sqlglot.dialects.snowflake

   1from __future__ import annotations
   2
   3import typing as t
   4
   5from sqlglot import exp, generator, jsonpath, parser, tokens, transforms
   6from sqlglot.dialects.dialect import (
   7    Dialect,
   8    NormalizationStrategy,
   9    build_timetostr_or_tochar,
  10    binary_from_function,
  11    build_default_decimal_type,
  12    build_timestamp_from_parts,
  13    date_delta_sql,
  14    date_trunc_to_time,
  15    datestrtodate_sql,
  16    build_formatted_time,
  17    if_sql,
  18    inline_array_sql,
  19    max_or_greatest,
  20    min_or_least,
  21    rename_func,
  22    timestamptrunc_sql,
  23    timestrtotime_sql,
  24    var_map_sql,
  25    map_date_part,
  26    no_timestamp_sql,
  27    strposition_sql,
  28    timestampdiff_sql,
  29    no_make_interval_sql,
  30    groupconcat_sql,
  31)
  32from sqlglot.generator import unsupported_args
  33from sqlglot.helper import flatten, is_float, is_int, seq_get
  34from sqlglot.tokens import TokenType
  35
  36if t.TYPE_CHECKING:
  37    from sqlglot._typing import E, B
  38
  39
  40# from https://docs.snowflake.com/en/sql-reference/functions/to_timestamp.html
  41def _build_datetime(
  42    name: str, kind: exp.DataType.Type, safe: bool = False
  43) -> t.Callable[[t.List], exp.Func]:
  44    def _builder(args: t.List) -> exp.Func:
  45        value = seq_get(args, 0)
  46        scale_or_fmt = seq_get(args, 1)
  47
  48        int_value = value is not None and is_int(value.name)
  49        int_scale_or_fmt = scale_or_fmt is not None and scale_or_fmt.is_int
  50
  51        if isinstance(value, exp.Literal) or (value and scale_or_fmt):
  52            # Converts calls like `TO_TIME('01:02:03')` into casts
  53            if len(args) == 1 and value.is_string and not int_value:
  54                return (
  55                    exp.TryCast(this=value, to=exp.DataType.build(kind))
  56                    if safe
  57                    else exp.cast(value, kind)
  58                )
  59
  60            # Handles `TO_TIMESTAMP(str, fmt)` and `TO_TIMESTAMP(num, scale)` as special
  61            # cases so we can transpile them, since they're relatively common
  62            if kind == exp.DataType.Type.TIMESTAMP:
  63                if not safe and (int_value or int_scale_or_fmt):
  64                    # TRY_TO_TIMESTAMP('integer') is not parsed into exp.UnixToTime as
  65                    # it's not easily transpilable
  66                    return exp.UnixToTime(this=value, scale=scale_or_fmt)
  67                if not int_scale_or_fmt and not is_float(value.name):
  68                    expr = build_formatted_time(exp.StrToTime, "snowflake")(args)
  69                    expr.set("safe", safe)
  70                    return expr
  71
  72        if kind in (exp.DataType.Type.DATE, exp.DataType.Type.TIME) and not int_value:
  73            klass = exp.TsOrDsToDate if kind == exp.DataType.Type.DATE else exp.TsOrDsToTime
  74            formatted_exp = build_formatted_time(klass, "snowflake")(args)
  75            formatted_exp.set("safe", safe)
  76            return formatted_exp
  77
  78        return exp.Anonymous(this=name, expressions=args)
  79
  80    return _builder
  81
  82
  83def _build_object_construct(args: t.List) -> t.Union[exp.StarMap, exp.Struct]:
  84    expression = parser.build_var_map(args)
  85
  86    if isinstance(expression, exp.StarMap):
  87        return expression
  88
  89    return exp.Struct(
