sqlglot.parsers.snowflake

   1from __future__ import annotations
   2
   3import typing as t
   4
   5from sqlglot import exp, parser
   6from sqlglot.dialects.dialect import (
   7    Dialect,
   8    binary_from_function,
   9    build_default_decimal_type,
  10    build_formatted_time,
  11    build_like,
  12    build_replace_with_optional_replacement,
  13    build_timetostr_or_tochar,
  14    build_trunc,
  15    date_trunc_to_time,
  16    map_date_part,
  17)
  18from sqlglot.helper import is_date_unit, is_int, seq_get
  19from sqlglot.tokens import TokenType
  20from sqlglot.trie import new_trie
  21
  22if t.TYPE_CHECKING:
  23    from collections.abc import Collection
  24
  25    from sqlglot._typing import B, E
  26    from sqlglot.dialects.dialect import Dialect
  27
  28
  29def _build_approx_top_k(args: list) -> exp.ApproxTopK:
  30    """
  31    Normalizes APPROX_TOP_K arguments to match Snowflake semantics.
  32
  33    Snowflake APPROX_TOP_K signature: APPROX_TOP_K(column [, k] [, counters])
  34    - k defaults to 1 if omitted (per Snowflake documentation)
  35    - counters is optional precision parameter
  36    """
  37    # Add default k=1 if only column is provided
  38    if len(args) == 1:
  39        args.append(exp.Literal.number(1))
  40
  41    return exp.ApproxTopK.from_arg_list(args)
  42
  43
  44def _build_to_number(args: list, safe: bool = False) -> exp.ToNumber:
  45    second_arg = seq_get(args, 1)
  46    if second_arg and second_arg.is_number:
  47        fmt = None
  48        precision = second_arg
  49        scale = seq_get(args, 2) or exp.Literal.number(0)
  50    else:
  51        fmt = second_arg
  52        precision = seq_get(args, 2) or exp.Literal.number(38)
  53        scale = seq_get(args, 3) or exp.Literal.number(0)
  54
  55    return exp.ToNumber(
  56        this=seq_get(args, 0),
  57        format=fmt,
  58        precision=precision,
  59        scale=scale,
  60        safe=safe,
  61    )
  62
  63
  64def _build_date_from_parts(args: list) -> exp.DateFromParts:
  65    return exp.DateFromParts(
  66        year=seq_get(args, 0),
  67        month=seq_get(args, 1),
  68        day=seq_get(args, 2),
  69        allow_overflow=True,
  70    )
  71
  72
  73# Timestamp types used in _build_datetime
  74TIMESTAMP_TYPES = {
  75    exp.DType.TIMESTAMP: "TO_TIMESTAMP",
  76    exp.DType.TIMESTAMPLTZ: "TO_TIMESTAMP_LTZ",
  77    exp.DType.TIMESTAMPNTZ: "TO_TIMESTAMP_NTZ",
  78    exp.DType.TIMESTAMPTZ: "TO_TIMESTAMP_TZ",
  79}
  80
  81
  82def _build_datetime(name: str, kind: exp.DType, safe: bool = False) -> t.Callable:
  83    def _builder(args: list, dialect: Dialect) -> exp.Func:
  84        value = seq_get(args, 0)
  85        scale_or_fmt = seq_get(args, 1)
  86
  87        int_value = value is not None and is_int(value.name)
  88        int_scale_or_fmt = scale_or_fmt is not None and scale_or_fmt.is_int
  89
  90        if isinstance(value, (exp.Literal, exp.Neg)) or (value and scale_or_fmt):
  91            # Converts calls like `TO_TIME('01:02:03')` into casts
  92            if len(args) == 1 and value.is_string and not int_value:
  93                cast = (
  94                    exp.TryCast(this=value, to=kind.into_expr(), requires_string=True)
  95                    if safe
  96                    else exp.cast(value, kind)
  97                )
  98                if safe and kind == exp.DType.DATE:
  99                    cast.set("probe_date_format", True)
 100                return cast
 101
 102            # Handles `TO_TIMESTAMP(str, fmt)` and `TO_TIMESTAMP(num, scale)` as special
 103            # cases so we can transpile them, since they're relatively common
 104            if kind in TIMESTAMP_TYPES:
 105                if not safe and (int_scale_or_fmt or (int_value and scale_or_fmt is None)):
 106                    # TRY_TO_TIMESTAMP('integer') is not parsed into exp.UnixToTime as
 107                    # it's not easily transpilable. Also, numeric-looking strings with
 108                    # format strings (e.g., TO_TIMESTAMP('20240115', 'YYYYMMDD')) should
 109                    # use StrToTime, not UnixToTime.
 110                    unix_expr = exp.UnixToTime(this=value, scale=scale_or_fmt)
 111                    unix_expr.set("target_type", kind.into_expr())
 112                    return unix_expr
 113                if scale_or_fmt and not int_scale_or_fmt:
 114                    # Format string provided (e.g., 'YYYY-MM-DD'), use StrToTime
 115                    strtotime_expr = build_formatted_time(exp.StrToTime)(args, dialect)
 116                    strtotime_expr.set("safe", safe)
 117                    strtotime_expr.set("target_type", kind.into_expr())
 118                    return strtotime_expr
 119
 120        # Handle DATE/TIME with format strings - allow int_value if a format string is provided
 121        has_format_string = scale_or_fmt and not int_scale_or_fmt
 122        if kind in (exp.DType.DATE, exp.DType.TIME) and (not int_value or has_format_string):
 123            klass = exp.TsOrDsToDate if kind == exp.DType.DATE else exp.TsOrDsToTime
 124            formatted_exp = build_formatted_time(klass)(args, dialect)
 125            formatted_exp.set("safe", safe)
 126            return formatted_exp
 127
 128        return exp.Anonymous(this=name, expressions=args)
 129
 130    return _builder
 131
 132
 133def _build_bitwise(expr_type: type[B], name: str) -> t.Callable[[list], B | exp.Anonymous]:
 134    def _builder(args: list) -> B | exp.Anonymous:
 135        if len(args) == 3:
 136            # Special handling for bitwise operations with padside argument
 137            if expr_type in (exp.BitwiseAnd, exp.BitwiseOr, exp.BitwiseXor):
 138                return expr_type(
 139                    this=seq_get(args, 0), expression=seq_get(args, 1), padside=seq_get(args, 2)
 140                )
 141            return exp.Anonymous(this=name, expressions=args)
 142
 143        result = binary_from_function(expr_type)(args)
 144
 145        # Snowflake specifies INT128 for bitwise shifts
 146        if expr_type in (exp.BitwiseLeftShift, exp.BitwiseRightShift):
 147            result.set("requires_int128", True)
 148
 149        return result
 150
 151    return _builder
 152
 153
 154# https://docs.snowflake.com/en/sql-reference/functions/div0
 155def _build_if_from_div0(args: list) -> exp.If:
 156    lhs = exp._wrap(seq_get(args, 0), exp.Binary)
 157    rhs = exp._wrap(seq_get(args, 1), exp.Binary)
 158
 159    cond = exp.EQ(this=rhs, expression=exp.Literal.number(0)).and_(
 160        exp.Is(this=lhs, expression=exp.null()).not_()
 161    )
 162    true = exp.Literal.number(0)
 163    false = exp.Div(this=lhs, expression=rhs)
 164    return exp.If(this=cond, true=true, false=false)
 165
 166
 167# https://docs.snowflake.com/en/sql-reference/functions/div0null
 168def _build_if_from_div0null(args: list) -> exp.If:
 169    lhs = exp._wrap(seq_get(args, 0), exp.Binary)
 170    rhs = exp._wrap(seq_get(args, 1), exp.Binary)
 171
 172    # Returns 0 when divisor is 0 OR NULL
 173    cond = exp.EQ(this=rhs, expression=exp.Literal.number(0)).or_(
