sqlglot.parsers.snowflake

   1from __future__ import annotations
   2
   3import typing as t
   4
   5from sqlglot import exp, parser
   6from sqlglot.dialects.dialect import (
   7    Dialect,
   8    binary_from_function,
   9    build_default_decimal_type,
  10    build_formatted_time,
  11    build_like,
  12    build_replace_with_optional_replacement,
  13    build_timetostr_or_tochar,
  14    build_trunc,
  15    date_trunc_to_time,
  16    map_date_part,
  17)
  18from sqlglot.helper import is_date_unit, is_int, seq_get
  19from sqlglot.tokens import TokenType
  20from sqlglot.trie import new_trie
  21
  22if t.TYPE_CHECKING:
  23    from collections.abc import Collection
  24
  25    from sqlglot._typing import B, E
  26    from sqlglot.dialects.dialect import Dialect
  27
  28
  29def _build_approx_top_k(args: list) -> exp.ApproxTopK:
  30    """
  31    Normalizes APPROX_TOP_K arguments to match Snowflake semantics.
  32
  33    Snowflake APPROX_TOP_K signature: APPROX_TOP_K(column [, k] [, counters])
  34    - k defaults to 1 if omitted (per Snowflake documentation)
  35    - counters is optional precision parameter
  36    """
  37    # Add default k=1 if only column is provided
  38    if len(args) == 1:
  39        args.append(exp.Literal.number(1))
  40
  41    return exp.ApproxTopK.from_arg_list(args)
  42
  43
  44def _build_to_number(args: list, safe: bool = False) -> exp.ToNumber:
  45    second_arg = seq_get(args, 1)
  46    if second_arg and second_arg.is_number:
  47        fmt = None
  48        precision = second_arg
  49        scale = seq_get(args, 2) or exp.Literal.number(0)
  50    else:
  51        fmt = second_arg
  52        precision = seq_get(args, 2) or exp.Literal.number(38)
  53        scale = seq_get(args, 3) or exp.Literal.number(0)
  54
  55    return exp.ToNumber(
  56        this=seq_get(args, 0),
  57        format=fmt,
  58        precision=precision,
  59        scale=scale,
  60        safe=safe,
  61    )
  62
  63
  64def _build_date_from_parts(args: list) -> exp.DateFromParts:
  65    return exp.DateFromParts(
  66        year=seq_get(args, 0),
  67        month=seq_get(args, 1),
  68        day=seq_get(args, 2),
  69        allow_overflow=True,
  70    )
  71
  72
  73# Timestamp types used in _build_datetime
  74TIMESTAMP_TYPES = {
  75    exp.DType.TIMESTAMP: "TO_TIMESTAMP",
  76    exp.DType.TIMESTAMPLTZ: "TO_TIMESTAMP_LTZ",
  77    exp.DType.TIMESTAMPNTZ: "TO_TIMESTAMP_NTZ",
  78    exp.DType.TIMESTAMPTZ: "TO_TIMESTAMP_TZ",
  79}
  80
  81
  82def _build_datetime(name: str, kind: exp.DType, safe: bool = False) -> t.Callable:
  83    def _builder(args: list, dialect: Dialect) -> exp.Func:
  84        value = seq_get(args, 0)
  85        scale_or_fmt = seq_get(args, 1)
  86
  87        int_value = value is not None and is_int(value.name)
  88        int_scale_or_fmt = scale_or_fmt is not None and scale_or_fmt.is_int
  89
  90        if isinstance(value, (exp.Literal, exp.Neg)) or (value and scale_or_fmt):
  91            # Converts calls like `TO_TIME('01:02:03')` into casts
  92            if len(args) == 1 and value.is_string and not int_value:
  93                cast = (
  94                    exp.TryCast(this=value, to=kind.into_expr(), requires_string=True)
  95                    if safe
  96                    else exp.cast(value, kind)
  97                )
  98                if safe and kind == exp.DType.DATE:
  99                    cast.set("probe_date_format", True)
 100                return cast
 101
 102            # Handles `TO_TIMESTAMP(str, fmt)` and `TO_TIMESTAMP(num, scale)` as special
 103            # cases so we can transpile them, since they're relatively common
 104            if kind in TIMESTAMP_TYPES:
 105                if not safe and (int_scale_or_fmt or (int_value and scale_or_fmt is None)):
 106                    # TRY_TO_TIMESTAMP('integer') is not parsed into exp.UnixToTime as
 107                    # it's not easily transpilable. Also, numeric-looking strings with
 108                    # format strings (e.g., TO_TIMESTAMP('20240115', 'YYYYMMDD')) should
 109                    # use StrToTime, not UnixToTime.
 110                    unix_expr = exp.UnixToTime(this=value, scale=scale_or_fmt)
 111                    unix_expr.set("target_type", kind.into_expr())
 112                    return unix_expr
 113                if scale_or_fmt and not int_scale_or_fmt:
 114                    # Format string provided (e.g., 'YYYY-MM-DD'), use StrToTime
 115                    strtotime_expr = build_formatted_time(exp.StrToTime)(args, dialect)
 116                    strtotime_expr.set("safe", safe)
 117                    strtotime_expr.set("target_type", kind.into_expr())
 118                    return strtotime_expr
 119
 120        # Handle DATE/TIME with format strings - allow int_value if a format string is provided
 121        has_format_string = scale_or_fmt and not int_scale_or_fmt
 122        if kind in (exp.DType.DATE, exp.DType.TIME) and (not int_value or has_format_string):
 123            klass = exp.TsOrDsToDate if kind == exp.DType.DATE else exp.TsOrDsToTime
 124            formatted_exp = build_formatted_time(klass)(args, dialect)
 125            formatted_exp.set("safe", safe)
 126            return formatted_exp
 127
 128        return exp.Anonymous(this=name, expressions=args)
 129
 130    return _builder
 131
 132
 133def _build_bitwise(expr_type: type[B], name: str) -> t.Callable[[list], B | exp.Anonymous]:
 134    def _builder(args: list) -> B | exp.Anonymous:
 135        if len(args) == 3:
 136            # Special handling for bitwise operations with padside argument
 137            if expr_type in (exp.BitwiseAnd, exp.BitwiseOr, exp.BitwiseXor):
 138                return expr_type(
 139                    this=seq_get(args, 0), expression=seq_get(args, 1), padside=seq_get(args, 2)
 140                )
 141            return exp.Anonymous(this=name, expressions=args)
 142
 143        result = binary_from_function(expr_type)(args)
 144
 145        # Snowflake specifies INT128 for bitwise shifts
 146        if expr_type in (exp.BitwiseLeftShift, exp.BitwiseRightShift):
 147            result.set("requires_int128", True)
 148
 149        return result
 150
 151    return _builder
 152
 153
 154# https://docs.snowflake.com/en/sql-reference/functions/div0
 155def _build_if_from_div0(args: list) -> exp.If:
 156    lhs = exp._wrap(seq_get(args, 0), exp.Binary)
 157    rhs = exp._wrap(seq_get(args, 1), exp.Binary)
 158
 159    cond = exp.EQ(this=rhs, expression=exp.Literal.number(0)).and_(
 160        exp.Is(this=lhs, expression=exp.null()).not_()
 161    )
 162    true = exp.Literal.number(0)
 163    false = exp.Div(this=lhs, expression=rhs)
 164    return exp.If(this=cond, true=true, false=false)
 165
 166
 167# https://docs.snowflake.com/en/sql-reference/functions/div0null
 168def _build_if_from_div0null(args: list) -> exp.If:
 169    lhs = exp._wrap(seq_get(args, 0), exp.Binary)
 170    rhs = exp._wrap(seq_get(args, 1), exp.Binary)
 171
 172    # Returns 0 when divisor is 0 OR NULL
