved_keywords.list#L1-L77
 802        RESERVED_KEYWORDS = {
 803            "array",
 804            "analyse",
 805            "union",
 806            "all",
 807            "when",
 808            "in_p",
 809            "default",
 810            "create_p",
 811            "window",
 812            "asymmetric",
 813            "to",
 814            "else",
 815            "localtime",
 816            "from",
 817            "end_p",
 818            "select",
 819            "current_date",
 820            "foreign",
 821            "with",
 822            "grant",
 823            "session_user",
 824            "or",
 825            "except",
 826            "references",
 827            "fetch",
 828            "limit",
 829            "group_p",
 830            "leading",
 831            "into",
 832            "collate",
 833            "offset",
 834            "do",
 835            "then",
 836            "localtimestamp",
 837            "check_p",
 838            "lateral_p",
 839            "current_role",
 840            "where",
 841            "asc_p",
 842            "placing",
 843            "desc_p",
 844            "user",
 845            "unique",
 846            "initially",
 847            "column",
 848            "both",
 849            "some",
 850            "as",
 851            "any",
 852            "only",
 853            "deferrable",
 854            "null_p",
 855            "current_time",
 856            "true_p",
 857            "table",
 858            "case",
 859            "trailing",
 860            "variadic",
 861            "for",
 862            "on",
 863            "distinct",
 864            "false_p",
 865            "not",
 866            "constraint",
 867            "current_timestamp",
 868            "returning",
 869            "primary",
 870            "intersect",
 871            "having",
 872            "analyze",
 873            "current_user",
 874            "and",
 875            "cast",
 876            "symmetric",
 877            "using",
 878            "order",
 879            "current_catalog",
 880        }
 881
 882        UNWRAPPED_INTERVAL_VALUES = (exp.Literal, exp.Paren)
 883
 884        # DuckDB doesn't generally support CREATE TABLE .. properties
 885        # https://duckdb.org/docs/sql/statements/create_table.html
 886        PROPERTIES_LOCATION = {
 887            prop: exp.Properties.Location.UNSUPPORTED
 888            for prop in generator.Generator.PROPERTIES_LOCATION
 889        }
 890
 891        # There are a few exceptions (e.g. temporary tables) which are supported or
 892        # can be transpiled to DuckDB, so we explicitly override them accordingly
 893        PROPERTIES_LOCATION[exp.LikeProperty] = exp.Properties.Location.POST_SCHEMA
 894        PROPERTIES_LOCATION[exp.TemporaryProperty] = exp.Properties.Location.POST_CREATE
 895        PROPERTIES_LOCATION[exp.ReturnsProperty] = exp.Properties.Location.POST_ALIAS
 896        PROPERTIES_LOCATION[exp.SequenceProperties] = exp.Properties.Location.POST_EXPRESSION
 897
 898        IGNORE_RESPECT_NULLS_WINDOW_FUNCTIONS = (
 899            exp.FirstValue,
 900            exp.Lag,
 901            exp.LastValue,
 902            exp.Lead,
 903            exp.NthValue,
 904        )
 905
 906        def lambda_sql(
 907            self, expression: exp.Lambda, arrow_sep: str = "->", wrap: bool = True
 908        ) -> str:
 909            if expression.args.get("colon"):
 910                prefix = "LAMBDA "
 911                arrow_sep = ":"
 912                wrap = False
 913            else:
 914                prefix = ""
 915
 916            lambda_sql = super().lambda_sql(expression, arrow_sep=arrow_sep, wrap=wrap)
 917            return f"{prefix}{lambda_sql}"
 918
 919        def show_sql(self, expression: exp.Show) -> str:
 920            return f"SHOW {expression.name}"
 921
 922        def fromiso8601timestamp_sql(self, expression: exp.FromISO8601Timestamp) -> str:
 923            return self.sql(exp.cast(expression.this, exp.DataType.Type.TIMESTAMPTZ))
