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
 337    TYPE_TOKENS = {*parser.Parser.TYPE_TOKENS, TokenType.FILE}
 338    STRUCT_TYPE_TOKENS = {*parser.Parser.STRUCT_TYPE_TOKENS, TokenType.FILE}
 339    NESTED_TYPE_TOKENS = {*parser.Parser.NESTED_TYPE_TOKENS, TokenType.FILE}
 340
 341    ID_VAR_TOKENS = {
 342        *parser.Parser.ID_VAR_TOKENS,
 343        TokenType.EXCEPT,
 344        TokenType.INTEGRATION,
 345        TokenType.MATCH_CONDITION,
 346        TokenType.PACKAGE,
 347        TokenType.POLICY,
 348        TokenType.POOL,
 349        TokenType.ROLE,
 350        TokenType.RULE,
 351        TokenType.VOLUME,
 352    }
 353
 354    ALIAS_TOKENS = parser.Parser.ALIAS_TOKENS | {
 355        TokenType.INTEGRATION,
 356        TokenType.PACKAGE,
 357        TokenType.POLICY,
 358        TokenType.POOL,
 359        TokenType.ROLE,
 360        TokenType.RULE,
 361        TokenType.VOLUME,
 362    }
 363
 364    TABLE_ALIAS_TOKENS = (
 365        parser.Parser.TABLE_ALIAS_TOKENS
 366        | {
 367            TokenType.ANTI,
 368            TokenType.INTEGRATION,
 369            TokenType.PACKAGE,
 370            TokenType.POLICY,
 371            TokenType.POOL,
 372            TokenType.ROLE,
 373            TokenType.RULE,
 374            TokenType.SEMI,
 375            TokenType.VOLUME,
 376            TokenType.WINDOW,
 377        }
 378    ) - {TokenType.MATCH_CONDITION}
 379
 380    COLON_PLACEHOLDER_TOKENS = ID_VAR_TOKENS | {TokenType.NUMBER}
 381
 382    NO_PAREN_FUNCTIONS = {
 383        **parser.Parser.NO_PAREN_FUNCTIONS,
 384        TokenType.LOCALTIME: exp.Localtime,
 385        TokenType.LOCALTIMESTAMP: exp.Localtimestamp,
 386        TokenType.CURRENT_TIME: exp.Localtime,
 387    }
 388
 389    RANGE_PARSERS = {
 390        **parser.Parser.RANGE_PARSERS,
 391        TokenType.RLIKE: lambda self, this: self.expression(
 392            exp.RegexpLike(this=this, expression=self._parse_bitwise(), full_match=True)
 393        ),
 394    }
 395
 396    FUNCTIONS = {
 397        **parser.Parser.FUNCTIONS,
 398        "CHARINDEX": lambda args: exp.StrPosition(
 399            this=seq_get(args, 1),
 400            substr=seq_get(args, 0),
 401            position=seq_get(args, 2),
 402            clamp_position=True,
 403        ),
 404        "ADD_MONTHS": lambda args: exp.AddMonths(
 405            this=seq_get(args, 0),
 406            expression=seq_get(args, 1),
 407            preserve_end_of_month=True,
 408        ),
 409        "APPROX_PERCENTILE": exp.ApproxQuantile.from_arg_list,
 410        "CURRENT_TIME": lambda args: exp.Localtime(this=seq_get(args, 0)),
 411        "APPROX_TOP_K": _build_approx_top_k,
 412        "ARRAY_CONSTRUCT": lambda args: exp.Array(expressions=args),
 413        "ARRAY_CONTAINS": lambda args: exp.ArrayContains(
 414            this=seq_get(args, 1),
 415            expression=seq_get(args, 0),
 416            ensure_variant=False,
 417            check_null=True,
 418        ),
 419        "ARRAY_DISTINCT": lambda args: exp.ArrayDistinct(
 420            this=seq_get(args, 0),
 421            check_null=True,
 422        ),
 423        "ARRAY_GENERATE_RANGE": lambda args: exp.GenerateSeries(
 424            # Snowflake has exclusive end semantics
 425            start=seq_get(args, 0),
 426            end=seq_get(args, 1),
 427            step=seq_get(args, 2),
 428            is_end_exclusive=True,
 429        ),
 430        "ARRAY_EXCEPT": lambda args: exp.ArrayExcept(
 431            this=seq_get(args, 0),
 432            expression=seq_get(args, 1),
 433            is_multiset=True,
 434        ),
 435        "ARRAY_INTERSECTION": lambda args: exp.ArrayIntersect(
 436            expressions=args,
 437            is_multiset=True,
 438        ),
 439        "ARRAY_POSITION": lambda args: exp.ArrayPosition(
 440            this=seq_get(args, 1),
 441            expression=seq_get(args, 0),
 442            zero_based=True,
 443        ),
 444        "ARRAY_SLICE": lambda args: exp.ArraySlice(
 445            this=seq_get(args, 0),
 446            start=seq_get(args, 1),
 447            end=seq_get(args, 2),
 448            zero_based=True,
 449        ),
 450        "ARRAY_SORT": _build_array_sort,
 451        "ARRAY_FLATTEN": exp.Flatten.from_arg_list,
 452        "ARRAY_TO_STRING": lambda args: exp.ArrayToString(
 453            this=seq_get(args, 0),
 454            expression=seq_get(args, 1),
 455            null_is_empty=True,
 456            null_delim_is_null=True,
 457        ),
 458        "ARRAYS_OVERLAP": lambda args: exp.ArrayOverlaps(
 459            this=seq_get(args, 0), expression=seq_get(args, 1), null_safe=True
 460        ),
 461        "BITAND": _build_bitwise(exp.BitwiseAnd, "BITAND"),
 462        "BIT_AND": _build_bitwise(exp.BitwiseAnd, "BITAND"),
 463        "BITNOT": lambda args: exp.BitwiseNot(this=seq_get(args, 0)),
 464        "BIT_NOT": lambda args: exp.BitwiseNot(this=seq_get(args, 0)),
 465        "BITXOR": _build_bitwise(exp.BitwiseXor, "BITXOR"),
 466        "BIT_XOR": _build_bitwise(exp.BitwiseXor, "BITXOR"),
 467        "BITOR": _build_bitwise(exp.BitwiseOr, "BITOR"),
 468        "BIT_OR": _build_bitwise(exp.BitwiseOr, "BITOR"),
 469        "BITSHIFTLEFT": _build_bitwise(exp.BitwiseLeftShift, "BITSHIFTLEFT"),
 470        "BIT_SHIFTLEFT": _build_bitwise(exp.BitwiseLeftShift, "BIT_SHIFTLEFT"),
 471        "BITSHIFTRIGHT": _build_bitwise(exp.BitwiseRightShift, "BITSHIFTRIGHT"),
 472        "BIT_SHIFTRIGHT": _build_bitwise(exp.BitwiseRightShift, "BIT_SHIFTRIGHT"),
 473        "BITANDAGG": exp.BitwiseAndAgg.from_arg_list,
 474        "BITAND_AGG": exp.BitwiseAndAgg.from_arg_list,
 475        "BIT_AND_AGG": exp.BitwiseAndAgg.from_arg_list,
 476        "BIT_ANDAGG": exp.BitwiseAndAgg.from_arg_list,
 477        "BITORAGG": exp.BitwiseOrAgg.from_arg_list,