 173    cond = exp.EQ(this=rhs, expression=exp.Literal.number(0)).or_(
 174        exp.Is(this=rhs, expression=exp.null())
 175    )
 176    true = exp.Literal.number(0)
 177    false = exp.Div(this=lhs, expression=rhs)
 178    return exp.If(this=cond, true=true, false=false)
 179
 180
 181# https://docs.snowflake.com/en/sql-reference/functions/zeroifnull
 182def _build_if_from_zeroifnull(args: list) -> exp.If:
 183    cond = exp.Is(this=seq_get(args, 0), expression=exp.Null())
 184    return exp.If(this=cond, true=exp.Literal.number(0), false=seq_get(args, 0))
 185
 186
 187def _build_search(args: list) -> exp.Search:
 188    kwargs = {
 189        "this": seq_get(args, 0),
 190        "expression": seq_get(args, 1),
 191        **{arg.name.lower(): arg for arg in args[2:] if isinstance(arg, exp.Kwarg)},
 192    }
 193    return exp.Search(**kwargs)
 194
 195
 196# https://docs.snowflake.com/en/sql-reference/functions/zeroifnull
 197def _build_if_from_nullifzero(args: list) -> exp.If:
 198    cond = exp.EQ(this=seq_get(args, 0), expression=exp.Literal.number(0))
 199    return exp.If(this=cond, true=exp.Null(), false=seq_get(args, 0))
 200
 201
 202def _build_regexp_replace(args: list) -> exp.RegexpReplace:
 203    regexp_replace = exp.RegexpReplace.from_arg_list(args)
 204
 205    if not regexp_replace.args.get("replacement"):
 206        regexp_replace.set("replacement", exp.Literal.string(""))
 207
 208    return regexp_replace
 209
 210
 211def _build_regexp_like(args: list) -> exp.RegexpLike:
 212    return exp.RegexpLike(
 213        this=seq_get(args, 0),
 214        expression=seq_get(args, 1),
 215        flag=seq_get(args, 2),
 216        full_match=True,
 217    )
 218
 219
 220def _date_trunc_to_time(args: list) -> exp.DateTrunc | exp.TimestampTrunc:
 221    trunc = date_trunc_to_time(args)
 222    unit = map_date_part(trunc.args["unit"])
 223    trunc.set("unit", unit)
 224    is_time_input = trunc.this.is_type(exp.DType.TIME, exp.DType.TIMETZ)
 225    if (isinstance(trunc, exp.TimestampTrunc) and is_date_unit(unit) or is_time_input) or (
 226        isinstance(trunc, exp.DateTrunc) and not is_date_unit(unit)
 227    ):
 228        trunc.set("input_type_preserved", True)
 229    return trunc
 230
 231
 232def _build_regexp_extract(expr_type: type[E]) -> t.Callable[[list, Dialect], E]:
 233    def _builder(args: list, dialect: Dialect) -> E:
 234        return expr_type(
 235            this=seq_get(args, 0),
 236            expression=seq_get(args, 1),
 237            position=seq_get(args, 2),
 238            occurrence=seq_get(args, 3),
 239            parameters=seq_get(args, 4),
 240            group=seq_get(args, 5) or exp.Literal.number(0),
 241            **(
 242                {"null_if_pos_overflow": dialect.REGEXP_EXTRACT_POSITION_OVERFLOW_RETURNS_NULL}
 243                if expr_type is exp.RegexpExtract
 244                else {}
 245            ),
 246        )
 247
 248    return _builder
 249
 250
 251def _build_timestamp_from_parts(args: list) -> exp.Func:
 252    """Build TimestampFromParts with support for both syntaxes:
 253    1. TIMESTAMP_FROM_PARTS(year, month, day, hour, minute, second [, nanosecond] [, time_zone])
 254    2. TIMESTAMP_FROM_PARTS(date_expr, time_expr) - Snowflake specific
 255    """
 256    if len(args) == 2:
 257        return exp.TimestampFromParts(this=seq_get(args, 0), expression=seq_get(args, 1))
 258
 259    return exp.TimestampFromParts.from_arg_list(args)
 260
 261
 262def _build_round(args: list) -> exp.Round:
 263    """
 264    Build Round expression, unwrapping Snowflake's named parameters.
 265
 266    Maps EXPR => this, SCALE => decimals, ROUNDING_MODE => truncate.
 267
 268    Note: Snowflake does not support mixing named and positional arguments.
 269    Arguments are either all named or all positional.
 270    """
 271    kwarg_map = {"EXPR": "this", "SCALE": "decimals", "ROUNDING_MODE": "truncate"}
 272    round_args = {}
 273    positional_keys = ["this", "decimals", "truncate"]
 274    positional_idx = 0
 275
 276    for arg in args:
 277        if isinstance(arg, exp.Kwarg):
 278            key = arg.this.name.upper()
 279            round_key = kwarg_map.get(key)
 280            if round_key:
 281                round_args[round_key] = arg.expression
 282        else:
 283            if positional_idx < len(positional_keys):
 284                round_args[positional_keys[positional_idx]] = arg
 285                positional_idx += 1
 286
 287    expression = exp.Round(**round_args)
 288    expression.set("casts_non_integer_decimals", True)
 289    return expression
 290
 291
 292def _build_array_sort(args: list) -> exp.SortArray:
 293    asc = seq_get(args, 1)
 294    nulls_first = seq_get(args, 2)
 295    if nulls_first is None and isinstance(asc, exp.Boolean):
 296        nulls_first = exp.Boolean(this=not asc.this)
 297    return exp.SortArray(this=seq_get(args, 0), asc=asc, nulls_first=nulls_first)
 298
 299
 300def _build_generator(args: list) -> exp.Generator:
 301    """
 302    Build Generator expression, unwrapping Snowflake's named parameters.
 303
 304    Maps ROWCOUNT => rowcount, TIMELIMIT => timelimit.
 305    """
 306    kwarg_map = {"ROWCOUNT": "rowcount", "TIMELIMIT": "timelimit"}
 307    gen_args = {}
 308
 309    positional_keys = ("rowcount", "timelimit")
 310
 311    for i, arg in enumerate(args):
 312        if isinstance(arg, exp.Kwarg):
 313            key = arg.this.name.upper()
 314            gen_key = kwarg_map.get(key)
 315            if gen_key:
 316                gen_args[gen_key] = arg.expression
 317        elif i < len(positional_keys):
 318            gen_args[positional_keys[i]] = arg
 319
 320    return exp.Generator(**gen_args)
 321
 322
 323def _show_parser(*args: t.Any, **kwargs: t.Any) -> t.Callable[[SnowflakeParser], exp.Show]:
 324    def _parse(self: SnowflakeParser) -> exp.Show:
 325        return self._parse_show_snowflake(*args, **kwargs)
 326
 327    return _parse
 328
 329
 330class SnowflakeParser(parser.Parser):
 331    IDENTIFY_PIVOT_STRINGS = True
 332    TYPED_LAMBDA_ARGS = True
 333    DEFAULT_SAMPLING_METHOD = "BERNOULLI"
 334    COLON_IS_VARIANT_EXTRACT = True
 335    JSON_EXTRACT_REQUIRES_JSON_EXPRESSION = True
 336    SUPPORTS_NTH_VALUE_FROM_MODIFIER = True
 337
 338    TYPE_TOKENS = {*parser.Parser.TYPE_TOKENS, TokenType.FILE}
 339    STRUCT_TYPE_TOKENS = {*parser.Parser.STRUCT_TYPE_TOKENS, TokenType.FILE}
 340    NESTED_TYPE_TOKENS = {*parser.Parser.NESTED_TYPE_TOKENS, TokenType.FILE}
 341
 342    ID_VAR_TOKENS = {
 343        *parser.Parser.ID_VAR_TOKENS,
 344        TokenType.EXCEPT,
 345        TokenType.INTEGRATION,
 346        TokenType.MATCH_CONDITION,
 347        TokenType.PACKAGE,
 348        TokenType.POLICY,
 349        TokenType.POOL,
 350        TokenType.ROLE,
 351        TokenType.RULE,
 352        TokenType.VOLUME,
 353    }
 354
 355    ALIAS_TOKENS = parser.Parser.ALIAS_TOKENS | {
 356        TokenType.INTEGRATION,
 357        TokenType.PACKAGE,
 358        TokenType.POLICY,
 359        TokenType.POOL,
 360        TokenType.ROLE,
 361        TokenType.RULE,
 362        TokenType.VOLUME,
 363    }
 364