  90        expressions=[
  91            exp.PropertyEQ(this=k, expression=v) for k, v in zip(expression.keys, expression.values)
  92        ]
  93    )
  94
  95
  96def _build_datediff(args: t.List) -> exp.DateDiff:
  97    return exp.DateDiff(
  98        this=seq_get(args, 2), expression=seq_get(args, 1), unit=map_date_part(seq_get(args, 0))
  99    )
 100
 101
 102def _build_date_time_add(expr_type: t.Type[E]) -> t.Callable[[t.List], E]:
 103    def _builder(args: t.List) -> E:
 104        return expr_type(
 105            this=seq_get(args, 2),
 106            expression=seq_get(args, 1),
 107            unit=map_date_part(seq_get(args, 0)),
 108        )
 109
 110    return _builder
 111
 112
 113def _build_bitwise(expr_type: t.Type[B], name: str) -> t.Callable[[t.List], B | exp.Anonymous]:
 114    def _builder(args: t.List) -> B | exp.Anonymous:
 115        if len(args) == 3:
 116            return exp.Anonymous(this=name, expressions=args)
 117
 118        return binary_from_function(expr_type)(args)
 119
 120    return _builder
 121
 122
 123# https://docs.snowflake.com/en/sql-reference/functions/div0
 124def _build_if_from_div0(args: t.List) -> exp.If:
 125    lhs = exp._wrap(seq_get(args, 0), exp.Binary)
 126    rhs = exp._wrap(seq_get(args, 1), exp.Binary)
 127
 128    cond = exp.EQ(this=rhs, expression=exp.Literal.number(0)).and_(
 129        exp.Is(this=lhs, expression=exp.null()).not_()
 130    )
 131    true = exp.Literal.number(0)
 132    false = exp.Div(this=lhs, expression=rhs)
 133    return exp.If(this=cond, true=true, false=false)
 134
 135
 136# https://docs.snowflake.com/en/sql-reference/functions/zeroifnull
 137def _build_if_from_zeroifnull(args: t.List) -> exp.If:
 138    cond = exp.Is(this=seq_get(args, 0), expression=exp.Null())
 139    return exp.If(this=cond, true=exp.Literal.number(0), false=seq_get(args, 0))
 140
 141
 142# https://docs.snowflake.com/en/sql-reference/functions/zeroifnull
 143def _build_if_from_nullifzero(args: t.List) -> exp.If:
 144    cond = exp.EQ(this=seq_get(args, 0), expression=exp.Literal.number(0))
 145    return exp.If(this=cond, true=exp.Null(), false=seq_get(args, 0))
 146
 147
 148def _regexpilike_sql(self: Snowflake.Generator, expression: exp.RegexpILike) -> str:
 149    flag = expression.text("flag")
 150
 151    if "i" not in flag:
 152        flag += "i"
 153
 154    return self.func(
 155        "REGEXP_LIKE", expression.this, expression.expression, exp.Literal.string(flag)
 156    )
 157
 158
 159def _build_regexp_replace(args: t.List) -> exp.RegexpReplace:
 160    regexp_replace = exp.RegexpReplace.from_arg_list(args)
 161
 162    if not regexp_replace.args.get("replacement"):
 163        regexp_replace.set("replacement", exp.Literal.string(""))
 164
 165    return regexp_replace
 166
 167
 168def _show_parser(*args: t.Any, **kwargs: t.Any) -> t.Callable[[Snowflake.Parser], exp.Show]:
 169    def _parse(self: Snowflake.Parser) -> exp.Show:
 170        return self._parse_show_snowflake(*args, **kwargs)
 171
 172    return _parse
 173
 174
 175def _date_trunc_to_time(args: t.List) -> exp.DateTrunc | exp.TimestampTrunc:
 176    trunc = date_trunc_to_time(args)
 177    trunc.set("unit", map_date_part(trunc.args["unit"]))
 178    return trunc
 179
 180
 181def _unqualify_pivot_columns(expression: exp.Expression) -> exp.Expression:
 182    """
 183    Snowflake doesn't allow columns referenced in UNPIVOT to be qualified,
 184    so we need to unqualify them. Same goes for ANY ORDER BY <column>.
