sqlglot.parsers.snowflake

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