 365    TABLE_ALIAS_TOKENS = (
 366        parser.Parser.TABLE_ALIAS_TOKENS
 367        | {
 368            TokenType.ANTI,
 369            TokenType.INTEGRATION,
 370            TokenType.PACKAGE,
 371            TokenType.POLICY,
 372            TokenType.POOL,
 373            TokenType.ROLE,
 374            TokenType.RULE,
 375            TokenType.SEMI,
 376            TokenType.VOLUME,
 377            TokenType.WINDOW,
 378        }
 379    ) - {TokenType.MATCH_CONDITION}
 380
 381    COLON_PLACEHOLDER_TOKENS = ID_VAR_TOKENS | {TokenType.NUMBER}
 382
 383    NO_PAREN_FUNCTIONS = {
 384        **parser.Parser.NO_PAREN_FUNCTIONS,
 385        TokenType.LOCALTIME: exp.Localtime,
 386        TokenType.LOCALTIMESTAMP: exp.Localtimestamp,
 387        TokenType.CURRENT_TIME: exp.Localtime,
 388    }
 389
 390    RANGE_PARSERS = {
 391        **parser.Parser.RANGE_PARSERS,
 392        TokenType.RLIKE: lambda self, this: self.expression(
 393            exp.RegexpLike(this=this, expression=self._parse_bitwise(), full_match=True)
 394        ),
 395    }
 396
 397    FUNCTIONS = {
 398        **parser.Parser.FUNCTIONS,
 399        "CHARINDEX": lambda args: exp.StrPosition(
 400            this=seq_get(args, 1),
 401            substr=seq_get(args, 0),
 402            position=seq_get(args, 2),
 403            clamp_position=True,
 404        ),
 405        "ADD_MONTHS": lambda args: exp.AddMonths(
 406            this=seq_get(args, 0),
 407            expression=seq_get(args, 1),
 408            preserve_end_of_month=True,
 409        ),
 410        "APPROX_PERCENTILE": exp.ApproxQuantile.from_arg_list,
 411        "CURRENT_TIME": lambda args: exp.Localtime(this=seq_get(args, 0)),
 412        "APPROX_TOP_K": _build_approx_top_k,
 413        "ARRAY_CONSTRUCT": lambda args: exp.Array(expressions=args),
 414        "ARRAY_CONTAINS": lambda args: exp.ArrayContains(
 415            this=seq_get(args, 1),
 416            expression=seq_get(args, 0),
 417            ensure_variant=False,
 418            check_null=True,
 419        ),
 420        "ARRAY_DISTINCT": lambda args: exp.ArrayDistinct(
 421            this=seq_get(args, 0),
 422            check_null=True,
 423        ),
 424        "ARRAY_GENERATE_RANGE": lambda args: exp.GenerateSeries(
 425            # Snowflake has exclusive end semantics
 426            start=seq_get(args, 0),
 427            end=seq_get(args, 1),
 428            step=seq_get(args, 2),
 429            is_end_exclusive=True,
 430        ),
 431        "ARRAY_EXCEPT": lambda args: exp.ArrayExcept(
 432            this=seq_get(args, 0),
 433            expression=seq_get(args, 1),
 434            is_multiset=True,
 435        ),
 436        "ARRAY_INTERSECTION": lambda args: exp.ArrayIntersect(
 437            expressions=args,
 438            is_multiset=True,
 439        ),
 440        "ARRAY_POSITION": lambda args: exp.ArrayPosition(
 441            this=seq_get(args, 1),
 442            expression=seq_get(args, 0),
 443            zero_based=True,
 444        ),
 445        "ARRAY_SLICE": lambda args: exp.ArraySlice(
 446            this=seq_get(args, 0),
 447            start=seq_get(args, 1),
 448            end=seq_get(args, 2),
 449            zero_based=True,
 450        ),
 451        "ARRAY_SORT": _build_array_sort,
 452        "ARRAY_FLATTEN": exp.Flatten.from_arg_list,
 453        "ARRAY_TO_STRING": lambda args: exp.ArrayToString(
 454            this=seq_get(args, 0),
 455            expression=seq_get(args, 1),
 456            null_is_empty=True,
 457            null_delim_is_null=True,
 458        ),
 459        "ARRAYS_OVERLAP": lambda args: exp.ArrayOverlaps(
 460            this=seq_get(args, 0), expression=seq_get(args, 1), null_safe=True
 461        ),
 462        "BITAND": _build_bitwise(exp.BitwiseAnd, "BITAND"),
 463        "BIT_AND": _build_bitwise(exp.BitwiseAnd, "BITAND"),
 464        "BITNOT": lambda args: exp.BitwiseNot(this=seq_get(args, 0)),
 465        "BIT_NOT": lambda args: exp.BitwiseNot(this=seq_get(args, 0)),
 466        "BITXOR": _build_bitwise(exp.BitwiseXor, "BITXOR"),
 467        "BIT_XOR": _build_bitwise(exp.BitwiseXor, "BITXOR"),
 468        "BITOR": _build_bitwise(exp.BitwiseOr, "BITOR"),
 469        "BIT_OR": _build_bitwise(exp.BitwiseOr, "BITOR"),
 470        "BITSHIFTLEFT": _build_bitwise(exp.BitwiseLeftShift, "BITSHIFTLEFT"),
 471        "BIT_SHIFTLEFT": _build_bitwise(exp.BitwiseLeftShift, "BIT_SHIFTLEFT"),
 472        "BITSHIFTRIGHT": _build_bitwise(exp.BitwiseRightShift, "BITSHIFTRIGHT"),
 473        "BIT_SHIFTRIGHT": _build_bitwise(exp.BitwiseRightShift, "BIT_SHIFTRIGHT"),
 474        "BITANDAGG": exp.BitwiseAndAgg.from_arg_list,
 475        "BITAND_AGG": exp.BitwiseAndAgg.from_arg_list,
 476        "BIT_AND_AGG": exp.BitwiseAndAgg.from_arg_list,
 477        "BIT_ANDAGG": exp.BitwiseAndAgg.from_arg_list,
 478        "BITORAGG": exp.BitwiseOrAgg.from_arg_list,
 479        "BITOR_AGG": exp.BitwiseOrAgg.from_arg_list,
 480        "BIT_OR_AGG": exp.BitwiseOrAgg.from_arg_list,
 481        "BIT_ORAGG": exp.BitwiseOrAgg.from_arg_list,
 482        "BITXORAGG": exp.BitwiseXorAgg.from_arg_list,
 483        "BITXOR_AGG": exp.BitwiseXorAgg.from_arg_list,
 484        "BIT_XOR_AGG": exp.BitwiseXorAgg.from_arg_list,
 485        "BIT_XORAGG": exp.BitwiseXorAgg.from_arg_list,
 486        "BITMAP_OR_AGG": exp.BitmapOrAgg.from_arg_list,
 487        "BOOLAND": lambda args: exp.Booland(
 488            this=seq_get(args, 0), expression=seq_get(args, 1), round_input=True
 489        ),
 490        "BOOLOR": lambda args: exp.Boolor(
 491            this=seq_get(args, 0), expression=seq_get(args, 1), round_input=True
 492        ),
 493        "BOOLNOT": lambda args: exp.Boolnot(this=seq_get(args, 0), round_input=True),
 494        "BOOLXOR": lambda args: exp.Xor(
 495            this=seq_get(args, 0), expression=seq_get(args, 1), round_input=True
 496        ),
 497        "CORR": lambda args: exp.Corr(
 498            this=seq_get(args, 0),
 499            expression=seq_get(args, 1),
 500            null_on_zero_variance=True,
 501        ),
 502        "COUNT_IF": lambda args: exp.CountIf(this=seq_get(args, 0), zero_on_all_null=True),
 503        "DATE": _build_datetime("DATE", exp.DType.DATE),
 504        "DATEFROMPARTS": _build_date_from_parts,
 505        "DATE_FROM_PARTS": _build_date_from_parts,
 506        "DATE_TRUNC": _date_trunc_to_time,
 507        "DATEADD": lambda args: exp.DateAdd(
 508            this=seq_get(args, 2),
 509            expression=seq_get(args, 1),
 510            unit=map_date_part(seq_get(args, 0)),
 511        ),
 512        "DATEDIFF": lambda args: exp.DateDiff(
 513            this=seq_get(args, 2),
 514            expression=seq_get(args, 1),
 515            unit=map_date_part(seq_get(args, 0)),
 516            date_part_boundary=True,
 517        ),
 518        "DAYNAME": lambda args: exp.Dayname(this=seq_get(args, 0), abbreviated=True),
 519        "DAYOFWEEKISO": exp.DayOfWeekIso.from_arg_list,
 520        "DIV0": _build_if_from_div0,
 521        "DIV0NULL": _build_if_from_div0null,
 522        "EDITDISTANCE": lambda args: exp.Levenshtein(