 185
 186    Example:
 187        >>> from sqlglot import parse_one
 188        >>> expr = parse_one("SELECT * FROM m_sales UNPIVOT(sales FOR month IN (m_sales.jan, feb, mar, april))")
 189        >>> print(_unqualify_pivot_columns(expr).sql(dialect="snowflake"))
 190        SELECT * FROM m_sales UNPIVOT(sales FOR month IN (jan, feb, mar, april))
 191    """
 192    if isinstance(expression, exp.Pivot):
 193        if expression.unpivot:
 194            expression = transforms.unqualify_columns(expression)
 195        else:
 196            for field in expression.fields:
 197                field_expr = seq_get(field.expressions if field else [], 0)
 198
 199                if isinstance(field_expr, exp.PivotAny):
 200                    unqualified_field_expr = transforms.unqualify_columns(field_expr)
 201                    t.cast(exp.Expression, field).set("expressions", unqualified_field_expr, 0)
 202
 203    return expression
 204
 205
 206def _flatten_structured_types_unless_iceberg(expression: exp.Expression) -> exp.Expression:
 207    assert isinstance(expression, exp.Create)
 208
 209    def _flatten_structured_type(expression: exp.DataType) -> exp.DataType:
 210        if expression.this in exp.DataType.NESTED_TYPES:
 211            expression.set("expressions", None)
 212        return expression
 213
 214    props = expression.args.get("properties")
 215    if isinstance(expression.this, exp.Schema) and not (props and props.find(exp.IcebergProperty)):
 216        for schema_expression in expression.this.expressions:
 217            if isinstance(schema_expression, exp.ColumnDef):
 218                column_type = schema_expression.kind
 219                if isinstance(column_type, exp.DataType):
 220                    column_type.transform(_flatten_structured_type, copy=False)
 221
 222    return expression
 223
 224
 225def _unnest_generate_date_array(unnest: exp.Unnest) -> None:
 226    generate_date_array = unnest.expressions[0]
 227    start = generate_date_array.args.get("start")
 228    end = generate_date_array.args.get("end")
 229    step = generate_date_array.args.get("step")
 230
 231    if not start or not end or not isinstance(step, exp.Interval) or step.name != "1":
 232        return
 233
 234    unit = step.args.get("unit")
 235
 236    unnest_alias = unnest.args.get("alias")
 237    if unnest_alias:
 238        unnest_alias = unnest_alias.copy()
 239        sequence_value_name = seq_get(unnest_alias.columns, 0) or "value"
 240    else:
 241        sequence_value_name = "value"
 242
 243    # We'll add the next sequence value to the starting date and project the result
 244    date_add = _build_date_time_add(exp.DateAdd)(
 245        [unit, exp.cast(sequence_value_name, "int"), exp.cast(start, "date")]
 246    ).as_(sequence_value_name)
 247
 248    # We use DATEDIFF to compute the number of sequence values needed
 249    number_sequence = Snowflake.Parser.FUNCTIONS["ARRAY_GENERATE_RANGE"](
 250        [exp.Literal.number(0), _build_datediff([unit, start, end]) + 1]
 251    )
 252
 253    unnest.set("expressions", [number_sequence])
 254    unnest.replace(exp.select(date_add).from_(unnest.copy()).subquery(unnest_alias))
 255
 256
 257def _transform_generate_date_array(expression: exp.Expression) -> exp.Expression:
 258    if isinstance(expression, exp.Select):
 259        for generate_date_array in expression.find_all(exp.GenerateDateArray):
 260            parent = generate_date_array.parent
 261
 262            # If GENERATE_DATE_ARRAY is used directly as an array (e.g passed into ARRAY_LENGTH), the transformed Snowflake
 263            # query is the following (it'll be unnested properly on the next iteration due to copy):
 264            # SELECT ref(GENERATE_DATE_ARRAY(...)) -> SELECT ref((SELECT ARRAY_AGG(*) FROM UNNEST(GENERATE_DATE_ARRAY(...))))