 478        "BITOR_AGG": exp.BitwiseOrAgg.from_arg_list,
 479        "BIT_OR_AGG": exp.BitwiseOrAgg.from_arg_list,
 480        "BIT_ORAGG": exp.BitwiseOrAgg.from_arg_list,
 481        "BITXORAGG": exp.BitwiseXorAgg.from_arg_list,
 482        "BITXOR_AGG": exp.BitwiseXorAgg.from_arg_list,
 483        "BIT_XOR_AGG": exp.BitwiseXorAgg.from_arg_list,
 484        "BIT_XORAGG": exp.BitwiseXorAgg.from_arg_list,
 485        "BITMAP_OR_AGG": exp.BitmapOrAgg.from_arg_list,
 486        "BOOLAND": lambda args: exp.Booland(
 487            this=seq_get(args, 0), expression=seq_get(args, 1), round_input=True
 488        ),
 489        "BOOLOR": lambda args: exp.Boolor(
 490            this=seq_get(args, 0), expression=seq_get(args, 1), round_input=True
 491        ),
 492        "BOOLNOT": lambda args: exp.Boolnot(this=seq_get(args, 0), round_input=True),
 493        "BOOLXOR": lambda args: exp.Xor(
 494            this=seq_get(args, 0), expression=seq_get(args, 1), round_input=True
 495        ),
 496        "CORR": lambda args: exp.Corr(
 497            this=seq_get(args, 0),
 498            expression=seq_get(args, 1),
 499            null_on_zero_variance=True,
 500        ),
 501        "DATE": _build_datetime("DATE", exp.DType.DATE),
 502        "DATEFROMPARTS": _build_date_from_parts,
 503        "DATE_FROM_PARTS": _build_date_from_parts,
 504        "DATE_TRUNC": _date_trunc_to_time,
 505        "DATEADD": lambda args: exp.DateAdd(
 506            this=seq_get(args, 2),
 507            expression=seq_get(args, 1),
 508            unit=map_date_part(seq_get(args, 0)),
 509        ),
 510        "DATEDIFF": lambda args: exp.DateDiff(
 511            this=seq_get(args, 2),
 512            expression=seq_get(args, 1),
 513            unit=map_date_part(seq_get(args, 0)),
 514            date_part_boundary=True,
 515        ),
 516        "DAYNAME": lambda args: exp.Dayname(this=seq_get(args, 0), abbreviated=True),
 517        "DAYOFWEEKISO": exp.DayOfWeekIso.from_arg_list,
 518        "DIV0": _build_if_from_div0,
 519        "DIV0NULL": _build_if_from_div0null,
 520        "EDITDISTANCE": lambda args: exp.Levenshtein(
 521            this=seq_get(args, 0), expression=seq_get(args, 1), max_dist=seq_get(args, 2)
 522        ),
 523        "FLATTEN": exp.Explode.from_arg_list,
 524        "GENERATOR": _build_generator,
 525        "GET": exp.GetExtract.from_arg_list,
 526        "GETDATE": exp.CurrentTimestamp.from_arg_list,
 527        "GET_PATH": lambda args, dialect: exp.JSONExtract(
 528            this=seq_get(args, 0),
 529            expression=dialect.to_json_path(seq_get(args, 1)),
 530            requires_json=True,
 531        ),
 532        "GREATEST_IGNORE_NULLS": lambda args: exp.Greatest(
 533            this=seq_get(args, 0), expressions=args[1:], ignore_nulls=True
 534        ),
 535        "LEAST_IGNORE_NULLS": lambda args: exp.Least(
 536            this=seq_get(args, 0), expressions=args[1:], ignore_nulls=True
 537        ),
 538        "LEFT": lambda args: exp.Left(
 539            this=seq_get(args, 0), expression=seq_get(args, 1), negative_length_returns_empty=True
 540        ),
 541        "HEX_DECODE_BINARY": exp.Unhex.from_arg_list,
 542        "HEX_ENCODE": exp.Hex.from_arg_list,
 543        "IFF": exp.If.from_arg_list,
 544        "JAROWINKLER_SIMILARITY": lambda args: exp.JarowinklerSimilarity(
 545            this=seq_get(args, 0),
 546            expression=seq_get(args, 1),
 547            case_insensitive=True,
 548            integer_scale=True,
 549        ),
 550        "MD5_HEX": exp.MD5.from_arg_list,
 551        "MD5_BINARY": exp.MD5Digest.from_arg_list,
 552        "MD5_NUMBER_LOWER64": exp.MD5NumberLower64.from_arg_list,
 553        "MD5_NUMBER_UPPER64": exp.MD5NumberUpper64.from_arg_list,
 554        "MONTHNAME": lambda args: exp.Monthname(this=seq_get(args, 0), abbreviated=True),
 555        "LAST_DAY": lambda args: exp.LastDay(
 556            this=seq_get(args, 0), unit=map_date_part(seq_get(args, 1))
 557        ),
 558        "LEN": lambda args: exp.Length(this=seq_get(args, 0), binary=True),
 559        "LENGTH": lambda args: exp.Length(this=seq_get(args, 0), binary=True),
 560        "LOCALTIMESTAMP": exp.CurrentTimestamp.from_arg_list,
 561        "NULLIFZERO": _build_if_from_nullifzero,
 562        "OBJECT_CONSTRUCT": lambda args: build_object_construct(args),
 563        "OBJECT_KEYS": exp.JSONKeys.from_arg_list,
 564        "OCTET_LENGTH": exp.ByteLength.from_arg_list,
 565        "PARSE_URL": lambda args: exp.ParseUrl(this=seq_get(args, 0), permissive=seq_get(args, 1)),
 566        "REGEXP_EXTRACT_ALL": _build_regexp_extract(exp.RegexpExtractAll),
 567        "REGEXP_LIKE": _build_regexp_like,
 568        "REGEXP_REPLACE": _build_regexp_replace,
 569        "REGEXP_SUBSTR": _build_regexp_extract(exp.RegexpExtract),
 570        "REGEXP_SUBSTR_ALL": _build_regexp_extract(exp.RegexpExtractAll),
 571        "RANDOM": lambda args: exp.Rand(
 572            this=seq_get(args, 0),
 573            lower=exp.Literal.number(-9223372036854775808.0),  # -2^63 as float to avoid overflow
 574            upper=exp.Literal.number(9223372036854775807.0),  # 2^63-1 as float
 575        ),
 576        "REPLACE": build_replace_with_optional_replacement,
 577        "RIGHT": lambda args: exp.Right(
 578            this=seq_get(args, 0), expression=seq_get(args, 1), negative_length_returns_empty=True
 579        ),
 580        "RLIKE": _build_regexp_like,
 581        "ROUND": _build_round,
 582        "SHA1_BINARY": exp.SHA1Digest.from_arg_list,
 583        "SHA1_HEX": exp.SHA.from_arg_list,
 584        "SHA2_BINARY": exp.SHA2Digest.from_arg_list,
 585        "SHA2_HEX": exp.SHA2.from_arg_list,
 586        "SPLIT": lambda args: exp.Split(
 587            this=seq_get(args, 0),
 588            expression=seq_get(args, 1),
 589            null_returns_null=True,