 174        exp.Is(this=rhs, expression=exp.null())
 175    )
 176    true = exp.Literal.number(0)
 177    false = exp.Div(this=lhs, expression=rhs)
 178    return exp.If(this=cond, true=true, false=false)
 179
 180
 181# https://docs.snowflake.com/en/sql-reference/functions/zeroifnull
 182def _build_if_from_zeroifnull(args: list) -> exp.If:
 183    cond = exp.Is(this=seq_get(args, 0), expression=exp.Null())
 184    return exp.If(this=cond, true=exp.Literal.number(0), false=seq_get(args, 0))
 185
 186
 187def _build_search(args: list) -> exp.Search:
 188    kwargs = {
 189        "this": seq_get(args, 0),
 190        "expression": seq_get(args, 1),
 191        **{arg.name.lower(): arg for arg in args[2:] if isinstance(arg, exp.Kwarg)},
 192    }
 193    return exp.Search(**kwargs)
 194
 195
 196# https://docs.snowflake.com/en/sql-reference/functions/zeroifnull
 197def _build_if_from_nullifzero(args: list) -> exp.If:
 198    cond = exp.EQ(this=seq_get(args, 0), expression=exp.Literal.number(0))
 199    return exp.If(this=cond, true=exp.Null(), false=seq_get(args, 0))
 200
 201
 202def _build_regexp_replace(args: list) -> exp.RegexpReplace:
 203    regexp_replace = exp.RegexpReplace.from_arg_list(args)
 204
 205    if not regexp_replace.args.get("replacement"):
 206        regexp_replace.set("replacement", exp.Literal.string(""))
 207
 208    return regexp_replace
 209
 210
 211def _build_regexp_like(args: list) -> exp.RegexpLike:
 212    return exp.RegexpLike(
 213        this=seq_get(args, 0),
 214        expression=seq_get(args, 1),
 215        flag=seq_get(args, 2),
 216        full_match=True,
 217    )
 218
 219
 220def _date_trunc_to_time(args: list) -> exp.DateTrunc | exp.TimestampTrunc:
 221    trunc = date_trunc_to_time(args)
 222    unit = map_date_part(trunc.args["unit"])
 223    trunc.set("unit", unit)
 224    is_time_input = trunc.this.is_type(exp.DType.TIME, exp.DType.TIMETZ)
 225    if (isinstance(trunc, exp.TimestampTrunc) and is_date_unit(unit) or is_time_input) or (
 226        isinstance(trunc, exp.DateTrunc) and not is_date_unit(unit)
 227    ):
 228        trunc.set("input_type_preserved", True)
 229    return trunc
 230
 231
 232def _build_regexp_extract(expr_type: type[E]) -> t.Callable[[list, Dialect], E]:
 233    def _builder(args: list, dialect: Dialect) -> E:
 234        return expr_type(
 235            this=seq_get(args, 0),
 236            expression=seq_get(args, 1),
 237            position=seq_get(args, 2),
 238            occurrence=seq_get(args, 3),
 239            parameters=seq_get(args, 4),
 240            group=seq_get(args, 5) or exp.Literal.number(0),
 241            **(
 242                {"null_if_pos_overflow": dialect.REGEXP_EXTRACT_POSITION_OVERFLOW_RETURNS_NULL}
 243                if expr_type is exp.RegexpExtract
 244                else {}
 245            ),
 246        )
 247
 248    return _builder
 249
 250
 251def _build_timestamp_from_parts(args: list) -> exp.Func:
 252    """Build TimestampFromParts with support for both syntaxes:
 253    1. TIMESTAMP_FROM_PARTS(year, month, day, hour, minute, second [, nanosecond] [, time_zone])
 254    2. TIMESTAMP_FROM_PARTS(date_expr, time_expr) - Snowflake specific
 255    """
 256    if len(args) == 2:
 257        return exp.TimestampFromParts(this=seq_get(args, 0), expression=seq_get(args, 1))
 258
 259    return exp.TimestampFromParts.from_arg_list(args)
 260
 261
 262def _build_round(args: list) -> exp.Round:
 263    """
 264    Build Round expression, unwrapping Snowflake's named parameters.
 265
 266    Maps EXPR => this, SCALE => decimals, ROUNDING_MODE => truncate.
 267
 268    Note: Snowflake does not support mixing named and positional arguments.
 269    Arguments are either all named or all positional.
 270    """
 271    kwarg_map = {"EXPR": "this", "SCALE": "decimals", "ROUNDING_MODE": "truncate"}
 272    round_args = {}
 273    positional_keys = ["this", "decimals", "truncate"]
 274    positional_idx = 0
 275
 276    for arg in args:
 277        if isinstance(arg, exp.Kwarg):
 278            key = arg.this.name.upper()
 279            round_key = kwarg_map.get(key)
 280            if round_key:
 281                round_args[round_key] = arg.expression
 282        else:
 283            if positional_idx < len(positional_keys):
 284                round_args[positional_keys[positional_idx]] = arg
 285                positional_idx += 1
 286
 287    expression = exp.Round(**round_args)
 288    expression.set("casts_non_integer_decimals", True)
 289    return expression
 290
 291
 292def _build_array_sort(args: list) -> exp.SortArray:
 293    asc = seq_get(args, 1)
 294    nulls_first = seq_get(args, 2)
 295    if nulls_first is None and isinstance(asc, exp.Boolean):
 296        nulls_first = exp.Boolean(this=not asc.this)
 297    return exp.SortArray(this=seq_get(args, 0), asc=asc, nulls_first=nulls_first)
 298
 299
 300def _build_generator(args: list) -> exp.Generator:
 301    """
 302    Build Generator expression, unwrapping Snowflake's named parameters.
 303
 304    Maps ROWCOUNT => rowcount, TIMELIMIT => timelimit.
 305    """
 306    kwarg_map = {"ROWCOUNT": "rowcount", "TIMELIMIT": "timelimit"}
 307    gen_args = {}
 308
 309    positional_keys = ("rowcount", "timelimit")
 310
 311    for i, arg in enumerate(args):
 312        if isinstance(arg, exp.Kwarg):
 313            key = arg.this.name.upper()
 314            gen_key = kwarg_map.get(key)
 315            if gen_key:
 316                gen_args[gen_key] = arg.expression
 317        elif i < len(positional_keys):
 318            gen_args[positional_keys[i]] = arg
 319
 320    return exp.Generator(**gen_args)
 321
 322
 323def _show_parser(*args: t.Any, **kwargs: t.Any) -> t.Callable[[SnowflakeParser], exp.Show]:
 324    def _parse(self: SnowflakeParser) -> exp.Show:
 325        return self._parse_show_snowflake(*args, **kwargs)
 326
 327    return _parse
 328
 329
 330class SnowflakeParser(parser.Parser):
 331    IDENTIFY_PIVOT_STRINGS = True
 332    TYPED_LAMBDA_ARGS = True
 333    DEFAULT_SAMPLING_METHOD = "BERNOULLI"
 334    COLON_IS_VARIANT_EXTRACT = True
 335    JSON_EXTRACT_REQUIRES_JSON_EXPRESSION = True
 336    SUPPORTS_NTH_VALUE_FROM_MODIFIER = True
 337
 338    TYPE_TOKENS = {*parser.Parser.TYPE_TOKENS, TokenType.FILE}
 339    STRUCT_TYPE_TOKENS = {*parser.Parser.STRUCT_TYPE_TOKENS, TokenType.FILE}
 340    NESTED_TYPE_TOKENS = {*parser.Parser.NESTED_TYPE_TOKENS, TokenType.FILE}
 341
 342    ID_VAR_TOKENS = {
 343        *parser.Parser.ID_VAR_TOKENS,
 344        TokenType.EXCEPT,
 345        TokenType.INTEGRATION,
 346        TokenType.MATCH_CONDITION,
 347        TokenType.PACKAGE,
 348        TokenType.POLICY,
 349        TokenType.POOL,
 350        TokenType.ROLE,
 351        TokenType.ROLLBACK,
 352        TokenType.RULE,
 353        TokenType.VOLUME,
 354    }
 355
 356    FUNC_TOKENS = parser.Parser.FUNC_TOKENS | {TokenType.ROLLBACK}
 357
 358    ALIAS_TOKENS = parser.Parser.ALIAS_TOKENS | {
 359        TokenType.INTEGRATION,
 360        TokenType.PACKAGE,
 361        TokenType.POLICY,
 362        TokenType.POOL,
 363        TokenType.ROLE,
 364        TokenType.ROLLBACK,