 265            if not isinstance(parent, exp.Unnest):
 266                unnest = exp.Unnest(expressions=[generate_date_array.copy()])
 267                generate_date_array.replace(
 268                    exp.select(exp.ArrayAgg(this=exp.Star())).from_(unnest).subquery()
 269                )
 270
 271            if (
 272                isinstance(parent, exp.Unnest)
 273                and isinstance(parent.parent, (exp.From, exp.Join))
 274                and len(parent.expressions) == 1
 275            ):
 276                _unnest_generate_date_array(parent)
 277
 278    return expression
 279
 280
 281def _build_regexp_extract(expr_type: t.Type[E]) -> t.Callable[[t.List], E]:
 282    def _builder(args: t.List) -> E:
 283        return expr_type(
 284            this=seq_get(args, 0),
 285            expression=seq_get(args, 1),
 286            position=seq_get(args, 2),
 287            occurrence=seq_get(args, 3),
 288            parameters=seq_get(args, 4),
 289            group=seq_get(args, 5) or exp.Literal.number(0),
 290        )
 291
 292    return _builder
 293
 294
 295def _regexpextract_sql(self, expression: exp.RegexpExtract | exp.RegexpExtractAll) -> str:
 296    # Other dialects don't support all of the following parameters, so we need to
 297    # generate default values as necessary to ensure the transpilation is correct
 298    group = expression.args.get("group")
 299
 300    # To avoid generating all these default values, we set group to None if
 301    # it's 0 (also default value) which doesn't trigger the following chain
 302    if group and group.name == "0":
 303        group = None
 304
 305    parameters = expression.args.get("parameters") or (group and exp.Literal.string("c"))
 306    occurrence = expression.args.get("occurrence") or (parameters and exp.Literal.number(1))
 307    position = expression.args.get("position") or (occurrence and exp.Literal.number(1))
 308
 309    return self.func(
 310        "REGEXP_SUBSTR" if isinstance(expression, exp.RegexpExtract) else "REGEXP_EXTRACT_ALL",
 311        expression.this,
 312        expression.expression,
 313        position,
 314        occurrence,
 315        parameters,
 316        group,
 317    )
 318
 319
 320def _json_extract_value_array_sql(
 321    self: Snowflake.Generator, expression: exp.JSONValueArray | exp.JSONExtractArray
 322) -> str:
 323    json_extract = exp.JSONExtract(this=expression.this, expression=expression.expression)
 324    ident = exp.to_identifier("x")
 325
 326    if isinstance(expression, exp.JSONValueArray):
 327        this: exp.Expression = exp.cast(ident, to=exp.DataType.Type.VARCHAR)
 328    else:
 329        this = exp.ParseJSON(this=f"TO_JSON({ident})")
 330
 331    transform_lambda = exp.Lambda(expressions=[ident], this=this)
 332
 333    return self.func("TRANSFORM", json_extract, transform_lambda)
 334
 335
 336class Snowflake(Dialect):
 337    # https://docs.snowflake.com/en/sql-reference/identifiers-syntax
 338    NORMALIZATION_STRATEGY = NormalizationStrategy.UPPERCASE
 339    NULL_ORDERING = "nulls_are_large"
 340    TIME_FORMAT = "'YYYY-MM-DD HH24:MI:SS'"
 341    SUPPORTS_USER_DEFINED_TYPES = False
 342    SUPPORTS_SEMI_ANTI_JOIN = False
 343    PREFER_CTE_ALIAS_COLUMN = True
 344    TABLESAMPLE_SIZE_IS_PERCENT = True
 345    COPY_PARAMS_ARE_CSV = False
 346    ARRAY_AGG_INCLUDES_NULLS = None
 347
 348    TIME_MAPPING = {
 349        "YYYY": "%Y",
 350        "yyyy": "%Y",
 351        "YY": "%y",
 352        "yy": "%y",
 353        "MMMM": "%B",
 354        "mmmm": "%B",
 355        "MON": "%b",
 356        "mon": "%b",
 357        "MM": "%m",
 358        "mm": "%m",
 359        "DD": "%d",
 360        "dd": "%-d",
 361        "DY": "%a",
 362        "dy": "%w",
 363        "HH24": "%H",