 590            empty_delimiter_returns_whole=True,
 591        ),
 592        "SQUARE": lambda args: exp.Pow(this=seq_get(args, 0), expression=exp.Literal.number(2)),
 593        "STDDEV_SAMP": exp.Stddev.from_arg_list,
 594        "SYSDATE": lambda args: exp.CurrentTimestamp(this=seq_get(args, 0), sysdate=True),
 595        "TABLE": lambda args: exp.TableFromRows(this=seq_get(args, 0)),
 596        "TIMEADD": lambda args: exp.TimeAdd(
 597            this=seq_get(args, 2),
 598            expression=seq_get(args, 1),
 599            unit=map_date_part(seq_get(args, 0)),
 600        ),
 601        "TIMEDIFF": lambda args: exp.DateDiff(
 602            this=seq_get(args, 2),
 603            expression=seq_get(args, 1),
 604            unit=map_date_part(seq_get(args, 0)),
 605            date_part_boundary=True,
 606        ),
 607        "TIME_FROM_PARTS": lambda args: exp.TimeFromParts(
 608            hour=seq_get(args, 0),
 609            min=seq_get(args, 1),
 610            sec=seq_get(args, 2),
 611            nano=seq_get(args, 3),
 612            overflow=True,
 613        ),
 614        "TIMESTAMPADD": lambda args: exp.DateAdd(
 615            this=seq_get(args, 2),
 616            expression=seq_get(args, 1),
 617            unit=map_date_part(seq_get(args, 0)),
 618        ),
 619        "TIMESTAMPDIFF": lambda args: exp.DateDiff(
 620            this=seq_get(args, 2),
 621            expression=seq_get(args, 1),
 622            unit=map_date_part(seq_get(args, 0)),
 623            date_part_boundary=True,
 624        ),
 625        "TIMESTAMPFROMPARTS": _build_timestamp_from_parts,
 626        "TIMESTAMP_FROM_PARTS": _build_timestamp_from_parts,
 627        "TIMESTAMPNTZFROMPARTS": _build_timestamp_from_parts,
 628        "TIMESTAMP_NTZ_FROM_PARTS": _build_timestamp_from_parts,
 629        "TRUNC": lambda args, dialect: build_trunc(
 630            args, dialect, date_trunc_requires_part=False, fractions_supported=True
 631        ),
 632        "TRUNCATE": lambda args, dialect: build_trunc(
 633            args, dialect, date_trunc_requires_part=False, fractions_supported=True
 634        ),
 635        "TRY_DECRYPT": lambda args: exp.Decrypt(
 636            this=seq_get(args, 0),
 637            passphrase=seq_get(args, 1),
 638            aad=seq_get(args, 2),
 639            encryption_method=seq_get(args, 3),
 640            safe=True,
 641        ),
 642        "TRY_DECRYPT_RAW": lambda args: exp.DecryptRaw(
 643            this=seq_get(args, 0),
 644            key=seq_get(args, 1),
 645            iv=seq_get(args, 2),
 646            aad=seq_get(args, 3),
 647            encryption_method=seq_get(args, 4),
 648            aead=seq_get(args, 5),
 649            safe=True,
 650        ),
 651        "TRY_PARSE_JSON": lambda args: exp.ParseJSON(this=seq_get(args, 0), safe=True),
 652        "TRY_TO_BINARY": lambda args: exp.ToBinary(
 653            this=seq_get(args, 0), format=seq_get(args, 1), safe=True
 654        ),
 655        "TRY_TO_BOOLEAN": lambda args: exp.ToBoolean(this=seq_get(args, 0), safe=True),
 656        "TRY_TO_DATE": _build_datetime("TRY_TO_DATE", exp.DType.DATE, safe=True),
 657        **dict.fromkeys(
 658            ("TRY_TO_DECIMAL", "TRY_TO_NUMBER", "TRY_TO_NUMERIC"),
 659            lambda args: _build_to_number(args, safe=True),
 660        ),
 661        "TRY_TO_DOUBLE": lambda args: exp.ToDouble(
 662            this=seq_get(args, 0), format=seq_get(args, 1), safe=True
 663        ),
 664        "TRY_TO_FILE": lambda args: exp.ToFile(
 665            this=seq_get(args, 0), path=seq_get(args, 1), safe=True
 666        ),
 667        "TRY_TO_TIME": _build_datetime("TRY_TO_TIME", exp.DType.TIME, safe=True),
 668        "TRY_TO_TIMESTAMP": _build_datetime("TRY_TO_TIMESTAMP", exp.DType.TIMESTAMP, safe=True),
 669        "TRY_TO_TIMESTAMP_LTZ": _build_datetime(
 670            "TRY_TO_TIMESTAMP_LTZ", exp.DType.TIMESTAMPLTZ, safe=True
 671        ),
 672        "TRY_TO_TIMESTAMP_NTZ": _build_datetime(
 673            "TRY_TO_TIMESTAMP_NTZ", exp.DType.TIMESTAMPNTZ, safe=True
 674        ),
 675        "TRY_TO_TIMESTAMP_TZ": _build_datetime(
 676            "TRY_TO_TIMESTAMP_TZ", exp.DType.TIMESTAMPTZ, safe=True
 677        ),
 678        "TO_CHAR": build_timetostr_or_tochar,
 679        "TO_DATE": _build_datetime("TO_DATE", exp.DType.DATE),
 680        **dict.fromkeys(
 681            ("TO_DECIMAL", "TO_NUMBER", "TO_NUMERIC"),
 682            lambda args: _build_to_number(args),
 683        ),
 684        "TO_TIME": _build_datetime("TO_TIME", exp.DType.TIME),
 685        "TO_TIMESTAMP": _build_datetime("TO_TIMESTAMP", exp.DType.TIMESTAMP),
 686        "TO_TIMESTAMP_LTZ": _build_datetime("TO_TIMESTAMP_LTZ", exp.DType.TIMESTAMPLTZ),
 687        "TO_TIMESTAMP_NTZ": _build_datetime("TO_TIMESTAMP_NTZ", exp.DType.TIMESTAMPNTZ),
 688        "TO_TIMESTAMP_TZ": _build_datetime("TO_TIMESTAMP_TZ", exp.DType.TIMESTAMPTZ),
 689        "TO_GEOGRAPHY": lambda args: (
 690            exp.cast(args[0], exp.DType.GEOGRAPHY)
 691            if len(args) == 1
 692            else exp.Anonymous(this="TO_GEOGRAPHY", expressions=args)
 693        ),
 694        "TO_GEOMETRY": lambda args: (
 695            exp.cast(args[0], exp.DType.GEOMETRY)
 696            if len(args) == 1
 697            else exp.Anonymous(this="TO_GEOMETRY", expressions=args)
 698        ),
 699        "TO_VARCHAR": build_timetostr_or_tochar,
 700        "TO_JSON": exp.JSONFormat.from_arg_list,
 701        "VECTOR_COSINE_SIMILARITY": exp.CosineDistance.from_arg_list,
 702        "VECTOR_INNER_PRODUCT": exp.DotProduct.from_arg_list,
 703        "VECTOR_L1_DISTANCE": exp.ManhattanDistance.from_arg_list,
 704        "VECTOR_L2_DISTANCE": exp.EuclideanDistance.from_arg_list,
 705        "ZEROIFNULL": _build_if_from_zeroifnull,
 706        "LIKE": build_like(exp.Like),