 365        TokenType.RULE,
 366        TokenType.VOLUME,
 367    }
 368
 369    TABLE_ALIAS_TOKENS = (
 370        parser.Parser.TABLE_ALIAS_TOKENS
 371        | {
 372            TokenType.ANTI,
 373            TokenType.INTEGRATION,
 374            TokenType.PACKAGE,
 375            TokenType.POLICY,
 376            TokenType.POOL,
 377            TokenType.ROLE,
 378            TokenType.ROLLBACK,
 379            TokenType.RULE,
 380            TokenType.SEMI,
 381            TokenType.VOLUME,
 382            TokenType.WINDOW,
 383        }
 384    ) - {TokenType.MATCH_CONDITION}
 385
 386    COLON_PLACEHOLDER_TOKENS = ID_VAR_TOKENS | {TokenType.NUMBER}
 387
 388    NO_PAREN_FUNCTIONS = {
 389        **parser.Parser.NO_PAREN_FUNCTIONS,
 390        TokenType.LOCALTIME: exp.Localtime,
 391        TokenType.LOCALTIMESTAMP: exp.Localtimestamp,
 392        TokenType.CURRENT_TIME: exp.Localtime,
 393    }
 394
 395    RANGE_PARSERS = {
 396        **parser.Parser.RANGE_PARSERS,
 397        TokenType.RLIKE: lambda self, this: self.expression(
 398            exp.RegexpLike(this=this, expression=self._parse_bitwise(), full_match=True)
 399        ),
 400    }
 401
 402    FUNCTIONS = {
 403        **parser.Parser.FUNCTIONS,
 404        "CHARINDEX": lambda args: exp.StrPosition(
 405            this=seq_get(args, 1),
 406            substr=seq_get(args, 0),
 407            position=seq_get(args, 2),
 408            clamp_position=True,
 409        ),
 410        "ADD_MONTHS": lambda args: exp.AddMonths(
 411            this=seq_get(args, 0),
 412            expression=seq_get(args, 1),
 413            preserve_end_of_month=True,
 414        ),
 415        "APPROX_PERCENTILE": exp.ApproxQuantile.from_arg_list,
 416        "CURRENT_TIME": lambda args: exp.Localtime(this=seq_get(args, 0)),
 417        "APPROX_TOP_K": _build_approx_top_k,
 418        "ARRAY_CONSTRUCT": lambda args: exp.Array(expressions=args),
 419        "ARRAY_CONTAINS": lambda args: exp.ArrayContains(
 420            this=seq_get(args, 1),
 421            expression=seq_get(args, 0),
 422            ensure_variant=False,
 423            check_null=True,
 424        ),
 425        "ARRAY_DISTINCT": lambda args: exp.ArrayDistinct(
 426            this=seq_get(args, 0),
 427            check_null=True,
 428        ),
 429        "ARRAY_GENERATE_RANGE": lambda args: exp.GenerateSeries(
 430            # Snowflake has exclusive end semantics
 431            start=seq_get(args, 0),
 432            end=seq_get(args, 1),
 433            step=seq_get(args, 2),
 434            is_end_exclusive=True,
 435        ),
 436        "ARRAY_EXCEPT": lambda args: exp.ArrayExcept(
 437            this=seq_get(args, 0),
 438            expression=seq_get(args, 1),
 439            is_multiset=True,
 440        ),
 441        "ARRAY_INTERSECTION": lambda args: exp.ArrayIntersect(
 442            expressions=args,
 443            is_multiset=True,
 444        ),
 445        "ARRAY_POSITION": lambda args: exp.ArrayPosition(
 446            this=seq_get(args, 1),
 447            expression=seq_get(args, 0),
 448            zero_based=True,
 449        ),
 450        "ARRAY_SLICE": lambda args: exp.ArraySlice(
 451            this=seq_get(args, 0),
 452            start=seq_get(args, 1),
 453            end=seq_get(args, 2),
 454            zero_based=True,
 455        ),
 456        "ARRAY_SORT": _build_array_sort,
 457        "ARRAY_FLATTEN": exp.Flatten.from_arg_list,
 458        "ARRAY_TO_STRING": lambda args: exp.ArrayToString(
 459            this=seq_get(args, 0),
 460            expression=seq_get(args, 1),
 461            null_is_empty=True,
 462            null_delim_is_null=True,
 463        ),
 464        "ARRAYS_OVERLAP": lambda args: exp.ArrayOverlaps(
 465            this=seq_get(args, 0), expression=seq_get(args, 1), null_safe=True
 466        ),
 467        "BITAND": _build_bitwise(exp.BitwiseAnd, "BITAND"),
 468        "BIT_AND": _build_bitwise(exp.BitwiseAnd, "BITAND"),
 469        "BITNOT": lambda args: exp.BitwiseNot(this=seq_get(args, 0)),
 470        "BIT_NOT": lambda args: exp.BitwiseNot(this=seq_get(args, 0)),
 471        "BITXOR": _build_bitwise(exp.BitwiseXor, "BITXOR"),
 472        "BIT_XOR": _build_bitwise(exp.BitwiseXor, "BITXOR"),
 473        "BITOR": _build_bitwise(exp.BitwiseOr, "BITOR"),
 474        "BIT_OR": _build_bitwise(exp.BitwiseOr, "BITOR"),
 475        "BITSHIFTLEFT": _build_bitwise(exp.BitwiseLeftShift, "BITSHIFTLEFT"),
 476        "BIT_SHIFTLEFT": _build_bitwise(exp.BitwiseLeftShift, "BIT_SHIFTLEFT"),
 477        "BITSHIFTRIGHT": _build_bitwise(exp.BitwiseRightShift, "BITSHIFTRIGHT"),
 478        "BIT_SHIFTRIGHT": _build_bitwise(exp.BitwiseRightShift, "BIT_SHIFTRIGHT"),
 479        "BITANDAGG": exp.BitwiseAndAgg.from_arg_list,
 480        "BITAND_AGG": exp.BitwiseAndAgg.from_arg_list,
 481        "BIT_AND_AGG": exp.BitwiseAndAgg.from_arg_list,
 482        "BIT_ANDAGG": exp.BitwiseAndAgg.from_arg_list,
 483        "BITORAGG": exp.BitwiseOrAgg.from_arg_list,
 484        "BITOR_AGG": exp.BitwiseOrAgg.from_arg_list,
 485        "BIT_OR_AGG": exp.BitwiseOrAgg.from_arg_list,
 486        "BIT_ORAGG": exp.BitwiseOrAgg.from_arg_list,
 487        "BITXORAGG": exp.BitwiseXorAgg.from_arg_list,
 488        "BITXOR_AGG": exp.BitwiseXorAgg.from_arg_list,
 489        "BIT_XOR_AGG": exp.BitwiseXorAgg.from_arg_list,
 490        "BIT_XORAGG": exp.BitwiseXorAgg.from_arg_list,
 491        "BITMAP_OR_AGG": exp.BitmapOrAgg.from_arg_list,
 492        "BOOLAND": lambda args: exp.Booland(
 493            this=seq_get(args, 0), expression=seq_get(args, 1), round_input=True
 494        ),
 495        "BOOLOR": lambda args: exp.Boolor(
 496            this=seq_get(args, 0), expression=seq_get(args, 1), round_input=True
 497        ),
 498        "BOOLNOT": lambda args: exp.Boolnot(this=seq_get(args, 0), round_input=True),
 499        "BOOLXOR": lambda args: exp.Xor(
 500            this=seq_get(args, 0), expression=seq_get(args, 1), round_input=True
 501        ),
 502        "CORR": lambda args: exp.Corr(
 503            this=seq_get(args, 0),
 504            expression=seq_get(args, 1),
 505            null_on_zero_variance=True,
 506        ),
 507        "COUNT_IF": lambda args: exp.CountIf(this=seq_get(args, 0), zero_on_all_null=True),
 508        "DATE": _build_datetime("DATE", exp.DType.DATE),
 509        "DATEFROMPARTS": _build_date_from_parts,
 510        "DATE_FROM_PARTS": _build_date_from_parts,
 511        "DATE_TRUNC": _date_trunc_to_time,
 512        "DATEADD": lambda args: exp.DateAdd(
 513            this=seq_get(args, 2),
 514            expression=seq_get(args, 1),
 515            unit=map_date_part(seq_get(args, 0)),
 516        ),
 517        "DATEDIFF": lambda args: exp.DateDiff(
 518            this=seq_get(args, 2),
 519            expression=seq_get(args, 1),
 520            unit=map_date_part(seq_get(args, 0)),
 521            date_part_boundary=True,
 522        ),
 523        "DAYNAME": lambda args: exp.Dayname(this=seq_get(args, 0), abbreviated=True),
 524        "DAYOFWEEKISO": exp.DayOfWeekIso.from_arg_list,
 525        "DIV0": _build_if_from_div0,