 924
 925        def strtotime_sql(self, expression: exp.StrToTime) -> str:
 926            if expression.args.get("safe"):
 927                formatted_time = self.format_time(expression)
 928                return f"CAST({self.func('TRY_STRPTIME', expression.this, formatted_time)} AS TIMESTAMP)"
 929            return str_to_time_sql(self, expression)
 930
 931        def strtodate_sql(self, expression: exp.StrToDate) -> str:
 932            if expression.args.get("safe"):
 933                formatted_time = self.format_time(expression)
 934                return f"CAST({self.func('TRY_STRPTIME', expression.this, formatted_time)} AS DATE)"
 935            return f"CAST({str_to_time_sql(self, expression)} AS DATE)"
 936
 937        def parsejson_sql(self, expression: exp.ParseJSON) -> str:
 938            arg = expression.this
 939            if expression.args.get("safe"):
 940                return self.sql(exp.case().when(exp.func("json_valid", arg), arg).else_(exp.null()))
 941            return self.func("JSON", arg)
 942
 943        def timefromparts_sql(self, expression: exp.TimeFromParts) -> str:
 944            nano = expression.args.get("nano")
 945            if nano is not None:
 946                expression.set(
 947                    "sec", expression.args["sec"] + nano.pop() / exp.Literal.number(1000000000.0)
 948                )
 949
 950            return rename_func("MAKE_TIME")(self, expression)
 951
 952        def timestampfromparts_sql(self, expression: exp.TimestampFromParts) -> str:
 953            sec = expression.args["sec"]
 954
 955            milli = expression.args.get("milli")
 956            if milli is not None:
 957                sec += milli.pop() / exp.Literal.number(1000.0)
 958
 959            nano = expression.args.get("nano")
 960            if nano is not None:
 961                sec += nano.pop() / exp.Literal.number(1000000000.0)
 962
 963            if milli or nano:
 964                expression.set("sec", sec)
 965
 966            return rename_func("MAKE_TIMESTAMP")(self, expression)
 967
 968        def tablesample_sql(
 969            self,
 970            expression: exp.TableSample,
 971            tablesample_keyword: t.Optional[str] = None,
 972        ) -> str:
 973            if not isinstance(expression.parent, exp.Select):
 974                # This sample clause only applies to a single source, not the entire resulting relation
 975                tablesample_keyword = "TABLESAMPLE"
 976
 977            if expression.args.get("size"):
 978                method = expression.args.get("method")
 979                if method and method.name.upper() != "RESERVOIR":
 980                    self.unsupported(
 981                        f"Sampling method {method} is not supported with a discrete sample count, "
 982                        "defaulting to reservoir sampling"
 983                    )
 984                    expression.set("method", exp.var("RESERVOIR"))
 985
 986            return super().tablesample_sql(expression, tablesample_keyword=tablesample_keyword)
 987
 988        def columndef_sql(self, expression: exp.ColumnDef, sep: str = " ") -> str:
 989            if isinstance(expression.parent, exp.UserDefinedFunction):
 990                return self.sql(expression, "this")
 991            return super().columndef_sql(expression, sep)
 992
 993        def join_sql(self, expression: exp.Join) -> str:
 994            if (
 995                expression.side == "LEFT"
 996                and not expression.args.get("on")
 997                and isinstance(expression.this, exp.Unnest)
 998            ):
 999                # Some dialects support `LEFT JOIN UNNEST(...)` without an explicit ON clause
1000                # DuckDB doesn't, but we can just add a dummy ON clause that is always true
1001                return super().join_sql(expression.on(exp.true()))
1002
1003            return super().join_sql(expression)
1004
1005        def generateseries_sql(self, expression: exp.GenerateSeries) -> str:
1006            # GENERATE_SERIES(a, b) -> [a, b], RANGE(a, b) -> [a, b)
1007            if expression.args.get("is_end_exclusive"):
1008                return rename_func("RANGE")(self, expression)
1009
1010            return self.function_fallback_sql(expression)
1011
1012        def countif_sql(self, expression: exp.CountIf) -> str:
1013            if self.dialect.version >= Version("1.2"):
1014                return self.function_fallback_sql(expression)
1015
1016            # https://github.com/tobymao/sqlglot/pull/4749
1017            return count_if_to_sum(self, expression)
1018
1019        def bracket_sql(self, expression: exp.Bracket) -> str:
1020            if self.dialect.version >= Version("1.2"):
1021                return super().bracket_sql(expression)
1022
1023            # https://duckdb.org/2025/02/05/announcing-duckdb-120.html#breaking-changes
1024            this = expression.this
1025            if isinstance(this, exp.Array):
1026                this.replace(exp.paren(this))
1027
1028            bracket = super().bracket_sql(expression)
1029
1030            if not expression.args.get("returns_list_for_maps"):
1031                if not this.type:
1032                    from sqlglot.optimizer.annotate_types import annotate_types
1033
1034                    this = annotate_types(this, dialect=self.dialect)
1035
1036                if this.is_type(exp.DataType.Type.MAP):
1037                    bracket = f"({bracket})[1]"
1038
1039            return bracket
1040
1041        def withingroup_sql(self, expression: exp.WithinGroup) -> str:
1042            expression_sql = self.sql(expression, "expression")
1043
1044            func = expression.this
1045            if isinstance(func, exp.PERCENTILES):
1046                # Make the order key the first arg and slide the fraction to the right
1047                # https://duckdb.org/docs/sql/aggregates#ordered-set-aggregate-functions
1048                order_col = expression.find(exp.Ordered)
1049                if order_col:
1050                    func.set("expression", func.this)
1051                    func.set("this", order_col.this)
1052
1053            this = self.sql(expression, "this").rstrip(")")
1054
1055            return f"{this}{expression_sql})"
1056
1057        def length_sql(self, expression: exp.Length) -> str:
1058            arg = expression.this
1059
1060            # Dialects like BQ and Snowflake also accept binary values as args, so
1061            # DDB will attempt to infer the type or resort to case/when resolution
1062            if not expression.args.get("binary") or arg.is_string:
1063                return self.func("LENGTH", arg)
1064
1065            if not arg.type:
1066                from sqlglot.optimizer.annotate_types import annotate_types
1067
1068                arg = annotate_types(arg, dialect=self.dialect)
1069
1070            if arg.is_type(*exp.DataType.TEXT_TYPES):
1071                return self.func("LENGTH", arg)
1072
1073            # We need these casts to make duckdb's static type checker happy
1074            blob = exp.cast(arg, exp.DataType.Type.VARBINARY)
1075            varchar = exp.cast(arg, exp.DataType.Type.VARCHAR)
1076
1077            case = (
1078                exp.case(self.func("TYPEOF", arg))
1079                .when("'BLOB'", self.func("OCTET_LENGTH", blob))
1080                .else_(
1081                    exp.Anonymous(this="LENGTH", expressions=[varchar])
1082                )  # anonymous to break length_sql recursion
1083            )
1084
1085            return self.sql(case)
1086
1087        def objectinsert_sql(self, expression: exp.ObjectInsert) -> str:
1088            this = expression.this
1089            key = expression.args.get("key")
1090            key_sql = key.name if isinstance(key, exp.Expression) else ""
1091            value_sql = self.sql(expression, "value")
1092
1093            kv_sql = f"{key_sql} := {value_sql}"
1094
1095            # If the input struct is empty e.g. transpiling OBJECT_INSERT(OBJECT_CONSTRUCT(), key, value) from Snowflake