 707        "ILIKE": build_like(exp.ILike),
 708        "SEARCH": _build_search,
 709        "SKEW": exp.Skewness.from_arg_list,
 710        "SPLIT_PART": lambda args: exp.SplitPart(
 711            this=seq_get(args, 0),
 712            delimiter=seq_get(args, 1),
 713            part_index=seq_get(args, 2),
 714            part_index_zero_as_one=True,
 715            empty_delimiter_returns_whole=True,
 716        ),
 717        "STRTOK": lambda args: exp.Strtok(
 718            this=seq_get(args, 0),
 719            delimiter=seq_get(args, 1) or exp.Literal.string(" "),
 720            part_index=seq_get(args, 2) or exp.Literal.number("1"),
 721        ),
 722        "STRTOK_TO_ARRAY": lambda args: exp.StrtokToArray(
 723            this=seq_get(args, 0),
 724            expression=seq_get(args, 1) or exp.Literal.string(" "),
 725        ),
 726        "SYSTIMESTAMP": exp.CurrentTimestamp.from_arg_list,
 727        "IDENTIFIER": exp.DynamicIdentifier.from_arg_list,
 728        "UNICODE": lambda args: exp.Unicode(this=seq_get(args, 0), empty_is_zero=True),
 729        "WEEKISO": exp.WeekOfYear.from_arg_list,
 730        "WEEKOFYEAR": exp.Week.from_arg_list,
 731    }
 732    FUNCTIONS = {k: v for k, v in FUNCTIONS.items() if k != "PREDICT"}
 733
 734    FUNCTION_PARSERS = {
 735        **parser.Parser.FUNCTION_PARSERS,
 736        "DATE_PART": lambda self: self._parse_date_part(),
 737        "DIRECTORY": lambda self: self._parse_directory(),
 738        "OBJECT_CONSTRUCT_KEEP_NULL": lambda self: self._parse_json_object(),
 739        "LISTAGG": lambda self: self._parse_string_agg(),
 740        "SEMANTIC_VIEW": lambda self: self._parse_semantic_view(),
 741        "SUBSTR": lambda self: self._parse_substring(),
 742    }
 743    FUNCTION_PARSERS = {k: v for k, v in FUNCTION_PARSERS.items() if k != "TRIM"}
 744
 745    TIMESTAMPS = parser.Parser.TIMESTAMPS - {TokenType.TIME}
 746
 747    ALTER_PARSERS = {
 748        **parser.Parser.ALTER_PARSERS,
 749        "MODIFY": lambda self: self._parse_alter_table_alter(),
 750        "SESSION": lambda self: self._parse_alter_session(),
 751        "UNSET": lambda self: self.expression(
 752            exp.Set(
 753                tag=self._match_text_seq("TAG"),
 754                expressions=self._parse_csv(self._parse_id_var),
 755                unset=True,
 756            )
 757        ),
 758    }
 759
 760    STATEMENT_PARSERS = {
 761        **parser.Parser.STATEMENT_PARSERS,
 762        TokenType.GET: lambda self: self._parse_get(),
 763        TokenType.PUT: lambda self: self._parse_put(),
 764        TokenType.SHOW: lambda self: self._parse_show(),
 765        TokenType.UNDROP: lambda self: self._parse_undrop(),
 766    }
 767
 768    PROPERTY_PARSERS = {
 769        **parser.Parser.PROPERTY_PARSERS,
 770        "CREDENTIALS": lambda self: self._parse_credentials_property(),
 771        "FILE_FORMAT": lambda self: self._parse_file_format_property(),
 772        "LOCATION": lambda self: self._parse_location_property(),
 773        "ROW": lambda self: (
 774            self._parse_row_access_policy()
 775            if self._match_text_seq("ACCESS", "POLICY")
 776            else self._parse_row()
 777        ),
 778        "TAG": lambda self: self._parse_tag(),
 779        "USING": lambda self: (
 780            self._match_text_seq("TEMPLATE")
 781            and self.expression(exp.UsingTemplateProperty(this=self._parse_statement()))
 782        ),
 783    }
 784
 785    DESCRIBE_QUALIFIER_PARSERS: t.ClassVar[dict[str, t.Callable]] = {
 786        "API": lambda self: self.expression(exp.ApiProperty()),
 787        "APPLICATION": lambda self: self.expression(exp.ApplicationProperty()),
 788        "CATALOG": lambda self: self.expression(exp.CatalogProperty()),
 789        "COMPUTE": lambda self: self.expression(exp.ComputeProperty()),
 790        "DATABASE": lambda self: (
 791            self.expression(exp.DatabaseProperty())
 792            if self._curr and self._curr.text.upper() == "ROLE"
 793            else None
 794        ),
 795        "DYNAMIC": lambda self: self.expression(exp.DynamicProperty()),
 796        "EXTERNAL": lambda self: self.expression(exp.ExternalProperty()),
 797        "HYBRID": lambda self: self.expression(exp.HybridProperty()),
 798        "ICEBERG": lambda self: self.expression(exp.IcebergProperty()),
 799        "MASKING": lambda self: self.expression(exp.MaskingProperty()),
 800        "MATERIALIZED": lambda self: self.expression(exp.MaterializedProperty()),
 801        "NETWORK": lambda self: self.expression(exp.NetworkProperty()),
 802        "ROW": lambda self: (
 803            self.expression(exp.RowAccessProperty()) if self._match_text_seq("ACCESS") else None
 804        ),
 805        "SECURITY": lambda self: (
 806            self.expression(exp.SecurityIntegrationProperty())
 807            if self._curr and self._curr.text.upper() == "INTEGRATION"
 808            else None
 809        ),
 810    }
 811
 812    TYPE_CONVERTERS = {
 813        # https://docs.snowflake.com/en/sql-reference/data-types-numeric#number
 814        exp.DType.DECIMAL: build_default_decimal_type(precision=38, scale=0),
 815    }
 816
 817    SHOW_PARSERS = {
 818        "DATABASES": _show_parser("DATABASES"),
 819        "SCHEMAS": _show_parser("SCHEMAS"),
 820        "OBJECTS": _show_parser("OBJECTS"),
 821        "TABLES": _show_parser("TABLES"),
 822        "VIEWS": _show_parser("VIEWS"),
 823        "PRIMARY KEYS": _show_parser("PRIMARY KEYS"),
 824        "IMPORTED KEYS": _show_parser("IMPORTED KEYS"),
 825        "UNIQUE KEYS": _show_parser("UNIQUE KEYS"),
 826        "SEQUENCES": _show_parser("SEQUENCES"),
 827        "STAGES": _show_parser("STAGES"),
 828        "COLUMNS": _show_parser("COLUMNS"),
 829        "USERS": _show_parser("USERS"),