 526        "DIV0NULL": _build_if_from_div0null,
 527        "EDITDISTANCE": lambda args: exp.Levenshtein(
 528            this=seq_get(args, 0), expression=seq_get(args, 1), max_dist=seq_get(args, 2)
 529        ),
 530        "FLATTEN": exp.Explode.from_arg_list,
 531        "GENERATOR": _build_generator,
 532        "GET": exp.GetExtract.from_arg_list,
 533        "GETDATE": exp.CurrentTimestamp.from_arg_list,
 534        "GET_PATH": lambda args, dialect: exp.JSONExtract(
 535            this=seq_get(args, 0),
 536            expression=dialect.to_json_path(seq_get(args, 1)),
 537            requires_json=True,
 538        ),
 539        "GREATEST_IGNORE_NULLS": lambda args: exp.Greatest(
 540            this=seq_get(args, 0), expressions=args[1:], ignore_nulls=True
 541        ),
 542        "LEAST_IGNORE_NULLS": lambda args: exp.Least(
 543            this=seq_get(args, 0), expressions=args[1:], ignore_nulls=True
 544        ),
 545        "LEFT": lambda args: exp.Left(
 546            this=seq_get(args, 0), expression=seq_get(args, 1), negative_length_returns_empty=True
 547        ),
 548        "HEX_DECODE_BINARY": exp.Unhex.from_arg_list,
 549        "HEX_ENCODE": exp.Hex.from_arg_list,
 550        "IFF": exp.If.from_arg_list,
 551        "JAROWINKLER_SIMILARITY": lambda args: exp.JarowinklerSimilarity(
 552            this=seq_get(args, 0),
 553            expression=seq_get(args, 1),
 554            case_insensitive=True,
 555            integer_scale=True,
 556        ),
 557        "MD5_HEX": exp.MD5.from_arg_list,
 558        "MD5_BINARY": exp.MD5Digest.from_arg_list,
 559        "MD5_NUMBER_LOWER64": exp.MD5NumberLower64.from_arg_list,
 560        "MD5_NUMBER_UPPER64": exp.MD5NumberUpper64.from_arg_list,
 561        "MONTHNAME": lambda args: exp.Monthname(this=seq_get(args, 0), abbreviated=True),
 562        "LAST_DAY": lambda args: exp.LastDay(
 563            this=seq_get(args, 0), unit=map_date_part(seq_get(args, 1))
 564        ),
 565        "LEN": lambda args: exp.Length(this=seq_get(args, 0), binary=True),
 566        "LENGTH": lambda args: exp.Length(this=seq_get(args, 0), binary=True),
 567        "LOCALTIMESTAMP": exp.CurrentTimestamp.from_arg_list,
 568        "NULLIFZERO": _build_if_from_nullifzero,
 569        "OBJECT_CONSTRUCT": lambda args: build_object_construct(args),
 570        "OBJECT_KEYS": exp.JSONKeys.from_arg_list,
 571        "OCTET_LENGTH": exp.ByteLength.from_arg_list,
 572        "PARSE_URL": lambda args: exp.ParseUrl(this=seq_get(args, 0), permissive=seq_get(args, 1)),
 573        "REGEXP_EXTRACT_ALL": _build_regexp_extract(exp.RegexpExtractAll),
 574        "REGEXP_LIKE": _build_regexp_like,
 575        "REGEXP_REPLACE": _build_regexp_replace,
 576        "REGEXP_SUBSTR": _build_regexp_extract(exp.RegexpExtract),
 577        "REGEXP_SUBSTR_ALL": _build_regexp_extract(exp.RegexpExtractAll),
 578        "RANDOM": lambda args: exp.Rand(
 579            this=seq_get(args, 0),
 580            lower=exp.Literal.number(-9223372036854775808.0),  # -2^63 as float to avoid overflow
 581            upper=exp.Literal.number(9223372036854775807.0),  # 2^63-1 as float
 582        ),
 583        "REPLACE": build_replace_with_optional_replacement,
 584        "RIGHT": lambda args: exp.Right(
 585            this=seq_get(args, 0), expression=seq_get(args, 1), negative_length_returns_empty=True
 586        ),
 587        "RLIKE": _build_regexp_like,
 588        "ROUND": _build_round,
 589        "SHA1_BINARY": exp.SHA1Digest.from_arg_list,
 590        "SHA1_HEX": exp.SHA.from_arg_list,
 591        "SHA2_BINARY": exp.SHA2Digest.from_arg_list,
 592        "SHA2_HEX": exp.SHA2.from_arg_list,
 593        "SPLIT": lambda args: exp.Split(
 594            this=seq_get(args, 0),
 595            expression=seq_get(args, 1),
 596            null_returns_null=True,
 597            empty_delimiter_returns_whole=True,
 598        ),
 599        "SQUARE": lambda args: exp.Pow(this=seq_get(args, 0), expression=exp.Literal.number(2)),
 600        "STDDEV_SAMP": exp.Stddev.from_arg_list,
 601        "SYSDATE": lambda args: exp.CurrentTimestamp(this=seq_get(args, 0), sysdate=True),
 602        "TABLE": lambda args: exp.TableFromRows(this=seq_get(args, 0)),
 603        "TIMEADD": lambda args: exp.TimeAdd(
 604            this=seq_get(args, 2),
 605            expression=seq_get(args, 1),
 606            unit=map_date_part(seq_get(args, 0)),
 607        ),
 608        "TIMEDIFF": lambda args: exp.DateDiff(
 609            this=seq_get(args, 2),
 610            expression=seq_get(args, 1),
 611            unit=map_date_part(seq_get(args, 0)),
 612            date_part_boundary=True,
 613        ),
 614        "TIME_FROM_PARTS": lambda args: exp.TimeFromParts(
 615            hour=seq_get(args, 0),
 616            min=seq_get(args, 1),
 617            sec=seq_get(args, 2),
 618            nano=seq_get(args, 3),
 619            overflow=True,
 620        ),
 621        "TIMESTAMPADD": lambda args: exp.DateAdd(
 622            this=seq_get(args, 2),
 623            expression=seq_get(args, 1),
 624            unit=map_date_part(seq_get(args, 0)),
 625        ),
 626        "TIMESTAMPDIFF": lambda args: exp.DateDiff(
 627            this=seq_get(args, 2),
 628            expression=seq_get(args, 1),
 629            unit=map_date_part(seq_get(args, 0)),
 630            date_part_boundary=True,
 631        ),
 632        "TIMESTAMPFROMPARTS": _build_timestamp_from_parts,
 633        "TIMESTAMP_FROM_PARTS": _build_timestamp_from_parts,
 634        "TIMESTAMPNTZFROMPARTS": _build_timestamp_from_parts,
 635        "TIMESTAMP_NTZ_FROM_PARTS": _build_timestamp_from_parts,
 636        "TRUNC": lambda args, dialect: build_trunc(
 637            args, dialect, date_trunc_requires_part=False, fractions_supported=True
 638        ),
 639        "TRUNCATE": lambda args, dialect: build_trunc(
 640            args, dialect, date_trunc_requires_part=False, fractions_supported=True
 641        ),
 642        "TRY_DECRYPT": lambda args: exp.Decrypt(
 643            this=seq_get(args, 0),
 644            passphrase=seq_get(args, 1),
 645            aad=seq_get(args, 2),
 646            encryption_method=seq_get(args, 3),
 647            safe=True,
 648        ),
 649        "TRY_DECRYPT_RAW": lambda args: exp.DecryptRaw(
 650            this=seq_get(args, 0),
 651            key=seq_get(args, 1),
 652            iv=seq_get(args, 2),
 653            aad=seq_get(args, 3),
 654            encryption_method=seq_get(args, 4),
 655            aead=seq_get(args, 5),
 656            safe=True,
 657        ),
 658        "TRY_PARSE_JSON": lambda args: exp.ParseJSON(this=seq_get(args, 0), safe=True),
 659        "TRY_TO_BINARY": lambda args: exp.ToBinary(
 660            this=seq_get(args, 0), format=seq_get(args, 1), safe=True
 661        ),
 662        "TRY_TO_BOOLEAN": lambda args: exp.ToBoolean(this=seq_get(args, 0), safe=True),
 663        "TRY_TO_DATE": _build_datetime("TRY_TO_DATE", exp.DType.DATE, safe=True),
 664        **dict.fromkeys(
 665            ("TRY_TO_DECIMAL", "TRY_TO_NUMBER", "TRY_TO_NUMERIC"),
 666            lambda args: _build_to_number(args, safe=True),
 667        ),