 523            this=seq_get(args, 0), expression=seq_get(args, 1), max_dist=seq_get(args, 2)
 524        ),
 525        "FLATTEN": exp.Explode.from_arg_list,
 526        "GENERATOR": _build_generator,
 527        "GET": exp.GetExtract.from_arg_list,
 528        "GETDATE": exp.CurrentTimestamp.from_arg_list,
 529        "GET_PATH": lambda args, dialect: exp.JSONExtract(
 530            this=seq_get(args, 0),
 531            expression=dialect.to_json_path(seq_get(args, 1)),
 532            requires_json=True,
 533        ),
 534        "GREATEST_IGNORE_NULLS": lambda args: exp.Greatest(
 535            this=seq_get(args, 0), expressions=args[1:], ignore_nulls=True
 536        ),
 537        "LEAST_IGNORE_NULLS": lambda args: exp.Least(
 538            this=seq_get(args, 0), expressions=args[1:], ignore_nulls=True
 539        ),
 540        "LEFT": lambda args: exp.Left(
 541            this=seq_get(args, 0), expression=seq_get(args, 1), negative_length_returns_empty=True
 542        ),
 543        "HEX_DECODE_BINARY": exp.Unhex.from_arg_list,
 544        "HEX_ENCODE": exp.Hex.from_arg_list,
 545        "IFF": exp.If.from_arg_list,
 546        "JAROWINKLER_SIMILARITY": lambda args: exp.JarowinklerSimilarity(
 547            this=seq_get(args, 0),
 548            expression=seq_get(args, 1),
 549            case_insensitive=True,
 550            integer_scale=True,
 551        ),
 552        "MD5_HEX": exp.MD5.from_arg_list,
 553        "MD5_BINARY": exp.MD5Digest.from_arg_list,
 554        "MD5_NUMBER_LOWER64": exp.MD5NumberLower64.from_arg_list,
 555        "MD5_NUMBER_UPPER64": exp.MD5NumberUpper64.from_arg_list,
 556        "MONTHNAME": lambda args: exp.Monthname(this=seq_get(args, 0), abbreviated=True),
 557        "LAST_DAY": lambda args: exp.LastDay(
 558            this=seq_get(args, 0), unit=map_date_part(seq_get(args, 1))
 559        ),
 560        "LEN": lambda args: exp.Length(this=seq_get(args, 0), binary=True),
 561        "LENGTH": lambda args: exp.Length(this=seq_get(args, 0), binary=True),
 562        "LOCALTIMESTAMP": exp.CurrentTimestamp.from_arg_list,
 563        "NULLIFZERO": _build_if_from_nullifzero,
 564        "OBJECT_CONSTRUCT": lambda args: build_object_construct(args),
 565        "OBJECT_KEYS": exp.JSONKeys.from_arg_list,
 566        "OCTET_LENGTH": exp.ByteLength.from_arg_list,
 567        "PARSE_URL": lambda args: exp.ParseUrl(this=seq_get(args, 0), permissive=seq_get(args, 1)),
 568        "REGEXP_EXTRACT_ALL": _build_regexp_extract(exp.RegexpExtractAll),
 569        "REGEXP_LIKE": _build_regexp_like,
 570        "REGEXP_REPLACE": _build_regexp_replace,
 571        "REGEXP_SUBSTR": _build_regexp_extract(exp.RegexpExtract),
 572        "REGEXP_SUBSTR_ALL": _build_regexp_extract(exp.RegexpExtractAll),
 573        "RANDOM": lambda args: exp.Rand(
 574            this=seq_get(args, 0),
 575            lower=exp.Literal.number(-9223372036854775808.0),  # -2^63 as float to avoid overflow
 576            upper=exp.Literal.number(9223372036854775807.0),  # 2^63-1 as float
 577        ),
 578        "REPLACE": build_replace_with_optional_replacement,
 579        "RIGHT": lambda args: exp.Right(
 580            this=seq_get(args, 0), expression=seq_get(args, 1), negative_length_returns_empty=True
 581        ),
 582        "RLIKE": _build_regexp_like,
 583        "ROUND": _build_round,
 584        "SHA1_BINARY": exp.SHA1Digest.from_arg_list,
 585        "SHA1_HEX": exp.SHA.from_arg_list,
 586        "SHA2_BINARY": exp.SHA2Digest.from_arg_list,
 587        "SHA2_HEX": exp.SHA2.from_arg_list,
 588        "SPLIT": lambda args: exp.Split(
 589            this=seq_get(args, 0),
 590            expression=seq_get(args, 1),
 591            null_returns_null=True,
 592            empty_delimiter_returns_whole=True,
 593        ),
 594        "SQUARE": lambda args: exp.Pow(this=seq_get(args, 0), expression=exp.Literal.number(2)),
 595        "STDDEV_SAMP": exp.Stddev.from_arg_list,
 596        "SYSDATE": lambda args: exp.CurrentTimestamp(this=seq_get(args, 0), sysdate=True),
 597        "TABLE": lambda args: exp.TableFromRows(this=seq_get(args, 0)),
 598        "TIMEADD": lambda args: exp.TimeAdd(
 599            this=seq_get(args, 2),
 600            expression=seq_get(args, 1),
 601            unit=map_date_part(seq_get(args, 0)),
 602        ),
 603        "TIMEDIFF": lambda args: exp.DateDiff(
 604            this=seq_get(args, 2),
 605            expression=seq_get(args, 1),
 606            unit=map_date_part(seq_get(args, 0)),
 607            date_part_boundary=True,
 608        ),
 609        "TIME_FROM_PARTS": lambda args: exp.TimeFromParts(
 610            hour=seq_get(args, 0),
 611            min=seq_get(args, 1),
 612            sec=seq_get(args, 2),
 613            nano=seq_get(args, 3),
 614            overflow=True,
 615        ),
 616        "TIMESTAMPADD": lambda args: exp.DateAdd(
 617            this=seq_get(args, 2),
 618            expression=seq_get(args, 1),
 619            unit=map_date_part(seq_get(args, 0)),
 620        ),
 621        "TIMESTAMPDIFF": lambda args: exp.DateDiff(
 622            this=seq_get(args, 2),
 623            expression=seq_get(args, 1),
 624            unit=map_date_part(seq_get(args, 0)),
 625            date_part_boundary=True,
 626        ),
 627        "TIMESTAMPFROMPARTS": _build_timestamp_from_parts,
 628        "TIMESTAMP_FROM_PARTS": _build_timestamp_from_parts,
 629        "TIMESTAMPNTZFROMPARTS": _build_timestamp_from_parts,
 630        "TIMESTAMP_NTZ_FROM_PARTS": _build_timestamp_from_parts,
 631        "TRUNC": lambda args, dialect: build_trunc(
 632            args, dialect, date_trunc_requires_part=False, fractions_supported=True
 633        ),
 634        "TRUNCATE": lambda args, dialect: build_trunc(
 635            args, dialect, date_trunc_requires_part=False, fractions_supported=True
 636        ),
 637        "TRY_DECRYPT": lambda args: exp.Decrypt(
 638            this=seq_get(args, 0),
 639            passphrase=seq_get(args, 1),
 640            aad=seq_get(args, 2),
 641            encryption_method=seq_get(args, 3),
 642            safe=True,
 643        ),
 644        "TRY_DECRYPT_RAW": lambda args: exp.DecryptRaw(
 645            this=seq_get(args, 0),
 646            key=seq_get(args, 1),
 647            iv=seq_get(args, 2),
 648            aad=seq_get(args, 3),
 649            encryption_method=seq_get(args, 4),
 650            aead=seq_get(args, 5),
 651            safe=True,
 652        ),
 653        "TRY_PARSE_JSON": lambda args: exp.ParseJSON(this=seq_get(args, 0), safe=True),
 654        "TRY_TO_BINARY": lambda args: exp.ToBinary(
 655            this=seq_get(args, 0), format=seq_get(args, 1), safe=True
 656        ),
 657        "TRY_TO_BOOLEAN": lambda args: exp.ToBoolean(this=seq_get(args, 0), safe=True),
 658        "TRY_TO_DATE": _build_datetime("TRY_TO_DATE", exp.DType.DATE, safe=True),
 659        **dict.fromkeys(
 660            ("TRY_TO_DECIMAL", "TRY_TO_NUMBER", "TRY_TO_NUMERIC"),
 661            lambda args: _build_to_number(args, safe=True),
 662        ),
 663        "TRY_TO_DOUBLE": lambda args: exp.ToDouble(