 830        "FILE FORMATS": _show_parser("FILE FORMATS"),
 831        "FUNCTIONS": _show_parser("FUNCTIONS"),
 832        "PROCEDURES": _show_parser("PROCEDURES"),
 833        "WAREHOUSES": _show_parser("WAREHOUSES"),
 834        "ICEBERG TABLES": _show_parser("TABLES", iceberg=True),
 835        "TERSE ICEBERG TABLES": _show_parser("TABLES", terse=True, iceberg=True),
 836        "TERSE DATABASES": _show_parser("DATABASES", terse=True),
 837        "TERSE SCHEMAS": _show_parser("SCHEMAS", terse=True),
 838        "TERSE OBJECTS": _show_parser("OBJECTS", terse=True),
 839        "TERSE TABLES": _show_parser("TABLES", terse=True),
 840        "TERSE VIEWS": _show_parser("VIEWS", terse=True),
 841        "TERSE SEQUENCES": _show_parser("SEQUENCES", terse=True),
 842        "TERSE USERS": _show_parser("USERS", terse=True),
 843        # TERSE has no semantic effect for KEYS, so we do not set the terse AST arg
 844        "TERSE PRIMARY KEYS": _show_parser("PRIMARY KEYS"),
 845        "TERSE IMPORTED KEYS": _show_parser("IMPORTED KEYS"),
 846        "TERSE UNIQUE KEYS": _show_parser("UNIQUE KEYS"),
 847    }
 848
 849    SHOW_TRIE = new_trie(key.split(" ") for key in SHOW_PARSERS)
 850
 851    CONSTRAINT_PARSERS = {
 852        **parser.Parser.CONSTRAINT_PARSERS,
 853        "WITH": lambda self: self._parse_with_constraint(),
 854        "MASKING": lambda self: self._parse_with_constraint(),
 855        "PROJECTION": lambda self: self._parse_with_constraint(),
 856        "TAG": lambda self: self._parse_with_constraint(),
 857    }
 858
 859    STAGED_FILE_SINGLE_TOKENS = {
 860        TokenType.DOT,
 861        TokenType.MOD,
 862        TokenType.SLASH,
 863    }
 864
 865    FLATTEN_COLUMNS = ["SEQ", "KEY", "PATH", "INDEX", "VALUE", "THIS"]
 866
 867    SCHEMA_KINDS = {"OBJECTS", "TABLES", "VIEWS", "SEQUENCES", "UNIQUE KEYS", "IMPORTED KEYS"}
 868
 869    NON_TABLE_CREATABLES = {"STORAGE INTEGRATION", "TAG", "WAREHOUSE", "STREAMLIT"}
 870
 871    UNDROP_OBJECTS: t.ClassVar[parser.OPTIONS_TYPE] = {
 872        **dict.fromkeys(
 873            (
 874                "ACCOUNT",
 875                "DATABASE",
 876                "NOTEBOOK",
 877                "SCHEMA",
 878                "SNAPSHOT",
 879                "STREAMLIT",
 880                "TABLE",
 881                "TAG",
 882                "TYPE",
 883            ),
 884            tuple(),
 885        ),
 886        "DYNAMIC": ("TABLE",),
 887        "EXTERNAL": ("VOLUME",),
 888        "ICEBERG": ("TABLE",),
 889    }
 890
 891    CREATABLES = {
 892        *parser.Parser.CREATABLES,
 893        TokenType.INTEGRATION,
 894        TokenType.PACKAGE,
 895        TokenType.POLICY,
 896        TokenType.POOL,
 897        TokenType.ROLE,
 898        TokenType.RULE,
 899        TokenType.VOLUME,
 900    }
 901
 902    LAMBDAS = {
 903        **parser.Parser.LAMBDAS,
 904        TokenType.ARROW: lambda self, expressions: self.expression(
 905            exp.Lambda(
 906                this=self._replace_lambda(
 907                    self._parse_assignment(),
 908                    expressions,
 909                ),
 910                expressions=[e.this if isinstance(e, exp.Cast) else e for e in expressions],
 911            )
 912        ),
 913    }
 914
 915    COLUMN_OPERATORS = {
 916        **parser.Parser.COLUMN_OPERATORS,
 917        TokenType.EXCLAMATION: lambda self, this, attr: self.expression(
 918            exp.ModelAttribute(this=this, expression=attr)
 919        ),
 920    }
 921
 922    def _parse_directory(self) -> exp.DirectoryStage:
 923        table = self._parse_table_parts()
 924        this = table.this if isinstance(table, exp.Table) else table
 925        return self.expression(exp.DirectoryStage(this=this))
 926
 927    def _parse_describe(self) -> exp.Describe:
 928        index = self._index
 929
 930        if self._match_texts(self.DESCRIBE_QUALIFIER_PARSERS):
 931            qualifier = self.DESCRIBE_QUALIFIER_PARSERS[self._prev.text.upper()](self)
 932
 933            if qualifier:
 934                kind = self._match_set(self.CREATABLES) and self._prev.text.upper()
 935
 936                if kind:
 937                    this = self._parse_table(schema=True)
 938                    properties = self.expression(exp.Properties(expressions=[qualifier]))
 939                    post_props = self._parse_properties()
 940                    expressions = post_props.expressions if post_props else None
 941                    return self.expression(
 942                        exp.Describe(
 943                            this=this,
 944                            kind=kind,
 945                            properties=properties,
 946                            expressions=expressions,
 947                        )
 948                    )
 949
 950        self._retreat(index)
 951        return super()._parse_describe()
 952
 953    def _parse_use(self) -> exp.Use:
 954        if self._match_text_seq("SECONDARY", "ROLES"):
 955            this = self._match_texts(("ALL", "NONE")) and exp.var(self._prev.text.upper())
 956            roles = None if this else self._parse_csv(lambda: self._parse_table(schema=False))
 957            return self.expression(exp.Use(kind="SECONDARY ROLES", this=this, expressions=roles))
 958
 959        return super()._parse_use()
 960
 961    def _negate_range(self, this: exp.Expr | None = None) -> exp.Expr | None:
 962        if not this:
 963            return this
 964
 965        query = this.args.get("query")
 966        if isinstance(this, exp.In) and isinstance(query, exp.Query):
 967            # Snowflake treats `value NOT IN (subquery)` as `VALUE <> ALL (subquery)`, so
 968            # we do this conversion here to avoid parsing it into `NOT value IN (subquery)`
 969            # which can produce different results (most likely a SnowFlake bug).