 668        "TRY_TO_DOUBLE": lambda args: exp.ToDouble(
 669            this=seq_get(args, 0), format=seq_get(args, 1), safe=True
 670        ),
 671        "TRY_TO_FILE": lambda args: exp.ToFile(
 672            this=seq_get(args, 0), path=seq_get(args, 1), safe=True
 673        ),
 674        "TRY_TO_TIME": _build_datetime("TRY_TO_TIME", exp.DType.TIME, safe=True),
 675        "TRY_TO_TIMESTAMP": _build_datetime("TRY_TO_TIMESTAMP", exp.DType.TIMESTAMP, safe=True),
 676        "TRY_TO_TIMESTAMP_LTZ": _build_datetime(
 677            "TRY_TO_TIMESTAMP_LTZ", exp.DType.TIMESTAMPLTZ, safe=True
 678        ),
 679        "TRY_TO_TIMESTAMP_NTZ": _build_datetime(
 680            "TRY_TO_TIMESTAMP_NTZ", exp.DType.TIMESTAMPNTZ, safe=True
 681        ),
 682        "TRY_TO_TIMESTAMP_TZ": _build_datetime(
 683            "TRY_TO_TIMESTAMP_TZ", exp.DType.TIMESTAMPTZ, safe=True
 684        ),
 685        "TO_CHAR": build_timetostr_or_tochar,
 686        "TO_DATE": _build_datetime("TO_DATE", exp.DType.DATE),
 687        **dict.fromkeys(
 688            ("TO_DECIMAL", "TO_NUMBER", "TO_NUMERIC"),
 689            lambda args: _build_to_number(args),
 690        ),
 691        "TO_TIME": _build_datetime("TO_TIME", exp.DType.TIME),
 692        "TO_TIMESTAMP": _build_datetime("TO_TIMESTAMP", exp.DType.TIMESTAMP),
 693        "TO_TIMESTAMP_LTZ": _build_datetime("TO_TIMESTAMP_LTZ", exp.DType.TIMESTAMPLTZ),
 694        "TO_TIMESTAMP_NTZ": _build_datetime("TO_TIMESTAMP_NTZ", exp.DType.TIMESTAMPNTZ),
 695        "TO_TIMESTAMP_TZ": _build_datetime("TO_TIMESTAMP_TZ", exp.DType.TIMESTAMPTZ),
 696        "TO_GEOGRAPHY": lambda args: (
 697            exp.cast(args[0], exp.DType.GEOGRAPHY)
 698            if len(args) == 1
 699            else exp.Anonymous(this="TO_GEOGRAPHY", expressions=args)
 700        ),
 701        "TO_GEOMETRY": lambda args: (
 702            exp.cast(args[0], exp.DType.GEOMETRY)
 703            if len(args) == 1
 704            else exp.Anonymous(this="TO_GEOMETRY", expressions=args)
 705        ),
 706        "TO_VARCHAR": build_timetostr_or_tochar,
 707        "TO_JSON": exp.JSONFormat.from_arg_list,
 708        "VECTOR_COSINE_SIMILARITY": exp.CosineDistance.from_arg_list,
 709        "VECTOR_INNER_PRODUCT": exp.DotProduct.from_arg_list,
 710        "VECTOR_L1_DISTANCE": exp.ManhattanDistance.from_arg_list,
 711        "VECTOR_L2_DISTANCE": exp.EuclideanDistance.from_arg_list,
 712        "ZEROIFNULL": _build_if_from_zeroifnull,
 713        "LIKE": build_like(exp.Like),
 714        "ILIKE": build_like(exp.ILike),
 715        "SEARCH": _build_search,
 716        "SKEW": exp.Skewness.from_arg_list,
 717        "SPLIT_PART": lambda args: exp.SplitPart(
 718            this=seq_get(args, 0),
 719            delimiter=seq_get(args, 1),
 720            part_index=seq_get(args, 2),
 721            part_index_zero_as_one=True,
 722            empty_delimiter_returns_whole=True,
 723        ),
 724        "STRTOK": lambda args: exp.Strtok(
 725            this=seq_get(args, 0),
 726            delimiter=seq_get(args, 1) or exp.Literal.string(" "),
 727            part_index=seq_get(args, 2) or exp.Literal.number("1"),
 728        ),
 729        "STRTOK_TO_ARRAY": lambda args: exp.StrtokToArray(
 730            this=seq_get(args, 0),
 731            expression=seq_get(args, 1) or exp.Literal.string(" "),
 732        ),
 733        "SYSTIMESTAMP": exp.CurrentTimestamp.from_arg_list,
 734        "IDENTIFIER": exp.DynamicIdentifier.from_arg_list,
 735        "UNICODE": lambda args: exp.Unicode(this=seq_get(args, 0), empty_is_zero=True),
 736        "WEEKISO": exp.WeekOfYear.from_arg_list,
 737        "WEEKOFYEAR": exp.Week.from_arg_list,
 738    }
 739    FUNCTIONS = {k: v for k, v in FUNCTIONS.items() if k != "PREDICT"}
 740
 741    FUNCTION_PARSERS = {
 742        **parser.Parser.FUNCTION_PARSERS,
 743        "DATE_PART": lambda self: self._parse_date_part(),
 744        "DIRECTORY": lambda self: self._parse_directory(),
 745        "OBJECT_CONSTRUCT_KEEP_NULL": lambda self: self._parse_object_construct_keep_null(),
 746        "LISTAGG": lambda self: self._parse_string_agg(),
 747        "SEMANTIC_VIEW": lambda self: self._parse_semantic_view(),
 748        "SUBSTR": lambda self: self._parse_substring(),
 749    }
 750    FUNCTION_PARSERS = {k: v for k, v in FUNCTION_PARSERS.items() if k != "TRIM"}
 751
 752    TIMESTAMPS = parser.Parser.TIMESTAMPS - {TokenType.TIME}
 753
 754    ALTER_PARSERS = {
 755        **parser.Parser.ALTER_PARSERS,
 756        "MODIFY": lambda self: self._parse_alter_table_alter(),
 757        "SESSION": lambda self: self._parse_alter_session(),
 758        "UNSET": lambda self: self.expression(
 759            exp.Set(
 760                tag=self._match_text_seq("TAG"),
 761                expressions=self._parse_csv(self._parse_id_var),
 762                unset=True,
 763            )
 764        ),
 765    }
 766
 767    STATEMENT_PARSERS = {
 768        **parser.Parser.STATEMENT_PARSERS,
 769        TokenType.GET: lambda self: self._parse_get(),
 770        TokenType.PUT: lambda self: self._parse_put(),
 771        TokenType.SHOW: lambda self: self._parse_show(),
 772        TokenType.UNDROP: lambda self: self._parse_undrop(),
 773    }
 774
 775    PROPERTY_PARSERS = {
 776        **parser.Parser.PROPERTY_PARSERS,
 777        "CREDENTIALS": lambda self: self._parse_credentials_property(),
 778        "FILE_FORMAT": lambda self: self._parse_file_format_property(),
 779        "LOCATION": lambda self: self._parse_location_property(),
 780        "ROW": lambda self: (
 781            self._parse_row_access_policy()
 782            if self._match_text_seq("ACCESS", "POLICY")
 783            else self._parse_row()
 784        ),
 785        "TAG": lambda self: self._parse_tag(),
 786        "USING": lambda self: (
 787            self._match_text_seq("TEMPLATE")
 788            and self.expression(exp.UsingTemplateProperty(this=self._parse_statement()))
 789        ),
 790    }
 791
 792    DESCRIBE_QUALIFIER_PARSERS: t.ClassVar[dict[str, t.Callable]] = {
 793        "API": lambda self: self.expression(exp.ApiProperty()),
 794        "APPLICATION": lambda self: self.expression(exp.ApplicationProperty()),
 795        "CATALOG": lambda self: self.expression(exp.CatalogProperty()),
 796        "COMPUTE": lambda self: self.expression(exp.ComputeProperty()),
 797        "DATABASE": lambda self: (
 798            self.expression(exp.DatabaseProperty())
 799            if self._curr and self._curr.text.upper() == "ROLE"
 800            else None
 801        ),
 802        "DYNAMIC": lambda self: self.expression(exp.DynamicProperty()),
 803        "EXTERNAL": lambda self: self.expression(exp.ExternalProperty()),
 804        "HYBRID": lambda self: self.expression(exp.HybridProperty()),
 805        "ICEBERG": lambda self: self.expression(exp.IcebergProperty()),
 806        "MASKING": lambda self: self.expression(exp.MaskingProperty()),
 807        "MATERIALIZED": lambda self: self.expression(exp.MaterializedProperty()),
 808        "NETWORK": lambda self: self.expression(exp.NetworkProperty()),
 809        "ROW": lambda self: (