 664            this=seq_get(args, 0), format=seq_get(args, 1), safe=True
 665        ),
 666        "TRY_TO_FILE": lambda args: exp.ToFile(
 667            this=seq_get(args, 0), path=seq_get(args, 1), safe=True
 668        ),
 669        "TRY_TO_TIME": _build_datetime("TRY_TO_TIME", exp.DType.TIME, safe=True),
 670        "TRY_TO_TIMESTAMP": _build_datetime("TRY_TO_TIMESTAMP", exp.DType.TIMESTAMP, safe=True),
 671        "TRY_TO_TIMESTAMP_LTZ": _build_datetime(
 672            "TRY_TO_TIMESTAMP_LTZ", exp.DType.TIMESTAMPLTZ, safe=True
 673        ),
 674        "TRY_TO_TIMESTAMP_NTZ": _build_datetime(
 675            "TRY_TO_TIMESTAMP_NTZ", exp.DType.TIMESTAMPNTZ, safe=True
 676        ),
 677        "TRY_TO_TIMESTAMP_TZ": _build_datetime(
 678            "TRY_TO_TIMESTAMP_TZ", exp.DType.TIMESTAMPTZ, safe=True
 679        ),
 680        "TO_CHAR": build_timetostr_or_tochar,
 681        "TO_DATE": _build_datetime("TO_DATE", exp.DType.DATE),
 682        **dict.fromkeys(
 683            ("TO_DECIMAL", "TO_NUMBER", "TO_NUMERIC"),
 684            lambda args: _build_to_number(args),
 685        ),
 686        "TO_TIME": _build_datetime("TO_TIME", exp.DType.TIME),
 687        "TO_TIMESTAMP": _build_datetime("TO_TIMESTAMP", exp.DType.TIMESTAMP),
 688        "TO_TIMESTAMP_LTZ": _build_datetime("TO_TIMESTAMP_LTZ", exp.DType.TIMESTAMPLTZ),
 689        "TO_TIMESTAMP_NTZ": _build_datetime("TO_TIMESTAMP_NTZ", exp.DType.TIMESTAMPNTZ),
 690        "TO_TIMESTAMP_TZ": _build_datetime("TO_TIMESTAMP_TZ", exp.DType.TIMESTAMPTZ),
 691        "TO_GEOGRAPHY": lambda args: (
 692            exp.cast(args[0], exp.DType.GEOGRAPHY)
 693            if len(args) == 1
 694            else exp.Anonymous(this="TO_GEOGRAPHY", expressions=args)
 695        ),
 696        "TO_GEOMETRY": lambda args: (
 697            exp.cast(args[0], exp.DType.GEOMETRY)
 698            if len(args) == 1
 699            else exp.Anonymous(this="TO_GEOMETRY", expressions=args)
 700        ),
 701        "TO_VARCHAR": build_timetostr_or_tochar,
 702        "TO_JSON": exp.JSONFormat.from_arg_list,
 703        "VECTOR_COSINE_SIMILARITY": exp.CosineDistance.from_arg_list,
 704        "VECTOR_INNER_PRODUCT": exp.DotProduct.from_arg_list,
 705        "VECTOR_L1_DISTANCE": exp.ManhattanDistance.from_arg_list,
 706        "VECTOR_L2_DISTANCE": exp.EuclideanDistance.from_arg_list,
 707        "ZEROIFNULL": _build_if_from_zeroifnull,
 708        "LIKE": build_like(exp.Like),
 709        "ILIKE": build_like(exp.ILike),
 710        "SEARCH": _build_search,
 711        "SKEW": exp.Skewness.from_arg_list,
 712        "SPLIT_PART": lambda args: exp.SplitPart(
 713            this=seq_get(args, 0),
 714            delimiter=seq_get(args, 1),
 715            part_index=seq_get(args, 2),
 716            part_index_zero_as_one=True,
 717            empty_delimiter_returns_whole=True,
 718        ),
 719        "STRTOK": lambda args: exp.Strtok(
 720            this=seq_get(args, 0),
 721            delimiter=seq_get(args, 1) or exp.Literal.string(" "),
 722            part_index=seq_get(args, 2) or exp.Literal.number("1"),
 723        ),
 724        "STRTOK_TO_ARRAY": lambda args: exp.StrtokToArray(
 725            this=seq_get(args, 0),
 726            expression=seq_get(args, 1) or exp.Literal.string(" "),
 727        ),
 728        "SYSTIMESTAMP": exp.CurrentTimestamp.from_arg_list,
 729        "IDENTIFIER": exp.DynamicIdentifier.from_arg_list,
 730        "UNICODE": lambda args: exp.Unicode(this=seq_get(args, 0), empty_is_zero=True),
 731        "WEEKISO": exp.WeekOfYear.from_arg_list,
 732        "WEEKOFYEAR": exp.Week.from_arg_list,
 733    }
 734    FUNCTIONS = {k: v for k, v in FUNCTIONS.items() if k != "PREDICT"}
 735
 736    FUNCTION_PARSERS = {
 737        **parser.Parser.FUNCTION_PARSERS,
 738        "DATE_PART": lambda self: self._parse_date_part(),
 739        "DIRECTORY": lambda self: self._parse_directory(),
 740        "OBJECT_CONSTRUCT_KEEP_NULL": lambda self: self._parse_object_construct_keep_null(),
 741        "LISTAGG": lambda self: self._parse_string_agg(),
 742        "SEMANTIC_VIEW": lambda self: self._parse_semantic_view(),
 743        "SUBSTR": lambda self: self._parse_substring(),
 744    }
 745    FUNCTION_PARSERS = {k: v for k, v in FUNCTION_PARSERS.items() if k != "TRIM"}
 746
 747    TIMESTAMPS = parser.Parser.TIMESTAMPS - {TokenType.TIME}
 748
 749    ALTER_PARSERS = {
 750        **parser.Parser.ALTER_PARSERS,
 751        "MODIFY": lambda self: self._parse_alter_table_alter(),
 752        "SESSION": lambda self: self._parse_alter_session(),
 753        "UNSET": lambda self: self.expression(
 754            exp.Set(
 755                tag=self._match_text_seq("TAG"),
 756                expressions=self._parse_csv(self._parse_id_var),
 757                unset=True,
 758            )
 759        ),
 760    }
 761
 762    STATEMENT_PARSERS = {
 763        **parser.Parser.STATEMENT_PARSERS,
 764        TokenType.GET: lambda self: self._parse_get(),
 765        TokenType.PUT: lambda self: self._parse_put(),
 766        TokenType.SHOW: lambda self: self._parse_show(),
 767        TokenType.UNDROP: lambda self: self._parse_undrop(),
 768    }
 769
 770    PROPERTY_PARSERS = {
 771        **parser.Parser.PROPERTY_PARSERS,
 772        "CREDENTIALS": lambda self: self._parse_credentials_property(),
 773        "FILE_FORMAT": lambda self: self._parse_file_format_property(),
 774        "LOCATION": lambda self: self._parse_location_property(),
 775        "ROW": lambda self: (
 776            self._parse_row_access_policy()
 777            if self._match_text_seq("ACCESS", "POLICY")
 778            else self._parse_row()
 779        ),
 780        "TAG": lambda self: self._parse_tag(),
 781        "USING": lambda self: (
 782            self._match_text_seq("TEMPLATE")
 783            and self.expression(exp.UsingTemplateProperty(this=self._parse_statement()))
 784        ),
 785    }
 786
 787    DESCRIBE_QUALIFIER_PARSERS: t.ClassVar[dict[str, t.Callable]] = {
 788        "API": lambda self: self.expression(exp.ApiProperty()),
 789        "APPLICATION": lambda self: self.expression(exp.ApplicationProperty()),
 790        "CATALOG": lambda self: self.expression(exp.CatalogProperty()),
 791        "COMPUTE": lambda self: self.expression(exp.ComputeProperty()),
 792        "DATABASE": lambda self: (
 793            self.expression(exp.DatabaseProperty())
 794            if self._curr and self._curr.text.upper() == "ROLE"
 795            else None
 796        ),
 797        "DYNAMIC": lambda self: self.expression(exp.DynamicProperty()),
 798        "EXTERNAL": lambda self: self.expression(exp.ExternalProperty()),
 799        "HYBRID": lambda self: self.expression(exp.HybridProperty()),
 800        "ICEBERG": lambda self: self.expression(exp.IcebergProperty()),
 801        "MASKING": lambda self: self.expression(exp.MaskingProperty()),
 802        "MATERIALIZED": lambda self: self.expression(exp.MaterializedProperty()),
 803        "NETWORK": lambda self: self.expression(exp.NetworkProperty()),
 804        "ROW": lambda self: (