 970            #
 971            # https://docs.snowflake.com/en/sql-reference/functions/in
 972            # Context: https://github.com/tobymao/sqlglot/issues/3890
 973            return self.expression(exp.NEQ(this=this.this, expression=exp.All(this=query.unnest())))
 974
 975        return self.expression(exp.Not(this=this))
 976
 977    def _parse_tag(self) -> exp.Tags:
 978        return self.expression(exp.Tags(expressions=self._parse_wrapped_csv(self._parse_property)))
 979
 980    def _parse_property_before(self) -> exp.Expr | list[exp.Expr] | None:
 981        prop = super()._parse_property_before()
 982        if prop:
 983            return prop
 984
 985        if not self._next or self._next.token_type != TokenType.EQ:
 986            return None
 987
 988        return self._parse_sequence_properties() or self._parse_key_value_property(
 989            self._parse_primary_or_var
 990        )
 991
 992    def _parse_with_constraint(self) -> exp.Expr | None:
 993        if self._prev.token_type != TokenType.WITH:
 994            self._retreat(self._index - 1)
 995
 996        if self._match_text_seq("MASKING", "POLICY"):
 997            policy = self._parse_column()
 998            return self.expression(
 999                exp.MaskingPolicyColumnConstraint(
1000                    this=policy.to_dot() if isinstance(policy, exp.Column) else policy,
1001                    expressions=self._match(TokenType.USING)
1002                    and self._parse_wrapped_csv(self._parse_id_var),
1003                )
1004            )
1005        if self._match_text_seq("PROJECTION", "POLICY"):
1006            policy = self._parse_column()
1007            return self.expression(
1008                exp.ProjectionPolicyColumnConstraint(
1009                    this=policy.to_dot() if isinstance(policy, exp.Column) else policy
1010                )
1011            )
1012        if self._match(TokenType.TAG):
1013            return self._parse_tag()
1014
1015        return None
1016
1017    def _parse_with_property(self) -> exp.Expr | None | list[exp.Expr]:
1018        if self._match(TokenType.TAG):
1019            return self._parse_tag()
1020
1021        if self._match_text_seq("ROW", "ACCESS", "POLICY"):
1022            return self._parse_row_access_policy()
1023
1024        return super()._parse_with_property()
1025
1026    def _parse_row_access_policy(self) -> exp.RowAccessProperty:
1027        # GET_DDL outputs #unknown_policy when the user lacks privileges to see the policy name
1028        if self._match(TokenType.HASH):
1029            policy: exp.Expr | None = self._parse_var(any_token=True)
1030            if policy:
1031                policy = exp.Var(this=f"#{policy.name}")
1032            expressions = None
1033        else:
1034            policy = self._parse_column()
1035            if isinstance(policy, exp.Column):
1036                policy = policy.to_dot()
1037            if not self._match(TokenType.ON):
1038                self.raise_error("Expected ON after ROW ACCESS POLICY name")
1039            expressions = self._parse_wrapped_csv(self._parse_id_var)
1040
1041        return self.expression(exp.RowAccessProperty(this=policy, expressions=expressions))
1042
1043    def _parse_create(self) -> exp.Create | exp.Command:
1044        expression = super()._parse_create()
1045        if isinstance(expression, exp.Create) and expression.kind in self.NON_TABLE_CREATABLES:
1046            # Replace the Table node with the enclosed Identifier
1047            expression.this.replace(expression.this.this)
1048
1049        return expression
1050
1051    # https://docs.snowflake.com/en/sql-reference/functions/date_part.html
1052    # https://docs.snowflake.com/en/sql-reference/functions-date-time.html#label-supported-date-time-parts
1053    def _parse_date_part(self) -> exp.Expr | None:
1054        this = self._parse_var() or self._parse_type()
1055
1056        if not this:
1057            return None
1058
1059        # Handle both syntaxes: DATE_PART(part, expr) and DATE_PART(part FROM expr)
1060        expression = self._match_set((TokenType.FROM, TokenType.COMMA)) and self._parse_bitwise()
1061        return self.expression(
1062            exp.Extract(this=map_date_part(this, self.dialect), expression=expression)
1063        )
1064
1065    def _parse_bracket_key_value(self, is_map: bool = False) -> exp.Expr | None:
1066        if is_map:
1067            # Keys are strings in Snowflake's objects, see also:
1068            # - https://docs.snowflake.com/en/sql-reference/data-types-semistructured
1069            # - https://docs.snowflake.com/en/sql-reference/functions/object_construct
1070            return self._parse_slice(self._parse_string()) or self._parse_assignment()
1071
1072        return self._parse_slice(self._parse_alias(self._parse_assignment(), explicit=True))
1073
1074    def _parse_lateral(self) -> exp.Lateral | None:
1075        lateral = super()._parse_lateral()
1076        if not lateral:
1077            return lateral
1078
1079        if isinstance(lateral.this, exp.Explode):
1080            table_alias = lateral.args.get("alias")
1081            columns = [exp.to_identifier(col) for col in self.FLATTEN_COLUMNS]
1082            if table_alias and not table_alias.args.get("columns"):
1083                table_alias.set("columns", columns)
1084            elif not table_alias:
1085                exp.alias_(lateral, "_flattened", table=columns, copy=False)
1086
1087        return lateral
1088
1089    def _parse_table_parts(
1090        self,
1091        schema: bool = False,
1092        is_db_reference: bool = False,
1093        wildcard: bool = False,
1094        fast: bool = False,
1095    ) -> exp.Table | exp.Dot | None:
1096        # https://docs.snowflake.com/en/user-guide/querying-stage
1097        if self._match(TokenType.STRING, advance=False):
1098            table = self._parse_string()
1099        elif self._match_text_seq("@", advance=False):
1100            table = self._parse_location_path()
1101        else:
1102            table = None
1103
1104        if table:
1105            file_format = None
1106            pattern = None
1107
1108            wrapped = self._match(TokenType.L_PAREN)
1109            while self._curr and wrapped and not self._match(TokenType.R_PAREN):
1110                if self._match_text_seq("FILE_FORMAT", "=>"):
1111                    file_format = self._parse_string() or super()._parse_table_parts(
1112                        is_db_reference=is_db_reference
1113                    )
1114                elif self._match_text_seq("PATTERN", "=>"):