 810            self.expression(exp.RowAccessProperty()) if self._match_text_seq("ACCESS") else None
 811        ),
 812        "SECURITY": lambda self: (
 813            self.expression(exp.SecurityIntegrationProperty())
 814            if self._curr and self._curr.text.upper() == "INTEGRATION"
 815            else None
 816        ),
 817    }
 818
 819    TYPE_CONVERTERS = {
 820        # https://docs.snowflake.com/en/sql-reference/data-types-numeric#number
 821        exp.DType.DECIMAL: build_default_decimal_type(precision=38, scale=0),
 822    }
 823
 824    SHOW_PARSERS = {
 825        "DATABASES": _show_parser("DATABASES"),
 826        "SCHEMAS": _show_parser("SCHEMAS"),
 827        "OBJECTS": _show_parser("OBJECTS"),
 828        "TABLES": _show_parser("TABLES"),
 829        "VIEWS": _show_parser("VIEWS"),
 830        "PRIMARY KEYS": _show_parser("PRIMARY KEYS"),
 831        "IMPORTED KEYS": _show_parser("IMPORTED KEYS"),
 832        "UNIQUE KEYS": _show_parser("UNIQUE KEYS"),
 833        "SEQUENCES": _show_parser("SEQUENCES"),
 834        "STAGES": _show_parser("STAGES"),
 835        "COLUMNS": _show_parser("COLUMNS"),
 836        "USERS": _show_parser("USERS"),
 837        "FILE FORMATS": _show_parser("FILE FORMATS"),
 838        "FUNCTIONS": _show_parser("FUNCTIONS"),
 839        "PROCEDURES": _show_parser("PROCEDURES"),
 840        "WAREHOUSES": _show_parser("WAREHOUSES"),
 841        "ICEBERG TABLES": _show_parser("TABLES", iceberg=True),
 842        "TERSE ICEBERG TABLES": _show_parser("TABLES", terse=True, iceberg=True),
 843        "TERSE DATABASES": _show_parser("DATABASES", terse=True),
 844        "TERSE SCHEMAS": _show_parser("SCHEMAS", terse=True),
 845        "TERSE OBJECTS": _show_parser("OBJECTS", terse=True),
 846        "TERSE TABLES": _show_parser("TABLES", terse=True),
 847        "TERSE VIEWS": _show_parser("VIEWS", terse=True),
 848        "TERSE SEQUENCES": _show_parser("SEQUENCES", terse=True),
 849        "TERSE USERS": _show_parser("USERS", terse=True),
 850        # TERSE has no semantic effect for KEYS, so we do not set the terse AST arg
 851        "TERSE PRIMARY KEYS": _show_parser("PRIMARY KEYS"),
 852        "TERSE IMPORTED KEYS": _show_parser("IMPORTED KEYS"),
 853        "TERSE UNIQUE KEYS": _show_parser("UNIQUE KEYS"),
 854    }
 855
 856    SHOW_TRIE = new_trie(key.split(" ") for key in SHOW_PARSERS)
 857
 858    CONSTRAINT_PARSERS = {
 859        **parser.Parser.CONSTRAINT_PARSERS,
 860        "WITH": lambda self: self._parse_with_constraint(),
 861        "MASKING": lambda self: self._parse_with_constraint(),
 862        "PROJECTION": lambda self: self._parse_with_constraint(),
 863        "TAG": lambda self: self._parse_with_constraint(),
 864    }
 865
 866    STAGED_FILE_SINGLE_TOKENS: t.ClassVar = {
 867        TokenType.DOT,
 868        TokenType.MOD,
 869        TokenType.SLASH,
 870    }
 871
 872    FLATTEN_COLUMNS: t.ClassVar = ["SEQ", "KEY", "PATH", "INDEX", "VALUE", "THIS"]
 873
 874    SCHEMA_KINDS: t.ClassVar = {
 875        "OBJECTS",
 876        "TABLES",
 877        "VIEWS",
 878        "SEQUENCES",
 879        "UNIQUE KEYS",
 880        "IMPORTED KEYS",
 881    }
 882
 883    NON_TABLE_CREATABLES: t.ClassVar = {"STORAGE INTEGRATION", "TAG", "WAREHOUSE", "STREAMLIT"}
 884
 885    UNDROP_OBJECTS: t.ClassVar[parser.OPTIONS_TYPE] = {
 886        **dict.fromkeys(
 887            (
 888                "ACCOUNT",
 889                "DATABASE",
 890                "NOTEBOOK",
 891                "SCHEMA",
 892                "SNAPSHOT",
 893                "STREAMLIT",
 894                "TABLE",
 895                "TAG",
 896                "TYPE",
 897            ),
 898            tuple(),
 899        ),
 900        "DYNAMIC": ("TABLE",),
 901        "EXTERNAL": ("VOLUME",),
 902        "ICEBERG": ("TABLE",),
 903    }
 904
 905    CREATABLES = {
 906        *parser.Parser.CREATABLES,
 907        TokenType.INTEGRATION,
 908        TokenType.PACKAGE,
 909        TokenType.POLICY,
 910        TokenType.POOL,
 911        TokenType.ROLE,
 912        TokenType.RULE,
 913        TokenType.VOLUME,
 914    }
 915
 916    LAMBDAS = {
 917        **parser.Parser.LAMBDAS,
 918        TokenType.ARROW: lambda self, expressions: self.expression(
 919            exp.Lambda(
 920                this=self._replace_lambda(
 921                    self._parse_assignment(),
 922                    expressions,
 923                ),
 924                expressions=[e.this if isinstance(e, exp.Cast) else e for e in expressions],
 925            )
 926        ),
 927    }
 928
 929    COLUMN_OPERATORS = {
 930        **parser.Parser.COLUMN_OPERATORS,
 931        TokenType.EXCLAMATION: lambda self, this, attr: self.expression(
 932            exp.ModelAttribute(this=this, expression=attr)
 933        ),
 934    }
 935
 936    def _parse_directory(self) -> exp.DirectoryStage:
 937        table = self._parse_table_parts()
 938        this = table.this if isinstance(table, exp.Table) else table
 939        return self.expression(exp.DirectoryStage(this=this))
 940
 941    def _parse_describe(self) -> exp.Describe:
 942        index = self._index
 943
 944        if self._match_texts(self.DESCRIBE_QUALIFIER_PARSERS):
 945            qualifier = self.DESCRIBE_QUALIFIER_PARSERS[self._prev.text.upper()](self)
 946
 947            if qualifier:
 948                kind = self._match_set(self.CREATABLES) and self._prev.text.upper()
 949
 950                if kind:
 951                    this = self._parse_table(schema=True)
 952                    properties = self.expression(exp.Properties(expressions=[qualifier]))
 953                    post_props = self._parse_properties()
 954                    expressions = post_props.expressions if post_props else None
 955                    return self.expression(
 956                        exp.Describe(
 957                            this=this,
 958                            kind=kind,
 959                            properties=properties,
 960                            expressions=expressions,
 961                        )
 962                    )
 963
 964        self._retreat(index)
 965        return super()._parse_describe()
 966
 967    def _parse_use(self) -> exp.Use:
 968        if self._match_text_seq("SECONDARY", "ROLES"):
 969            this = self._match_texts(("ALL", "NONE")) and exp.var(self._prev.text.upper())
 970            roles = None if this else self._parse_csv(lambda: self._parse_table(schema=False))
 971            return self.expression(exp.Use(kind="SECONDARY ROLES", this=this, expressions=roles))
 972
 973        return super()._parse_use()
 974
 975    def _negate_range(self, this: exp.Expr | None = None) -> exp.Expr | None:
 976        if not this:
 977            return this
 978
 979        query = this.args.get("query")
 980        if isinstance(this, exp.In) and isinstance(query, exp.Query):
 981            # Snowflake treats `value NOT IN (subquery)` as `VALUE <> ALL (subquery)`, so
 982            # we do this conversion here to avoid parsing it into `NOT value IN (subquery)`
 983            # which can produce different results (most likely a SnowFlake bug).