 364        "hh24": "%H",
 365        "HH12": "%I",
 366        "hh12": "%I",
 367        "MI": "%M",
 368        "mi": "%M",
 369        "SS": "%S",
 370        "ss": "%S",
 371        "FF6": "%f",
 372        "ff6": "%f",
 373    }
 374
 375    def quote_identifier(self, expression: E, identify: bool = True) -> E:
 376        # This disables quoting DUAL in SELECT ... FROM DUAL, because Snowflake treats an
 377        # unquoted DUAL keyword in a special way and does not map it to a user-defined table
 378        if (
 379            isinstance(expression, exp.Identifier)
 380            and isinstance(expression.parent, exp.Table)
 381            and expression.name.lower() == "dual"
 382        ):
 383            return expression  # type: ignore
 384
 385        return super().quote_identifier(expression, identify=identify)
 386
 387    class JSONPathTokenizer(jsonpath.JSONPathTokenizer):
 388        SINGLE_TOKENS = jsonpath.JSONPathTokenizer.SINGLE_TOKENS.copy()
 389        SINGLE_TOKENS.pop("$")
 390
 391    class Parser(parser.Parser):
 392        IDENTIFY_PIVOT_STRINGS = True
 393        DEFAULT_SAMPLING_METHOD = "BERNOULLI"
 394        COLON_IS_VARIANT_EXTRACT = True
 395
 396        ID_VAR_TOKENS = {
 397            *parser.Parser.ID_VAR_TOKENS,
 398            TokenType.MATCH_CONDITION,
 399        }
 400
 401        TABLE_ALIAS_TOKENS = parser.Parser.TABLE_ALIAS_TOKENS | {TokenType.WINDOW}
 402        TABLE_ALIAS_TOKENS.discard(TokenType.MATCH_CONDITION)
 403
 404        COLON_PLACEHOLDER_TOKENS = ID_VAR_TOKENS | {TokenType.NUMBER}
 405
 406        FUNCTIONS = {
 407            **parser.Parser.FUNCTIONS,
 408            "APPROX_PERCENTILE": exp.ApproxQuantile.from_arg_list,
 409            "ARRAY_CONSTRUCT": lambda args: exp.Array(expressions=args),
 410            "ARRAY_CONTAINS": lambda args: exp.ArrayContains(
 411                this=seq_get(args, 1), expression=seq_get(args, 0)
 412            ),
 413            "ARRAY_GENERATE_RANGE": lambda args: exp.GenerateSeries(
 414                # ARRAY_GENERATE_RANGE has an exlusive end; we normalize it to be inclusive
 415                start=seq_get(args, 0),
 416                end=exp.Sub(this=seq_get(args, 1), expression=exp.Literal.number(1)),
 417                step=seq_get(args, 2),
 418            ),
 419            "BITXOR": _build_bitwise(exp.BitwiseXor, "BITXOR"),
 420            "BIT_XOR": _build_bitwise(exp.BitwiseXor, "BITXOR"),
 421            "BITOR": _build_bitwise(exp.BitwiseOr, "BITOR"),
 422            "BIT_OR": _build_bitwise(exp.BitwiseOr, "BITOR"),
 423            "BITSHIFTLEFT": _build_bitwise(exp.BitwiseLeftShift, "BITSHIFTLEFT"),
 424            "BIT_SHIFTLEFT": _build_bitwise(exp.BitwiseLeftShift, "BIT_SHIFTLEFT"),
 425            "BITSHIFTRIGHT": _build_bitwise(exp.BitwiseRightShift, "BITSHIFTRIGHT"),
 426            "BIT_SHIFTRIGHT": _build_bitwise(exp.BitwiseRightShift, "BIT_SHIFTRIGHT"),
 427            "BOOLXOR": _build_bitwise(exp.Xor, "BOOLXOR"),
 428            "DATE": _build_datetime("DATE", exp.DataType.Type.DATE),
 429            "DATE_TRUNC": _date_trunc_to_time,
 430            "DATEADD": _build_date_time_add(exp.DateAdd),
 431            "DATEDIFF": _build_datediff,
 432            "DIV0": _build_if_from_div0,
 433            "EDITDISTANCE": lambda args: exp.Levenshtein(
 434                this=seq_get(args, 0), expression=seq_get(args, 1), max_dist=seq_get(args, 2)
 435            ),
 436            "FLATTEN": exp.Explode.from_arg_list,
 437            "GET_PATH": lambda args, dialect: exp.JSONExtract(
 438                this=seq_get(args, 0), expression=dialect.to_json_path(seq_get(args, 1))
 439            ),
 440            "HEX_DECODE_BINARY": exp.Unhex.from_arg_list,
 441            "IFF": exp.If.from_arg_list,