1096            # then we can generate STRUCT_PACK which will build it since STRUCT_INSERT({}, key := value) is not valid DuckDB
1097            if isinstance(this, exp.Struct) and not this.expressions:
1098                return self.func("STRUCT_PACK", kv_sql)
1099
1100            return self.func("STRUCT_INSERT", this, kv_sql)
1101
1102        def unnest_sql(self, expression: exp.Unnest) -> str:
1103            explode_array = expression.args.get("explode_array")
1104            if explode_array:
1105                # In BigQuery, UNNESTing a nested array leads to explosion of the top-level array & struct
1106                # This is transpiled to DDB by transforming "FROM UNNEST(...)" to "FROM (SELECT UNNEST(..., max_depth => 2))"
1107                expression.expressions.append(
1108                    exp.Kwarg(this=exp.var("max_depth"), expression=exp.Literal.number(2))
1109                )
1110
1111                # If BQ's UNNEST is aliased, we transform it from a column alias to a table alias in DDB
1112                alias = expression.args.get("alias")
1113                if isinstance(alias, exp.TableAlias):
1114                    expression.set("alias", None)
1115                    if alias.columns:
1116                        alias = exp.TableAlias(this=seq_get(alias.columns, 0))
1117
1118                unnest_sql = super().unnest_sql(expression)
1119                select = exp.Select(expressions=[unnest_sql]).subquery(alias)
1120                return self.sql(select)
1121
1122            return super().unnest_sql(expression)
1123
1124        def ignorenulls_sql(self, expression: exp.IgnoreNulls) -> str:
1125            this = expression.this
1126
1127            if isinstance(this, self.IGNORE_RESPECT_NULLS_WINDOW_FUNCTIONS):
1128                # DuckDB should render IGNORE NULLS only for the general-purpose
1129                # window functions that accept it e.g. FIRST_VALUE(... IGNORE NULLS) OVER (...)
1130                return super().ignorenulls_sql(expression)
1131
1132            if isinstance(this, exp.First):
1133                this = exp.AnyValue(this=this.this)
1134
1135            if not isinstance(this, exp.AnyValue):
1136                self.unsupported("IGNORE NULLS is not supported for non-window functions.")
1137
1138            return self.sql(this)
1139
1140        def respectnulls_sql(self, expression: exp.RespectNulls) -> str:
1141            if isinstance(expression.this, self.IGNORE_RESPECT_NULLS_WINDOW_FUNCTIONS):
1142                # DuckDB should render RESPECT NULLS only for the general-purpose
1143                # window functions that accept it e.g. FIRST_VALUE(... RESPECT NULLS) OVER (...)
1144                return super().respectnulls_sql(expression)
1145
1146            self.unsupported("RESPECT NULLS is not supported for non-window functions.")
1147            return self.sql(expression, "this")
1148
1149        def arraytostring_sql(self, expression: exp.ArrayToString) -> str:
1150            this = self.sql(expression, "this")
1151            null_text = self.sql(expression, "null")
1152
1153            if null_text:
1154                this = f"LIST_TRANSFORM({this}, x -> COALESCE(x, {null_text}))"
1155
1156            return self.func("ARRAY_TO_STRING", this, expression.expression)
1157
1158        @unsupported_args("position", "occurrence")
1159        def regexpextract_sql(self, expression: exp.RegexpExtract) -> str:
1160            group = expression.args.get("group")
1161            params = expression.args.get("parameters")
1162
1163            # Do not render group if there is no following argument,
1164            # and it's the default value for this dialect
1165            if (
1166                not params
1167                and group
1168                and group.name == str(self.dialect.REGEXP_EXTRACT_DEFAULT_GROUP)
1169            ):
1170                group = None
1171            return self.func(
1172                "REGEXP_EXTRACT", expression.this, expression.expression, group, params
1173            )
1174