 984            #
 985            # https://docs.snowflake.com/en/sql-reference/functions/in
 986            # Context: https://github.com/tobymao/sqlglot/issues/3890
 987            return self.expression(exp.NEQ(this=this.this, expression=exp.All(this=query.unnest())))
 988
 989        return self.expression(exp.Not(this=this))
 990
 991    def _parse_tag(self) -> exp.Tags:
 992        return self.expression(exp.Tags(expressions=self._parse_wrapped_csv(self._parse_property)))
 993
 994    def _parse_property_before(self) -> exp.Expr | list[exp.Expr] | None:
 995        prop = super()._parse_property_before()
 996        if prop:
 997            return prop
 998
 999        if not self._next or self._next.token_type != TokenType.EQ:
1000            return None
1001
1002        return self._parse_sequence_properties() or self._parse_key_value_property(
1003            self._parse_primary_or_var
1004        )
1005
1006    def _parse_with_constraint(self) -> exp.Expr | None:
1007        if self._prev.token_type != TokenType.WITH:
1008            self._retreat(self._index - 1)
1009
1010        if self._match_text_seq("MASKING", "POLICY"):
1011            policy = self._parse_column()
1012            return self.expression(
1013                exp.MaskingPolicyColumnConstraint(
1014                    this=policy.to_dot() if isinstance(policy, exp.Column) else policy,
1015                    expressions=self._match(TokenType.USING)
1016                    and self._parse_wrapped_csv(self._parse_id_var),
1017                )
1018            )
1019        if self._match_text_seq("PROJECTION", "POLICY"):
1020            policy = self._parse_column()
1021            return self.expression(
1022                exp.ProjectionPolicyColumnConstraint(
1023                    this=policy.to_dot() if isinstance(policy, exp.Column) else policy
1024                )
1025            )
1026        if self._match(TokenType.TAG):
1027            return self._parse_tag()
1028
1029        return None
1030
1031    def _parse_with_property(self) -> exp.Expr | None | list[exp.Expr]:
1032        if self._match(TokenType.TAG):
1033            return self._parse_tag()
1034
1035        if self._match_text_seq("ROW", "ACCESS", "POLICY"):
1036            return self._parse_row_access_policy()
1037
1038        return super()._parse_with_property()
1039
1040    def _parse_row_access_policy(self) -> exp.RowAccessProperty:
1041        # GET_DDL outputs #unknown_policy when the user lacks privileges to see the policy name
1042        if self._match(TokenType.HASH):
1043            policy: exp.Expr | None = self._parse_var(any_token=True)
1044            if policy:
1045                policy = exp.Var(this=f"#{policy.name}")
1046            expressions = None
1047        else:
1048            policy = self._parse_column()
1049            if isinstance(policy, exp.Column):
1050                policy = policy.to_dot()
1051            if not self._match(TokenType.ON):
1052                self.raise_error("Expected ON after ROW ACCESS POLICY name")
1053            expressions = self._parse_wrapped_csv(self._parse_id_var)
1054
1055        return self.expression(exp.RowAccessProperty(this=policy, expressions=expressions))
1056
1057    def _parse_create(self) -> exp.Create | exp.Command:
1058        expression = super()._parse_create()
1059        if isinstance(expression, exp.Create) and expression.kind in self.NON_TABLE_CREATABLES:
1060            # Replace the Table node with the enclosed Identifier
1061            expression.this.replace(expression.this.this)
1062
1063        return expression
1064
1065    def _parse_object_construct_keep_null(self) -> exp.JSONObject:
1066        index = self._index
1067        column = self._parse_column()
1068        if column and column.is_star:
1069            return self.expression(exp.JSONObject(expressions=[column]))
1070
1071        self._retreat(index)
1072        return self._parse_json_object(agg=False)
1073
1074    # https://docs.snowflake.com/en/sql-reference/functions/date_part.html
1075    # https://docs.snowflake.com/en/sql-reference/functions-date-time.html#label-supported-date-time-parts
1076    def _parse_date_part(self) -> exp.Expr | None:
1077        this = self._parse_var() or self._parse_type()
1078
1079        if not this:
1080            return None
1081
1082        # Handle both syntaxes: DATE_PART(part, expr) and DATE_PART(part FROM expr)
1083        expression = self._match_set((TokenType.FROM, TokenType.COMMA)) and self._parse_bitwise()
1084        return self.expression(
1085            exp.Extract(this=map_date_part(this, self.dialect), expression=expression)
1086        )
1087
1088    def _parse_bracket_key_value(self, is_map: bool = False) -> exp.Expr | None:
1089        if is_map:
1090            # Keys are strings in Snowflake's objects, see also:
1091            # - https://docs.snowflake.com/en/sql-reference/data-types-semistructured
1092            # - https://docs.snowflake.com/en/sql-reference/functions/object_construct
1093            return self._parse_slice(self._parse_string()) or self._parse_assignment()
1094
1095        return self._parse_slice(self._parse_alias(self._parse_assignment(), explicit=True))
1096
1097    def _parse_lateral(self) -> exp.Lateral | None:
1098        lateral = super()._parse_lateral()
1099        if not lateral:
1100            return lateral
1101
1102        if isinstance(lateral.this, exp.Explode):
1103            table_alias = lateral.args.get("alias")
1104            columns = [exp.to_identifier(col) for col in self.FLATTEN_COLUMNS]
1105            if table_alias and not table_alias.args.get("columns"):
1106                table_alias.set("columns", columns)
1107            elif not table_alias:
1108                exp.alias_(lateral, "_flattened", table=columns, copy=False)
1109
1110        return lateral
1111
1112    def _parse_table_parts(
1113        self,
1114        schema: bool = False,
1115        is_db_reference: bool = False,
1116        wildcard: bool = False,
1117        fast: bool = False,
1118    ) -> exp.Table | exp.Dot | None:
1119        # https://docs.snowflake.com/en/user-guide/querying-stage
1120        if self._match(TokenType.STRING, advance=False):
1121            table = self._parse_string()
1122        elif self._match_text_seq("@", advance=False):
1123            table = self._parse_location_path()
1124        else:
1125            table = None
1126
1127        if table:
1128            file_format = None
1129            pattern = None
1130
1131            wrapped = self._match(TokenType.L_PAREN)
1132            while self._curr and wrapped and not self._match(TokenType.R_PAREN):
1133                if self._match_text_seq("FILE_FORMAT", "=>"):
1134                    file_format = self._parse_string() or super()._parse_table_parts(
1135                        is_db_reference=is_db_reference
1136                    )
1137                elif self._match_text_seq("PATTERN", "=>"):
1138                    pattern = self._parse_string()
1139                else:
1140                    break
1141
1142                self._match(TokenType.COMMA)
1143
1144            table = self.expression(exp.Table(this=table, format=file_format, pattern=pattern))
1145        else:
1146            table = super()._parse_table_parts(
1147                schema=schema,
1148                is_db_reference=is_db_reference,
1149                fast=fast,
1150            )
1151
1152        return table
1153
1154    def _parse_function_call(
1155        self,
1156        functions: dict[str, t.Callable] | None = None,
1157        anonymous: bool = False,
1158        optional_parens: bool = True,
1159        any_token: bool = False,
1160    ) -> exp.Expr | None:
1161        this = super()._parse_function_call(
1162            functions=functions,
1163            anonymous=anonymous,
1164            optional_parens=optional_parens,
1165            any_token=any_token,
1166        )
1167
1168        # Snowflake can invoke a function whose name is dynamically resolved, e.g.
1169        # `IDENTIFIER('my_func')(1, 2)`. The trailing argument list is the call's arguments.