 805            self.expression(exp.RowAccessProperty()) if self._match_text_seq("ACCESS") else None
 806        ),
 807        "SECURITY": lambda self: (
 808            self.expression(exp.SecurityIntegrationProperty())
 809            if self._curr and self._curr.text.upper() == "INTEGRATION"
 810            else None
 811        ),
 812    }
 813
 814    TYPE_CONVERTERS = {
 815        # https://docs.snowflake.com/en/sql-reference/data-types-numeric#number
 816        exp.DType.DECIMAL: build_default_decimal_type(precision=38, scale=0),
 817    }
 818
 819    SHOW_PARSERS = {
 820        "DATABASES": _show_parser("DATABASES"),
 821        "SCHEMAS": _show_parser("SCHEMAS"),
 822        "OBJECTS": _show_parser("OBJECTS"),
 823        "TABLES": _show_parser("TABLES"),
 824        "VIEWS": _show_parser("VIEWS"),
 825        "PRIMARY KEYS": _show_parser("PRIMARY KEYS"),
 826        "IMPORTED KEYS": _show_parser("IMPORTED KEYS"),
 827        "UNIQUE KEYS": _show_parser("UNIQUE KEYS"),
 828        "SEQUENCES": _show_parser("SEQUENCES"),
 829        "STAGES": _show_parser("STAGES"),
 830        "COLUMNS": _show_parser("COLUMNS"),
 831        "USERS": _show_parser("USERS"),
 832        "FILE FORMATS": _show_parser("FILE FORMATS"),
 833        "FUNCTIONS": _show_parser("FUNCTIONS"),
 834        "PROCEDURES": _show_parser("PROCEDURES"),
 835        "WAREHOUSES": _show_parser("WAREHOUSES"),
 836        "ICEBERG TABLES": _show_parser("TABLES", iceberg=True),
 837        "TERSE ICEBERG TABLES": _show_parser("TABLES", terse=True, iceberg=True),
 838        "TERSE DATABASES": _show_parser("DATABASES", terse=True),
 839        "TERSE SCHEMAS": _show_parser("SCHEMAS", terse=True),
 840        "TERSE OBJECTS": _show_parser("OBJECTS", terse=True),
 841        "TERSE TABLES": _show_parser("TABLES", terse=True),
 842        "TERSE VIEWS": _show_parser("VIEWS", terse=True),
 843        "TERSE SEQUENCES": _show_parser("SEQUENCES", terse=True),
 844        "TERSE USERS": _show_parser("USERS", terse=True),
 845        # TERSE has no semantic effect for KEYS, so we do not set the terse AST arg
 846        "TERSE PRIMARY KEYS": _show_parser("PRIMARY KEYS"),
 847        "TERSE IMPORTED KEYS": _show_parser("IMPORTED KEYS"),
 848        "TERSE UNIQUE KEYS": _show_parser("UNIQUE KEYS"),
 849    }
 850
 851    SHOW_TRIE = new_trie(key.split(" ") for key in SHOW_PARSERS)
 852
 853    CONSTRAINT_PARSERS = {
 854        **parser.Parser.CONSTRAINT_PARSERS,
 855        "WITH": lambda self: self._parse_with_constraint(),
 856        "MASKING": lambda self: self._parse_with_constraint(),
 857        "PROJECTION": lambda self: self._parse_with_constraint(),
 858        "TAG": lambda self: self._parse_with_constraint(),
 859    }
 860
 861    STAGED_FILE_SINGLE_TOKENS: t.ClassVar = {
 862        TokenType.DOT,
 863        TokenType.MOD,
 864        TokenType.SLASH,
 865    }
 866
 867    FLATTEN_COLUMNS: t.ClassVar = ["SEQ", "KEY", "PATH", "INDEX", "VALUE", "THIS"]
 868
 869    SCHEMA_KINDS: t.ClassVar = {
 870        "OBJECTS",
 871        "TABLES",
 872        "VIEWS",
 873        "SEQUENCES",
 874        "UNIQUE KEYS",
 875        "IMPORTED KEYS",
 876    }
 877
 878    NON_TABLE_CREATABLES: t.ClassVar = {"STORAGE INTEGRATION", "TAG", "WAREHOUSE", "STREAMLIT"}
 879
 880    UNDROP_OBJECTS: t.ClassVar[parser.OPTIONS_TYPE] = {
 881        **dict.fromkeys(
 882            (
 883                "ACCOUNT",
 884                "DATABASE",
 885                "NOTEBOOK",
 886                "SCHEMA",
 887                "SNAPSHOT",
 888                "STREAMLIT",
 889                "TABLE",
 890                "TAG",
 891                "TYPE",
 892            ),
 893            tuple(),
 894        ),
 895        "DYNAMIC": ("TABLE",),
 896        "EXTERNAL": ("VOLUME",),
 897        "ICEBERG": ("TABLE",),
 898    }
 899
 900    CREATABLES = {
 901        *parser.Parser.CREATABLES,
 902        TokenType.INTEGRATION,
 903        TokenType.PACKAGE,
 904        TokenType.POLICY,
 905        TokenType.POOL,
 906        TokenType.ROLE,
 907        TokenType.RULE,
 908        TokenType.VOLUME,
 909    }
 910
 911    LAMBDAS = {
 912        **parser.Parser.LAMBDAS,
 913        TokenType.ARROW: lambda self, expressions: self.expression(
 914            exp.Lambda(
 915                this=self._replace_lambda(
 916                    self._parse_assignment(),
 917                    expressions,
 918                ),
 919                expressions=[e.this if isinstance(e, exp.Cast) else e for e in expressions],
 920            )
 921        ),
 922    }
 923
 924    COLUMN_OPERATORS = {
 925        **parser.Parser.COLUMN_OPERATORS,
 926        TokenType.EXCLAMATION: lambda self, this, attr: self.expression(
 927            exp.ModelAttribute(this=this, expression=attr)
 928        ),
 929    }
 930
 931    def _parse_directory(self) -> exp.DirectoryStage:
 932        table = self._parse_table_parts()
 933        this = table.this if isinstance(table, exp.Table) else table
 934        return self.expression(exp.DirectoryStage(this=this))
 935
 936    def _parse_describe(self) -> exp.Describe:
 937        index = self._index
 938
 939        if self._match_texts(self.DESCRIBE_QUALIFIER_PARSERS):
 940            qualifier = self.DESCRIBE_QUALIFIER_PARSERS[self._prev.text.upper()](self)
 941
 942            if qualifier:
 943                kind = self._match_set(self.CREATABLES) and self._prev.text.upper()
 944
 945                if kind:
 946                    this = self._parse_table(schema=True)
 947                    properties = self.expression(exp.Properties(expressions=[qualifier]))
 948                    post_props = self._parse_properties()
 949                    expressions = post_props.expressions if post_props else None
 950                    return self.expression(
 951                        exp.Describe(
 952                            this=this,
 953                            kind=kind,
 954                            properties=properties,
 955                            expressions=expressions,
 956                        )
 957                    )
 958
 959        self._retreat(index)
 960        return super()._parse_describe()
 961
 962    def _parse_use(self) -> exp.Use:
 963        if self._match_text_seq("SECONDARY", "ROLES"):
 964            this = self._match_texts(("ALL", "NONE")) and exp.var(self._prev.text.upper())
 965            roles = None if this else self._parse_csv(lambda: self._parse_table(schema=False))
 966            return self.expression(exp.Use(kind="SECONDARY ROLES", this=this, expressions=roles))
 967
 968        return super()._parse_use()
 969
 970    def _negate_range(self, this: exp.Expr | None = None) -> exp.Expr | None:
 971        if not this:
 972            return this
 973
 974        query = this.args.get("query")
 975        if isinstance(this, exp.In) and isinstance(query, exp.Query):
 976            # Snowflake treats `value NOT IN (subquery)` as `VALUE <> ALL (subquery)`, so
 977            # we do this conversion here to avoid parsing it into `NOT value IN (subquery)`
 978            # which can produce different results (most likely a SnowFlake bug).