1175        @unsupported_args("culture")
1176        def numbertostr_sql(self, expression: exp.NumberToStr) -> str:
1177            fmt = expression.args.get("format")
1178            if fmt and fmt.is_int:
1179                return self.func("FORMAT", f"'{{:,.{fmt.name}f}}'", expression.this)
1180
1181            self.unsupported("Only integer formats are supported by NumberToStr")
1182            return self.function_fallback_sql(expression)
1183
1184        def autoincrementcolumnconstraint_sql(self, _) -> str:
1185            self.unsupported("The AUTOINCREMENT column constraint is not supported by DuckDB")
1186            return ""
1187
1188        def aliases_sql(self, expression: exp.Aliases) -> str:
1189            this = expression.this
1190            if isinstance(this, exp.Posexplode):
1191                return self.posexplode_sql(this)
1192
1193            return super().aliases_sql(expression)
1194
1195        def posexplode_sql(self, expression: exp.Posexplode) -> str:
1196            this = expression.this
1197            parent = expression.parent
1198
1199            # The default Spark aliases are "pos" and "col", unless specified otherwise
1200            pos, col = exp.to_identifier("pos"), exp.to_identifier("col")
1201
1202            if isinstance(parent, exp.Aliases):
1203                # Column case: SELECT POSEXPLODE(col) [AS (a, b)]
1204                pos, col = parent.expressions
1205            elif isinstance(parent, exp.Table):
1206                # Table case: SELECT * FROM POSEXPLODE(col) [AS (a, b)]
1207                alias = parent.args.get("alias")
1208                if alias:
1209                    pos, col = alias.columns or [pos, col]
1210                    alias.pop()
1211
1212            # Translate POSEXPLODE to UNNEST + GENERATE_SUBSCRIPTS
1213            # Note: In Spark pos is 0-indexed, but in DuckDB it's 1-indexed, so we subtract 1 from GENERATE_SUBSCRIPTS
1214            unnest_sql = self.sql(exp.Unnest(expressions=[this], alias=col))
1215            gen_subscripts = self.sql(
1216                exp.Alias(
1217                    this=exp.Anonymous(
1218                        this="GENERATE_SUBSCRIPTS", expressions=[this, exp.Literal.number(1)]
1219                    )
1220                    - exp.Literal.number(1),
1221                    alias=pos,
1222                )
1223            )
1224
1225            posexplode_sql = self.format_args(gen_subscripts, unnest_sql)
1226
1227            if isinstance(parent, exp.From) or (parent and isinstance(parent.parent, exp.From)):
1228                # SELECT * FROM POSEXPLODE(col) -> SELECT * FROM (SELECT GENERATE_SUBSCRIPTS(...), UNNEST(...))
1229                return self.sql(exp.Subquery(this=exp.Select(expressions=[posexplode_sql])))
1230
1231            return posexplode_sql
1232
1233        def addmonths_sql(self, expression: exp.AddMonths) -> str:
1234            this = expression.this
1235
1236            if not this.type:
1237                from sqlglot.optimizer.annotate_types import annotate_types
1238
1239                this = annotate_types(this, dialect=self.dialect)
1240
1241            if this.is_type(*exp.DataType.TEXT_TYPES):
1242                this = exp.Cast(this=this, to=exp.DataType(this=exp.DataType.Type.TIMESTAMP))
1243
1244            func = self.func(
1245                "DATE_ADD", this, exp.Interval(this=expression.expression, unit=exp.var("MONTH"))
1246            )
1247
1248            # DuckDB's DATE_ADD function returns TIMESTAMP/DATETIME by default, even when the input is DATE
1249            # To match for example Snowflake's ADD_MONTHS behavior (which preserves the input type)
1250            # We need to cast the result back to the original type when the input is DATE or TIMESTAMPTZ
1251            # Example: ADD_MONTHS('2023-01-31'::date, 1) should return DATE, not TIMESTAMP
1252            if this.is_type(exp.DataType.Type.DATE, exp.DataType.Type.TIMESTAMPTZ):
1253                return self.sql(exp.Cast(this=func, to=this.type))
1254
1255            return self.sql(func)