 442            "LAST_DAY": lambda args: exp.LastDay(
 443                this=seq_get(args, 0), unit=map_date_part(seq_get(args, 1))
 444            ),
 445            "LEN": lambda args: exp.Length(this=seq_get(args, 0), binary=True),
 446            "LENGTH": lambda args: exp.Length(this=seq_get(args, 0), binary=True),
 447            "NULLIFZERO": _build_if_from_nullifzero,
 448            "OBJECT_CONSTRUCT": _build_object_construct,
 449            "REGEXP_EXTRACT_ALL": _build_regexp_extract(exp.RegexpExtractAll),
 450            "REGEXP_REPLACE": _build_regexp_replace,
 451            "REGEXP_SUBSTR": _build_regexp_extract(exp.RegexpExtract),
 452            "REGEXP_SUBSTR_ALL": _build_regexp_extract(exp.RegexpExtractAll),
 453            "RLIKE": exp.RegexpLike.from_arg_list,
 454            "SQUARE": lambda args: exp.Pow(this=seq_get(args, 0), expression=exp.Literal.number(2)),
 455            "TABLE": lambda args: exp.TableFromRows(this=seq_get(args, 0)),
 456            "TIMEADD": _build_date_time_add(exp.TimeAdd),
 457            "TIMEDIFF": _build_datediff,
 458            "TIMESTAMPADD": _build_date_time_add(exp.DateAdd),
 459            "TIMESTAMPDIFF": _build_datediff,
 460            "TIMESTAMPFROMPARTS": build_timestamp_from_parts,
 461            "TIMESTAMP_FROM_PARTS": build_timestamp_from_parts,
 462            "TIMESTAMPNTZFROMPARTS": build_timestamp_from_parts,
 463            "TIMESTAMP_NTZ_FROM_PARTS": build_timestamp_from_parts,
 464            "TRY_PARSE_JSON": lambda args: exp.ParseJSON(this=seq_get(args, 0), safe=True),
 465            "TRY_TO_DATE": _build_datetime("TRY_TO_DATE", exp.DataType.Type.DATE, safe=True),
 466            "TRY_TO_TIME": _build_datetime("TRY_TO_TIME", exp.DataType.Type.TIME, safe=True),
 467            "TRY_TO_TIMESTAMP": _build_datetime(
 468                "TRY_TO_TIMESTAMP", exp.DataType.Type.TIMESTAMP, safe=True
 469            ),
 470            "TO_CHAR": build_timetostr_or_tochar,
 471            "TO_DATE": _build_datetime("TO_DATE", exp.DataType.Type.DATE),
 472            "TO_NUMBER": lambda args: exp.ToNumber(
 473                this=seq_get(args, 0),
 474                format=seq_get(args, 1),
 475                precision=seq_get(args, 2),
 476                scale=seq_get(args, 3),
 477            ),
 478            "TO_TIME": _build_datetime("TO_TIME", exp.DataType.Type.TIME),
 479            "TO_TIMESTAMP": _build_datetime("TO_TIMESTAMP", exp.DataType.Type.TIMESTAMP),
 480            "TO_TIMESTAMP_LTZ": _build_datetime("TO_TIMESTAMP_LTZ", exp.DataType.Type.TIMESTAMPLTZ),
 481            "TO_TIMESTAMP_NTZ": _build_datetime("TO_TIMESTAMP_NTZ", exp.DataType.Type.TIMESTAMP),
 482            "TO_TIMESTAMP_TZ": _build_datetime("TO_TIMESTAMP_TZ", exp.DataType.Type.TIMESTAMPTZ),
 483            "TO_VARCHAR": exp.ToChar.from_arg_list,
 484            "ZEROIFNULL": _build_if_from_zeroifnull,
 485        }
 486
 487        FUNCTION_PARSERS = {
 488            **parser.Parser.FUNCTION_PARSERS,
 489            "DATE_PART": lambda self: self._parse_date_part(),
 490            "OBJECT_CONSTRUCT_KEEP_NULL": lambda self: self._parse_json_object(),
 491            "LISTAGG": lambda self: self._parse_string_agg(),
 492        }
 493        FUNCTION_PARSERS.pop("TRIM")
 494
 495        TIMESTAMPS = parser.Parser.TIMESTAMPS - {TokenType.TIME}
 496
 497        RANGE_PARSERS = {
 498            **parser.Parser.RANGE_PARSERS,
 499            TokenType.LIKE_ANY: parser.binary_range_parser(exp.LikeAny),
 500            TokenType.ILIKE_ANY: parser.binary_range_parser(exp.ILikeAny),
 501        }
 502
 503        ALTER_PARSERS = {
 504            **parser.Parser.ALTER_PARSERS,
 505            "UNSET": lambda self: self.expression(
 506                exp.Set,