 979            #
 980            # https://docs.snowflake.com/en/sql-reference/functions/in
 981            # Context: https://github.com/tobymao/sqlglot/issues/3890
 982            return self.expression(exp.NEQ(this=this.this, expression=exp.All(this=query.unnest())))
 983
 984        return self.expression(exp.Not(this=this))
 985
 986    def _parse_tag(self) -> exp.Tags:
 987        return self.expression(exp.Tags(expressions=self._parse_wrapped_csv(self._parse_property)))
 988
 989    def _parse_property_before(self) -> exp.Expr | list[exp.Expr] | None:
 990        prop = super()._parse_property_before()
 991        if prop:
 992            return prop
 993
 994        if not self._next or self._next.token_type != TokenType.EQ:
 995            return None
 996
 997        return self._parse_sequence_properties() or self._parse_key_value_property(
 998            self._parse_primary_or_var
 999        )
1000
1001    def _parse_with_constraint(self) -> exp.Expr | None:
1002        if self._prev.token_type != TokenType.WITH:
1003            self._retreat(self._index - 1)
1004
1005        if self._match_text_seq("MASKING", "POLICY"):
1006            policy = self._parse_column()
1007            return self.expression(
1008                exp.MaskingPolicyColumnConstraint(
1009                    this=policy.to_dot() if isinstance(policy, exp.Column) else policy,
1010                    expressions=self._match(TokenType.USING)
1011                    and self._parse_wrapped_csv(self._parse_id_var),
1012                )
1013            )
1014        if self._match_text_seq("PROJECTION", "POLICY"):
1015            policy = self._parse_column()
1016            return self.expression(
1017                exp.ProjectionPolicyColumnConstraint(
1018                    this=policy.to_dot() if isinstance(policy, exp.Column) else policy
1019                )
1020            )
1021        if self._match(TokenType.TAG):
1022            return self._parse_tag()
1023
1024        return None
1025
1026    def _parse_with_property(self) -> exp.Expr | None | list[exp.Expr]:
1027        if self._match(TokenType.TAG):
1028            return self._parse_tag()
1029
1030        if self._match_text_seq("ROW", "ACCESS", "POLICY"):
1031            return self._parse_row_access_policy()
1032
1033        return super()._parse_with_property()
1034
1035    def _parse_row_access_policy(self) -> exp.RowAccessProperty:
1036        # GET_DDL outputs #unknown_policy when the user lacks privileges to see the policy name
1037        if self._match(TokenType.HASH):
1038            policy: exp.Expr | None = self._parse_var(any_token=True)
1039            if policy:
1040                policy = exp.Var(this=f"#{policy.name}")
1041            expressions = None
1042        else:
1043            policy = self._parse_column()
1044            if isinstance(policy, exp.Column):
1045                policy = policy.to_dot()
1046            if not self._match(TokenType.ON):
1047                self.raise_error("Expected ON after ROW ACCESS POLICY name")
1048            expressions = self._parse_wrapped_csv(self._parse_id_var)
1049
1050        return self.expression(exp.RowAccessProperty(this=policy, expressions=expressions))
1051
1052    def _parse_create(self) -> exp.Create | exp.Command:
1053        expression = super()._parse_create()
1054        if isinstance(expression, exp.Create) and expression.kind in self.NON_TABLE_CREATABLES:
1055            # Replace the Table node with the enclosed Identifier
1056            expression.this.replace(expression.this.this)
1057
1058        return expression
1059
1060    def _parse_object_construct_keep_null(self) -> exp.JSONObject:
1061        index = self._index
1062        column = self._parse_column()
1063        if column and column.is_star:
1064            return self.expression(exp.JSONObject(expressions=[column]))
1065
1066        self._retreat(index)
1067        return self._parse_json_object(agg=False)
1068
1069    # https://docs.snowflake.com/en/sql-reference/functions/date_part.html
1070    # https://docs.snowflake.com/en/sql-reference/functions-date-time.html#label-supported-date-time-parts
1071    def _parse_date_part(self) -> exp.Expr | None:
1072        this = self._parse_var() or self._parse_type()
1073
1074        if not this:
1075            return None
1076
1077        # Handle both syntaxes: DATE_PART(part, expr) and DATE_PART(part FROM expr)
1078        expression = self._match_set((TokenType.FROM, TokenType.COMMA)) and self._parse_bitwise()
1079        return self.expression(
1080            exp.Extract(this=map_date_part(this, self.dialect), expression=expression)
1081        )
1082
1083    def _parse_bracket_key_value(self, is_map: bool = False) -> exp.Expr | None:
1084        if is_map:
1085            # Keys are strings in Snowflake's objects, see also:
1086            # - https://docs.snowflake.com/en/sql-reference/data-types-semistructured
1087            # - https://docs.snowflake.com/en/sql-reference/functions/object_construct
1088            return self._parse_slice(self._parse_string()) or self._parse_assignment()
1089
1090        return self._parse_slice(self._parse_alias(self._parse_assignment(), explicit=True))
1091
1092    def _parse_lateral(self) -> exp.Lateral | None:
1093        lateral = super()._parse_lateral()
1094        if not lateral:
1095            return lateral
1096
1097        if isinstance(lateral.this, exp.Explode):
1098            table_alias = lateral.args.get("alias")
1099            columns = [exp.to_identifier(col) for col in self.FLATTEN_COLUMNS]
1100            if table_alias and not table_alias.args.get("columns"):
1101                table_alias.set("columns", columns)
1102            elif not table_alias:
1103                exp.alias_(lateral, "_flattened", table=columns, copy=False)
1104
1105        return lateral
1106
1107    def _parse_table_parts(
1108        self,
1109        schema: bool = False,
1110        is_db_reference: bool = False,
1111        wildcard: bool = False,
1112        fast: bool = False,
1113    ) -> exp.Table | exp.Dot | None:
1114        # https://docs.snowflake.com/en/user-guide/querying-stage
1115        if self._match(TokenType.STRING, advance=False):
1116            table = self._parse_string()
1117        elif self._match_text_seq("@", advance=False):
1118            table = self._parse_location_path()
1119        else:
1120            table = None
1121
1122        if table:
1123            file_format = None
1124            pattern = None
1125
1126            wrapped = self._match(TokenType.L_PAREN)
1127            while self._curr and wrapped and not self._match(TokenType.R_PAREN):
1128                if self._match_text_seq("FILE_FORMAT", "=>"):
1129                    file_format = self._parse_string() or super()._parse_table_parts(
1130                        is_db_reference=is_db_reference
1131                    )
1132                elif self._match_text_seq("PATTERN", "=>"):
1133                    pattern = self._parse_string()
1134                else:
1135                    break
1136
1137                self._match(TokenType.COMMA)
1138
1139            table = self.expression(exp.Table(this=table, format=file_format, pattern=pattern))
1140        else:
1141            table = super()._parse_table_parts(
1142                schema=schema,
1143                is_db_reference=is_db_reference,
1144                fast=fast,
1145            )
1146
1147        return table
1148
1149    def _parse_function_call(
1150        self,
1151        functions: dict[str, t.Callable] | None = None,
1152        anonymous: bool = False,
1153        optional_parens: bool = True,
1154        any_token: bool = False,
1155    ) -> exp.Expr | None:
1156        this = super()._parse_function_call(
1157            functions=functions,
1158            anonymous=anonymous,
1159            optional_parens=optional_parens,
1160            any_token=any_token,
1161        )
1162
1163        # Snowflake can invoke a function whose name is dynamically resolved, e.g.
1164        # `IDENTIFIER('my_func')(1, 2)`. The trailing argument list is the call's arguments.