1170        #
1171        # https://docs.snowflake.com/en/sql-reference/identifier-literal
1172        if isinstance(this, exp.DynamicIdentifier) and self._match(
1173            TokenType.L_PAREN, advance=False
1174        ):
1175            this.set("expressions", self._parse_wrapped_csv(self._parse_lambda))
1176
1177        return this
1178
1179    def _parse_id_var(
1180        self,
1181        any_token: bool = True,
1182        tokens: Collection[TokenType] | None = None,
1183    ) -> exp.Expr | None:
1184        if self._match_text_seq("IDENTIFIER", "("):
1185            identifier = (
1186                super()._parse_id_var(any_token=any_token, tokens=tokens) or self._parse_string()
1187            )
1188            self._match_r_paren()
1189            return self.expression(exp.DynamicIdentifier(this=identifier))
1190
1191        return super()._parse_id_var(any_token=any_token, tokens=tokens)
1192
1193    def _parse_show_snowflake(
1194        self, this: str, terse: bool = False, iceberg: bool = False
1195    ) -> exp.Show:
1196        scope = None
1197        scope_kind = None
1198
1199        history = self._match_text_seq("HISTORY")
1200
1201        like = self._parse_string() if self._match(TokenType.LIKE) else None
1202
1203        if self._match(TokenType.IN):
1204            if self._match_text_seq("ACCOUNT"):
1205                scope_kind = "ACCOUNT"
1206            elif self._match_text_seq("CLASS"):
1207                scope_kind = "CLASS"
1208                scope = self._parse_table_parts()
1209            elif self._match_text_seq("APPLICATION"):
1210                scope_kind = "APPLICATION"
1211                if self._match_text_seq("PACKAGE"):
1212                    scope_kind += " PACKAGE"
1213                scope = self._parse_table_parts()
1214            elif self._match_set(self.DB_CREATABLES):
1215                scope_kind = self._prev.text.upper()
1216                if self._curr:
1217                    scope = self._parse_table_parts()
1218            elif self._curr:
1219                scope_kind = "SCHEMA" if this in self.SCHEMA_KINDS else "TABLE"
1220                scope = self._parse_table_parts()
1221
1222        return self.expression(
1223            exp.Show(
1224                terse=terse,
1225                iceberg=iceberg,
1226                this=this,
1227                history=history,
1228                like=like,
1229                scope=scope,
1230                scope_kind=scope_kind,
1231                starts_with=self._match_text_seq("STARTS", "WITH") and self._parse_string(),
1232                limit=self._parse_limit(),
1233                from_=self._parse_string() if self._match(TokenType.FROM) else None,
1234                privileges=self._match_text_seq("WITH", "PRIVILEGES")
1235                and self._parse_csv(lambda: self._parse_var(any_token=True, upper=True)),
1236            )
1237        )
1238
1239    def _parse_undrop(self) -> exp.Undrop | exp.Command:
1240        start = self._prev
1241        kind = self._parse_var_from_options(self.UNDROP_OBJECTS, raise_unmatched=False)
1242        if not kind:
1243            return self._parse_as_command(start)
1244
1245        this = self._parse_table_parts(
1246            is_db_reference=kind.name in ("ACCOUNT", "DATABASE", "SCHEMA")
1247        )
1248        rename = self._parse_table_parts() if self._match_text_seq("RENAME", "TO") else None
1249        return self.expression(exp.Undrop(this=this, kind=kind.name, rename=rename))
1250
1251    def _parse_put(self) -> exp.Put | exp.Command:
1252        if self._curr.token_type not in (TokenType.STRING, TokenType.URI_START):
1253            return self._parse_as_command(self._prev)
1254
1255        return self.expression(
1256            exp.Put(
1257                this=self._parse_string() or exp.Literal.string(self._parse_location_path().name),
1258                target=self._parse_location_path(),
1259                properties=self._parse_properties(),
1260            )
1261        )
1262
1263    def _parse_get(self) -> exp.Expr | None:
1264        start = self._prev
1265
1266        # If we detect GET( then we need to parse a function, not a statement
1267        if self._match(TokenType.L_PAREN):
1268            self._retreat(self._index - 2)
1269            return self._parse_expression()
1270
1271        target = self._parse_location_path()
1272
1273        # Parse as command if unquoted file path
1274        if self._curr.token_type == TokenType.URI_START:
1275            return self._parse_as_command(start)
1276
1277        return self.expression(
1278            exp.Get(this=self._parse_string(), target=target, properties=self._parse_properties())
1279        )
1280
1281    def _parse_location_property(self) -> exp.LocationProperty:
1282        self._match(TokenType.EQ)
1283        return self.expression(exp.LocationProperty(this=self._parse_location_path()))
1284
1285    def _parse_file_location(self) -> exp.Expr | None:
1286        # Parse either a subquery or a staged file
1287        return (
1288            self._parse_select(table=True, parse_subquery_alias=False)
1289            if self._match(TokenType.L_PAREN, advance=False)
1290            else self._parse_table_parts()
1291        )
1292
1293    def _parse_location_path(self) -> exp.Var:
1294        start = self._curr
1295        self._advance_any(ignore_reserved=True)
1296
1297        # We avoid consuming a comma token because external tables like @foo and @bar
1298        # can be joined in a query with a comma separator, as well as closing paren
1299        # in case of subqueries
1300        while self._is_connected() and not self._match_set(
1301            (TokenType.COMMA, TokenType.L_PAREN, TokenType.R_PAREN), advance=False
1302        ):
1303            self._advance_any(ignore_reserved=True)
1304
1305        return exp.var(self._find_sql(start, self._prev))
1306
1307    def _parse_lambda_arg(self) -> exp.Expr | None:
1308        this = super()._parse_lambda_arg()
1309
1310        if not this:
1311            return this
1312
1313        typ = self._parse_types()
1314
1315        if typ:
1316            return self.expression(exp.Cast(this=this, to=typ))
1317
1318        return this
1319
1320    def _parse_foreign_key(self) -> exp.ForeignKey:
1321        # inlineFK, the REFERENCES columns are implied
1322        if self._match(TokenType.REFERENCES, advance=False):
1323            return self.expression(exp.ForeignKey())
1324
1325        # outoflineFK, explicitly names the columns
1326        return super()._parse_foreign_key()
1327
1328    def _parse_file_format_property(self) -> exp.FileFormatProperty:
1329        self._match(TokenType.EQ)
1330        if self._match(TokenType.L_PAREN, advance=False):
1331            expressions = self._parse_wrapped_options()
1332        else:
1333            expressions = [self._parse_format_name()]
1334
1335        return self.expression(exp.FileFormatProperty(expressions=expressions))
1336
1337    def _parse_credentials_property(self) -> exp.CredentialsProperty:
1338        return self.expression(exp.CredentialsProperty(expressions=self._parse_wrapped_options()))
1339
1340    def _parse_semantic_view(self) -> exp.SemanticView:
1341        kwargs: dict[str, t.Any] = {"this": self._parse_table_parts()}
1342
1343        while self._curr and not self._match(TokenType.R_PAREN, advance=False):
1344            if self._match_texts(("DIMENSIONS", "METRICS", "FACTS")):
1345                keyword = self._prev.text.lower()
1346                kwargs[keyword] = self._parse_csv(
1347                    lambda: self._parse_alias(self._parse_disjunction(), explicit=True)
1348                )
1349            elif self._match_text_seq("WHERE"):
1350                kwargs["where"] = self._parse_expression()
1351            else:
1352                self.raise_error("Expecting ) or encountered unexpected keyword")
1353                break
1354
1355        return self.expression(exp.SemanticView(**kwargs))
1356
1357    def _parse_set(self, unset: bool = False, tag: bool = False) -> exp.Set | exp.Command:
1358        set = super()._parse_set(unset=unset, tag=tag)
1359
1360        if isinstance(set, exp.Set):
1361            for expr in set.expressions:
1362                if isinstance(expr, exp.SetItem):
1363                    expr.set("kind", "VARIABLE")
1364        return set
1365
1366    def _parse_position(self, haystack_first: bool = False) -> exp.StrPosition:
1367        result = super()._parse_position(haystack_first)
1368        result.set("clamp_position", True)
1369        return result
1370
1371    def _parse_substring(self) -> exp.Substring:
1372        result = super()._parse_substring()
1373        result.set("zero_start", True)
1374        return result
1375
1376    def build_cast(self, strict: bool, **kwargs) -> exp.Expr:
1377        to = kwargs.get("to")
1378        if not strict and to and to.this == exp.DataType.Type.BOOLEAN:
1379            return self.expression(exp.ToBoolean(this=kwargs.get("this"), safe=True))
1380        cast = super().build_cast(strict, **kwargs)
1381        if isinstance(cast, exp.TryCast) and to:
1382            if to.this in exp.DataType.TEXT_TYPES and to.expressions:
1383                cast.set("null_on_text_overflow", True)
1384            elif to.this == exp.DType.DATE and cast.this.is_string:
1385                cast.set("probe_date_format", True)
1386        return cast
1387
1388    def _parse_window(self, this: exp.Expr | None, alias: bool = False) -> exp.Expr | None:
1389        result = super()._parse_window(this, alias)
1390
1391        # Set default window frame for ranking functions if not present
1392        if (
1393            isinstance(result, exp.Window)
1394            and isinstance(this, RANKING_WINDOW_FUNCTIONS_WITH_FRAME)
1395            and not result.args.get("spec")
1396        ):
1397            frame = exp.WindowSpec(
1398                kind="ROWS",
1399                start="UNBOUNDED",
1400                start_side="PRECEDING",
1401                end="UNBOUNDED",
1402                end_side="FOLLOWING",
1403            )
1404            result.set("spec", frame)
1405        return result
1406
1407
1408# This is imported and used by both the parser (above) and the generator in the dialect file
1409RANKING_WINDOW_FUNCTIONS_WITH_FRAME = (
1410    exp.FirstValue,
1411    exp.LastValue,
1412    exp.NthValue,
1413)
1414
1415
1416def build_object_construct(args: list) -> exp.StarMap | exp.Struct:
1417    expression = parser.build_var_map(args)
1418
1419    if isinstance(expression, exp.StarMap):
1420        return expression
1421
1422    return exp.Struct(
1423        expressions=[
1424            exp.PropertyEQ(this=k, expression=v) for k, v in zip(expression.keys, expression.values)
1425        ]
1426    )