 507                tag=self._match_text_seq("TAG"),
 508                expressions=self._parse_csv(self._parse_id_var),
 509                unset=True,
 510            ),
 511        }
 512
 513        STATEMENT_PARSERS = {
 514            **parser.Parser.STATEMENT_PARSERS,
 515            TokenType.PUT: lambda self: self._parse_put(),
 516            TokenType.SHOW: lambda self: self._parse_show(),
 517        }
 518
 519        PROPERTY_PARSERS = {
 520            **parser.Parser.PROPERTY_PARSERS,
 521            "CREDENTIALS": lambda self: self._parse_credentials_property(),
 522            "FILE_FORMAT": lambda self: self._parse_file_format_property(),
 523            "LOCATION": lambda self: self._parse_location_property(),
 524            "TAG": lambda self: self._parse_tag(),
 525            "USING": lambda self: self._match_text_seq("TEMPLATE")
 526            and self.expression(exp.UsingTemplateProperty, this=self._parse_statement()),
 527        }
 528
 529        TYPE_CONVERTERS = {
 530            # https://docs.snowflake.com/en/sql-reference/data-types-numeric#number
 531            exp.DataType.Type.DECIMAL: build_default_decimal_type(precision=38, scale=0),
 532        }
 533
 534        SHOW_PARSERS = {
 535            "DATABASES": _show_parser("DATABASES"),
 536            "TERSE DATABASES": _show_parser("DATABASES"),
 537            "SCHEMAS": _show_parser("SCHEMAS"),
 538            "TERSE SCHEMAS": _show_parser("SCHEMAS"),
 539            "OBJECTS": _show_parser("OBJECTS"),
 540            "TERSE OBJECTS": _show_parser("OBJECTS"),
 541            "TABLES": _show_parser("TABLES"),
 542            "TERSE TABLES": _show_parser("TABLES"),
 543            "VIEWS": _show_parser("VIEWS"),
 544            "TERSE VIEWS": _show_parser("VIEWS"),
 545            "PRIMARY KEYS": _show_parser("PRIMARY KEYS"),
 546            "TERSE PRIMARY KEYS": _show_parser("PRIMARY KEYS"),
 547            "IMPORTED KEYS": _show_parser("IMPORTED KEYS"),
 548            "TERSE IMPORTED KEYS": _show_parser("IMPORTED KEYS"),
 549            "UNIQUE KEYS": _show_parser("UNIQUE KEYS"),
 550            "TERSE UNIQUE KEYS": _show_parser("UNIQUE KEYS"),
 551            "SEQUENCES": _show_parser("SEQUENCES"),
 552            "TERSE SEQUENCES": _show_parser("SEQUENCES"),
 553            "STAGES": _show_parser("STAGES"),
 554            "COLUMNS": _show_parser("COLUMNS"),
 555            "USERS": _show_parser("USERS"),
 556            "TERSE USERS": _show_parser("USERS"),
 557            "FILE FORMATS": _show_parser("FILE FORMATS"),
 558            "FUNCTIONS": _show_parser("FUNCTIONS"),
 559            "PROCEDURES": _show_parser("PROCEDURES"),
 560            "WAREHOUSES": _show_parser("WAREHOUSES"),
 561        }
 562
 563        CONSTRAINT_PARSERS = {
 564            **parser.Parser.CONSTRAINT_PARSERS,
 565            "WITH": lambda self: self._parse_with_constraint(),
 566            "MASKING": lambda self: self._parse_with_constraint(),
 567            "PROJECTION": lambda self: self._parse_with_constraint(),
 568            "TAG": lambda self: self._parse_with_constraint(),
 569        }
 570
 571        STAGED_FILE_SINGLE_TOKENS = {
 572            TokenType.DOT,
 573            TokenType.MOD,
 574            TokenType.SLASH,
 575        }
 576
 577        FLATTEN_COLUMNS = ["SEQ", "KEY", "PATH", "INDEX", "VALUE", "THIS"]
 578
 579        SCHEMA_KINDS = {"OBJECTS", "TABLES", "VIEWS", "SEQUENCES", "UNIQUE KEYS", "IMPORTED KEYS"}
 580
 581        NON_TABLE_CREATABLES = {"STORAGE INTEGRATION", "TAG", "WAREHOUSE", "STREAMLIT"}
 582
 583        LAMBDAS = {
 584            **parser.Parser.LAMBDAS,
 585            TokenType.ARROW: lambda self, expressions: self.expression(
 586                exp.Lambda,
 587                this=self._replace_lambda(