1165        #
1166        # https://docs.snowflake.com/en/sql-reference/identifier-literal
1167        if isinstance(this, exp.DynamicIdentifier) and self._match(
1168            TokenType.L_PAREN, advance=False
1169        ):
1170            this.set("expressions", self._parse_wrapped_csv(self._parse_lambda))
1171
1172        return this
1173
1174    def _parse_id_var(
1175        self,
1176        any_token: bool = True,
1177        tokens: Collection[TokenType] | None = None,
1178    ) -> exp.Expr | None:
1179        if self._match_text_seq("IDENTIFIER", "("):
1180            identifier = (
1181                super()._parse_id_var(any_token=any_token, tokens=tokens) or self._parse_string()
1182            )
1183            self._match_r_paren()
1184            return self.expression(exp.DynamicIdentifier(this=identifier))
1185
1186        return super()._parse_id_var(any_token=any_token, tokens=tokens)
1187
1188    def _parse_show_snowflake(
1189        self, this: str, terse: bool = False, iceberg: bool = False
1190    ) -> exp.Show:
1191        scope = None
1192        scope_kind = None
1193
1194        history = self._match_text_seq("HISTORY")
1195
1196        like = self._parse_string() if self._match(TokenType.LIKE) else None
1197
1198        if self._match(TokenType.IN):
1199            if self._match_text_seq("ACCOUNT"):
1200                scope_kind = "ACCOUNT"
1201            elif self._match_text_seq("CLASS"):
1202                scope_kind = "CLASS"
1203                scope = self._parse_table_parts()
1204            elif self._match_text_seq("APPLICATION"):
1205                scope_kind = "APPLICATION"
1206                if self._match_text_seq("PACKAGE"):
1207                    scope_kind += " PACKAGE"
1208                scope = self._parse_table_parts()
1209            elif self._match_set(self.DB_CREATABLES):
1210                scope_kind = self._prev.text.upper()
1211                if self._curr:
1212                    scope = self._parse_table_parts()
1213            elif self._curr:
1214                scope_kind = "SCHEMA" if this in self.SCHEMA_KINDS else "TABLE"
1215                scope = self._parse_table_parts()
1216
1217        return self.expression(
1218            exp.Show(
1219                terse=terse,
1220                iceberg=iceberg,
1221                this=this,
1222                history=history,
1223                like=like,
1224                scope=scope,
1225                scope_kind=scope_kind,
1226                starts_with=self._match_text_seq("STARTS", "WITH") and self._parse_string(),
1227                limit=self._parse_limit(),
1228                from_=self._parse_string() if self._match(TokenType.FROM) else None,
1229                privileges=self._match_text_seq("WITH", "PRIVILEGES")
1230                and self._parse_csv(lambda: self._parse_var(any_token=True, upper=True)),
1231            )
1232        )
1233
1234    def _parse_undrop(self) -> exp.Undrop | exp.Command:
1235        start = self._prev
1236        kind = self._parse_var_from_options(self.UNDROP_OBJECTS, raise_unmatched=False)
1237        if not kind:
1238            return self._parse_as_command(start)
1239
1240        this = self._parse_table_parts(
1241            is_db_reference=kind.name in ("ACCOUNT", "DATABASE", "SCHEMA")
1242        )
1243        rename = self._parse_table_parts() if self._match_text_seq("RENAME", "TO") else None
1244        return self.expression(exp.Undrop(this=this, kind=kind.name, rename=rename))
1245
1246    def _parse_put(self) -> exp.Put | exp.Command:
1247        if self._curr.token_type != TokenType.STRING:
1248            return self._parse_as_command(self._prev)
1249
1250        return self.expression(
1251            exp.Put(
1252                this=self._parse_string(),
1253                target=self._parse_location_path(),
1254                properties=self._parse_properties(),
1255            )
1256        )
1257
1258    def _parse_get(self) -> exp.Expr | None:
1259        start = self._prev
1260
1261        # If we detect GET( then we need to parse a function, not a statement
1262        if self._match(TokenType.L_PAREN):
1263            self._retreat(self._index - 2)
1264            return self._parse_expression()
1265
1266        target = self._parse_location_path()
1267
1268        # Parse as command if unquoted file path
1269        if self._curr.token_type == TokenType.URI_START:
1270            return self._parse_as_command(start)
1271
1272        return self.expression(
1273            exp.Get(this=self._parse_string(), target=target, properties=self._parse_properties())
1274        )
1275
1276    def _parse_location_property(self) -> exp.LocationProperty:
1277        self._match(TokenType.EQ)
1278        return self.expression(exp.LocationProperty(this=self._parse_location_path()))
1279
1280    def _parse_file_location(self) -> exp.Expr | None:
1281        # Parse either a subquery or a staged file
1282        return (
1283            self._parse_select(table=True, parse_subquery_alias=False)
1284            if self._match(TokenType.L_PAREN, advance=False)
1285            else self._parse_table_parts()
1286        )
1287
1288    def _parse_location_path(self) -> exp.Var:
1289        start = self._curr
1290        self._advance_any(ignore_reserved=True)
1291
1292        # We avoid consuming a comma token because external tables like @foo and @bar
1293        # can be joined in a query with a comma separator, as well as closing paren
1294        # in case of subqueries
1295        while self._is_connected() and not self._match_set(
1296            (TokenType.COMMA, TokenType.L_PAREN, TokenType.R_PAREN), advance=False
1297        ):
1298            self._advance_any(ignore_reserved=True)
1299
1300        return exp.var(self._find_sql(start, self._prev))
1301
1302    def _parse_lambda_arg(self) -> exp.Expr | None:
1303        this = super()._parse_lambda_arg()
1304
1305        if not this:
1306            return this
1307
1308        typ = self._parse_types()
1309
1310        if typ:
1311            return self.expression(exp.Cast(this=this, to=typ))
1312
1313        return this
1314
1315    def _parse_foreign_key(self) -> exp.ForeignKey:
1316        # inlineFK, the REFERENCES columns are implied
1317        if self._match(TokenType.REFERENCES, advance=False):
1318            return self.expression(exp.ForeignKey())
1319
1320        # outoflineFK, explicitly names the columns
1321        return super()._parse_foreign_key()
1322
1323    def _parse_file_format_property(self) -> exp.FileFormatProperty:
1324        self._match(TokenType.EQ)
1325        if self._match(TokenType.L_PAREN, advance=False):
1326            expressions = self._parse_wrapped_options()
1327        else:
1328            expressions = [self._parse_format_name()]
1329
1330        return self.expression(exp.FileFormatProperty(expressions=expressions))
1331
1332    def _parse_credentials_property(self) -> exp.CredentialsProperty:
1333        return self.expression(exp.CredentialsProperty(expressions=self._parse_wrapped_options()))
1334
1335    def _parse_semantic_view(self) -> exp.SemanticView:
1336        kwargs: dict[str, t.Any] = {"this": self._parse_table_parts()}
1337
1338        while self._curr and not self._match(TokenType.R_PAREN, advance=False):
1339            if self._match_texts(("DIMENSIONS", "METRICS", "FACTS")):
1340                keyword = self._prev.text.lower()
1341                kwargs[keyword] = self._parse_csv(
1342                    lambda: self._parse_alias(self._parse_disjunction(), explicit=True)
1343                )
1344            elif self._match_text_seq("WHERE"):
1345                kwargs["where"] = self._parse_expression()
1346            else:
1347                self.raise_error("Expecting ) or encountered unexpected keyword")
1348                break
1349
1350        return self.expression(exp.SemanticView(**kwargs))
1351
1352    def _parse_set(self, unset: bool = False, tag: bool = False) -> exp.Set | exp.Command:
1353        set = super()._parse_set(unset=unset, tag=tag)
1354
1355        if isinstance(set, exp.Set):
1356            for expr in set.expressions:
1357                if isinstance(expr, exp.SetItem):
1358                    expr.set("kind", "VARIABLE")
1359        return set
1360
1361    def _parse_position(self, haystack_first: bool = False) -> exp.StrPosition:
1362        result = super()._parse_position(haystack_first)
1363        result.set("clamp_position", True)
1364        return result
1365
1366    def _parse_substring(self) -> exp.Substring:
1367        result = super()._parse_substring()
1368        result.set("zero_start", True)
1369        return result
1370
1371    def build_cast(self, strict: bool, **kwargs) -> exp.Expr:
1372        to = kwargs.get("to")
1373        if not strict and to and to.this == exp.DataType.Type.BOOLEAN:
1374            return self.expression(exp.ToBoolean(this=kwargs.get("this"), safe=True))
1375        cast = super().build_cast(strict, **kwargs)
1376        if isinstance(cast, exp.TryCast) and to:
1377            if to.this in exp.DataType.TEXT_TYPES and to.expressions:
1378                cast.set("null_on_text_overflow", True)
1379            elif to.this == exp.DType.DATE and cast.this.is_string:
1380                cast.set("probe_date_format", True)
1381        return cast
1382
1383    def _parse_window(self, this: exp.Expr | None, alias: bool = False) -> exp.Expr | None:
1384        result = super()._parse_window(this, alias)
1385
1386        # Set default window frame for ranking functions if not present
1387        if (
1388            isinstance(result, exp.Window)
1389            and isinstance(this, RANKING_WINDOW_FUNCTIONS_WITH_FRAME)
1390            and not result.args.get("spec")
1391        ):
1392            frame = exp.WindowSpec(
1393                kind="ROWS",
1394                start="UNBOUNDED",
1395                start_side="PRECEDING",
1396                end="UNBOUNDED",
1397                end_side="FOLLOWING",
1398            )
1399            result.set("spec", frame)
1400        return result
1401
1402
1403# This is imported and used by both the parser (above) and the generator in the dialect file
1404RANKING_WINDOW_FUNCTIONS_WITH_FRAME = (
1405    exp.FirstValue,
1406    exp.LastValue,
1407    exp.NthValue,
1408)
1409
1410
1411def build_object_construct(args: list) -> exp.StarMap | exp.Struct:
1412    expression = parser.build_var_map(args)
1413
1414    if isinstance(expression, exp.StarMap):
1415        return expression
1416
1417    return exp.Struct(
1418        expressions=[
1419            exp.PropertyEQ(this=k, expression=v) for k, v in zip(expression.keys, expression.values)
1420        ]
1421    )