1115                    pattern = self._parse_string()
1116                else:
1117                    break
1118
1119                self._match(TokenType.COMMA)
1120
1121            table = self.expression(exp.Table(this=table, format=file_format, pattern=pattern))
1122        else:
1123            table = super()._parse_table_parts(
1124                schema=schema,
1125                is_db_reference=is_db_reference,
1126                fast=fast,
1127            )
1128
1129        return table
1130
1131    def _parse_table(
1132        self,
1133        schema: bool = False,
1134        joins: bool = False,
1135        alias_tokens: Collection[TokenType] | None = None,
1136        parse_bracket: bool = False,
1137        is_db_reference: bool = False,
1138        parse_partition: bool = False,
1139        consume_pipe: bool = False,
1140    ) -> exp.Expr | None:
1141        table = super()._parse_table(
1142            schema=schema,
1143            joins=joins,
1144            alias_tokens=alias_tokens,
1145            parse_bracket=parse_bracket,
1146            is_db_reference=is_db_reference,
1147            parse_partition=parse_partition,
1148        )
1149        if isinstance(table, exp.Table) and isinstance(table.this, exp.TableFromRows):
1150            table_from_rows = table.this
1151            for arg in exp.TableFromRows.arg_types:
1152                if arg != "this":
1153                    table_from_rows.set(arg, table.args.get(arg))
1154
1155            table = table_from_rows
1156
1157        return table
1158
1159    def _parse_function_call(
1160        self,
1161        functions: dict[str, t.Callable] | None = None,
1162        anonymous: bool = False,
1163        optional_parens: bool = True,
1164        any_token: bool = False,
1165    ) -> exp.Expr | None:
1166        this = super()._parse_function_call(
1167            functions=functions,
1168            anonymous=anonymous,
1169            optional_parens=optional_parens,
1170            any_token=any_token,
1171        )
1172
1173        # Snowflake can invoke a function whose name is dynamically resolved, e.g.
1174        # `IDENTIFIER('my_func')(1, 2)`. The trailing argument list is the call's arguments.
1175        #
1176        # https://docs.snowflake.com/en/sql-reference/identifier-literal
1177        if isinstance(this, exp.DynamicIdentifier) and self._match(
1178            TokenType.L_PAREN, advance=False
1179        ):
1180            this.set("expressions", self._parse_wrapped_csv(self._parse_lambda))
1181
1182        return this
1183
1184    def _parse_id_var(
1185        self,
1186        any_token: bool = True,
1187        tokens: Collection[TokenType] | None = None,
1188    ) -> exp.Expr | None:
1189        if self._match_text_seq("IDENTIFIER", "("):
1190            identifier = (
1191                super()._parse_id_var(any_token=any_token, tokens=tokens) or self._parse_string()
1192            )
1193            self._match_r_paren()
1194            return self.expression(exp.DynamicIdentifier(this=identifier))
1195
1196        return super()._parse_id_var(any_token=any_token, tokens=tokens)
1197
1198    def _parse_show_snowflake(
1199        self, this: str, terse: bool = False, iceberg: bool = False
1200    ) -> exp.Show:
1201        scope = None
1202        scope_kind = None
1203
1204        history = self._match_text_seq("HISTORY")
1205
1206        like = self._parse_string() if self._match(TokenType.LIKE) else None
1207
1208        if self._match(TokenType.IN):
1209            if self._match_text_seq("ACCOUNT"):
1210                scope_kind = "ACCOUNT"
1211            elif self._match_text_seq("CLASS"):
1212                scope_kind = "CLASS"
1213                scope = self._parse_table_parts()
1214            elif self._match_text_seq("APPLICATION"):
1215                scope_kind = "APPLICATION"
1216                if self._match_text_seq("PACKAGE"):
1217                    scope_kind += " PACKAGE"
1218                scope = self._parse_table_parts()
1219            elif self._match_set(self.DB_CREATABLES):
1220                scope_kind = self._prev.text.upper()
1221                if self._curr:
1222                    scope = self._parse_table_parts()
1223            elif self._curr:
1224                scope_kind = "SCHEMA" if this in self.SCHEMA_KINDS else "TABLE"
1225                scope = self._parse_table_parts()
1226
1227        return self.expression(
1228            exp.Show(
1229                terse=terse,
1230                iceberg=iceberg,
1231                this=this,
1232                history=history,
1233                like=like,
1234                scope=scope,
1235                scope_kind=scope_kind,
1236                starts_with=self._match_text_seq("STARTS", "WITH") and self._parse_string(),
1237                limit=self._parse_limit(),
1238                from_=self._parse_string() if self._match(TokenType.FROM) else None,
1239                privileges=self._match_text_seq("WITH", "PRIVILEGES")
1240                and self._parse_csv(lambda: self._parse_var(any_token=True, upper=True)),
1241            )
1242        )
1243
1244    def _parse_undrop(self) -> exp.Undrop | exp.Command:
1245        start = self._prev
1246        kind = self._parse_var_from_options(self.UNDROP_OBJECTS, raise_unmatched=False)
1247        if not kind:
1248            return self._parse_as_command(start)
1249
1250        this = self._parse_table_parts(
1251            is_db_reference=kind.name in ("ACCOUNT", "DATABASE", "SCHEMA")
1252        )
1253        rename = self._parse_table_parts() if self._match_text_seq("RENAME", "TO") else None
1254        return self.expression(exp.Undrop(this=this, kind=kind.name, rename=rename))
1255
1256    def _parse_put(self) -> exp.Put | exp.Command:
1257        if self._curr.token_type != TokenType.STRING:
1258            return self._parse_as_command(self._prev)
1259
1260        return self.expression(
1261            exp.Put(
1262                this=self._parse_string(),
1263                target=self._parse_location_path(),
1264                properties=self._parse_properties(),
1265            )
1266        )
1267
1268    def _parse_get(self) -> exp.Expr | None:
1269        start = self._prev
1270
1271        # If we detect GET( then we need to parse a function, not a statement
1272        if self._match(TokenType.L_PAREN):
1273            self._retreat(self._index - 2)
1274            return self._parse_expression()
1275
1276        target = self._parse_location_path()
1277
1278        # Parse as command if unquoted file path
1279        if self._curr.token_type == TokenType.URI_START:
1280            return self._parse_as_command(start)
1281
1282        return self.expression(
1283            exp.Get(this=self._parse_string(), target=target, properties=self._parse_properties())
1284        )
1285
1286    def _parse_location_property(self) -> exp.LocationProperty:
1287        self._match(TokenType.EQ)
1288        return self.expression(exp.LocationProperty(this=self._parse_location_path()))
1289
1290    def _parse_file_location(self) -> exp.Expr | None:
1291        # Parse either a subquery or a staged file
1292        return (
1293            self._parse_select(table=True, parse_subquery_alias=False)
1294            if self._match(TokenType.L_PAREN, advance=False)
1295            else self._parse_table_parts()
1296        )
1297
1298    def _parse_location_path(self) -> exp.Var:
1299        start = self._curr
1300        self._advance_any(ignore_reserved=True)
1301
1302        # We avoid consuming a comma token because external tables like @foo and @bar
1303        # can be joined in a query with a comma separator, as well as closing paren
1304        # in case of subqueries
1305        while self._is_connected() and not self._match_set(
1306            (TokenType.COMMA, TokenType.L_PAREN, TokenType.R_PAREN), advance=False
1307        ):
1308            self._advance_any(ignore_reserved=True)
1309
1310        return exp.var(self._find_sql(start, self._prev))
1311
1312    def _parse_lambda_arg(self) -> exp.Expr | None:
1313        this = super()._parse_lambda_arg()
1314
1315        if not this:
1316            return this
1317
1318        typ = self._parse_types()
1319
1320        if typ:
1321            return self.expression(exp.Cast(this=this, to=typ))
1322
1323        return this
1324
1325    def _parse_foreign_key(self) -> exp.ForeignKey:
1326        # inlineFK, the REFERENCES columns are implied
1327        if self._match(TokenType.REFERENCES, advance=False):
1328            return self.expression(exp.ForeignKey())
1329
1330        # outoflineFK, explicitly names the columns
1331        return super()._parse_foreign_key()
1332
1333    def _parse_file_format_property(self) -> exp.FileFormatProperty:
1334        self._match(TokenType.EQ)
1335        if self._match(TokenType.L_PAREN, advance=False):
1336            expressions = self._parse_wrapped_options()
1337        else:
1338            expressions = [self._parse_format_name()]
1339
1340        return self.expression(exp.FileFormatProperty(expressions=expressions))
1341
1342    def _parse_credentials_property(self) -> exp.CredentialsProperty:
1343        return self.expression(exp.CredentialsProperty(expressions=self._parse_wrapped_options()))
1344
1345    def _parse_semantic_view(self) -> exp.SemanticView:
1346        kwargs: dict[str, t.Any] = {"this": self._parse_table_parts()}
1347
1348        while self._curr and not self._match(TokenType.R_PAREN, advance=False):
1349            if self._match_texts(("DIMENSIONS", "METRICS", "FACTS")):
1350                keyword = self._prev.text.lower()
1351                kwargs[keyword] = self._parse_csv(
1352                    lambda: self._parse_alias(self._parse_disjunction(), explicit=True)
1353                )
1354            elif self._match_text_seq("WHERE"):
1355                kwargs["where"] = self._parse_expression()
1356            else:
1357                self.raise_error("Expecting ) or encountered unexpected keyword")
1358                break
1359
1360        return self.expression(exp.SemanticView(**kwargs))
1361
1362    def _parse_set(self, unset: bool = False, tag: bool = False) -> exp.Set | exp.Command:
1363        set = super()._parse_set(unset=unset, tag=tag)
1364
1365        if isinstance(set, exp.Set):
1366            for expr in set.expressions:
1367                if isinstance(expr, exp.SetItem):
1368                    expr.set("kind", "VARIABLE")
1369        return set
1370
1371    def _parse_position(self, haystack_first: bool = False) -> exp.StrPosition:
1372        result = super()._parse_position(haystack_first)
1373        result.set("clamp_position", True)
1374        return result
1375
1376    def _parse_substring(self) -> exp.Substring:
1377        result = super()._parse_substring()
1378        result.set("zero_start", True)
1379        return result
1380
1381    def build_cast(self, strict: bool, **kwargs) -> exp.Expr:
1382        to = kwargs.get("to")
1383        if not strict and to and to.this == exp.DataType.Type.BOOLEAN:
1384            return self.expression(exp.ToBoolean(this=kwargs.get("this"), safe=True))
1385        cast = super().build_cast(strict, **kwargs)
1386        if isinstance(cast, exp.TryCast) and to:
1387            if to.this in exp.DataType.TEXT_TYPES and to.expressions:
1388                cast.set("null_on_text_overflow", True)
1389            elif to.this == exp.DType.DATE and cast.this.is_string:
1390                cast.set("probe_date_format", True)
1391        return cast
1392
1393    def _parse_window(self, this: exp.Expr | None, alias: bool = False) -> exp.Expr | None:
1394        if isinstance(this, exp.NthValue):
1395            if self._match_text_seq("FROM"):
1396                if self._match_texts(("FIRST", "LAST")):
1397                    from_first = self._prev.text.upper() == "FIRST"
1398                    this.set("from_first", from_first)
1399
1400        result = super()._parse_window(this, alias)
1401
1402        # Set default window frame for ranking functions if not present
1403        if (
1404            isinstance(result, exp.Window)
1405            and isinstance(this, RANKING_WINDOW_FUNCTIONS_WITH_FRAME)
1406            and not result.args.get("spec")
1407        ):
1408            frame = exp.WindowSpec(
1409                kind="ROWS",
1410                start="UNBOUNDED",
1411                start_side="PRECEDING",
1412                end="UNBOUNDED",
1413                end_side="FOLLOWING",
1414            )
1415            result.set("spec", frame)
1416        return result
1417
1418
1419# This is imported and used by both the parser (above) and the generator in the dialect file
1420RANKING_WINDOW_FUNCTIONS_WITH_FRAME = (
1421    exp.FirstValue,
1422    exp.LastValue,
1423    exp.NthValue,
1424)
1425
1426
1427def build_object_construct(args: list) -> exp.StarMap | exp.Struct:
1428    expression = parser.build_var_map(args)
1429
1430    if isinstance(expression, exp.StarMap):
1431        return expression
1432
1433    return exp.Struct(
1434        expressions=[
1435            exp.PropertyEQ(this=k, expression=v) for k, v in zip(expression.keys, expression.values)
1436        ]
1437    )