 588                    self._parse_assignment(),
 589                    expressions,
 590                ),
 591                expressions=[e.this if isinstance(e, exp.Cast) else e for e in expressions],
 592            ),
 593        }
 594
 595        def _parse_use(self) -> exp.Use:
 596            if self._match_text_seq("SECONDARY", "ROLES"):
 597                this = self._match_texts(("ALL", "NONE")) and exp.var(self._prev.text.upper())
 598                roles = None if this else self._parse_csv(lambda: self._parse_table(schema=False))
 599                return self.expression(
 600                    exp.Use, kind="SECONDARY ROLES", this=this, expressions=roles
 601                )
 602
 603            return super()._parse_use()
 604
 605        def _negate_range(
 606            self, this: t.Optional[exp.Expression] = None
 607        ) -> t.Optional[exp.Expression]:
 608            if not this:
 609                return this
 610
 611            query = this.args.get("query")
 612            if isinstance(this, exp.In) and isinstance(query, exp.Query):
 613                # Snowflake treats `value NOT IN (subquery)` as `VALUE <> ALL (subquery)`, so
 614                # we do this conversion here to avoid parsing it into `NOT value IN (subquery)`
 615                # which can produce different results (most likely a SnowFlake bug).
 616                #
 617                # https://docs.snowflake.com/en/sql-reference/functions/in
 618                # Context: https://github.com/tobymao/sqlglot/issues/3890
 619                return self.expression(
 620                    exp.NEQ, this=this.this, expression=exp.All(this=query.unnest())
 621                )
 622
 623            return self.expression(exp.Not, this=this)
 624
 625        def _parse_tag(self) -> exp.Tags:
 626            return self.expression(
 627                exp.Tags,
 628                expressions=self._parse_wrapped_csv(self._parse_property),
 629            )
 630
 631        def _parse_with_constraint(self) -> t.Optional[exp.Expression]:
 632            if self._prev.token_type != TokenType.WITH:
 633                self._retreat(self._index - 1)
 634
 635            if self._match_text_seq("MASKING", "POLICY"):
 636                policy = self._parse_column()
 637                return self.expression(
 638                    exp.MaskingPolicyColumnConstraint,
 639                    this=policy.to_dot() if isinstance(policy, exp.Column) else policy,
 640                    expressions=self._match(TokenType.USING)
 641                    and self._parse_wrapped_csv(self._parse_id_var),
 642                )
 643            if self._match_text_seq("PROJECTION", "POLICY"):
 644                policy = self._parse_column()
 645                return self.expression(
 646                    exp.ProjectionPolicyColumnConstraint,
 647                    this=policy.to_dot() if isinstance(policy, exp.Column) else policy,
 648                )
 649            if self._match(TokenType.TAG):
 650                return self._parse_tag()
 651
 652            return None
 653
 654        def _parse_with_property(self) -> t.Optional[exp.Expression] | t.List[exp.Expression]:
 655            if self._match(TokenType.TAG):
 656                return self._parse_tag()
 657
 658            return super()._parse_with_property()
 659
 660        def _parse_create(self) -> exp.Create | exp.Command:
 661            expression = super()._parse_create()
 662            if isinstance(expression, exp.Create) and expression.kind in self.NON_TABLE_CREATABLES:
 663                # Replace the Table node with the enclosed Identifier
 664                expression.this.replace(expression.this.this)
 665
 666            return expression
 667
 668        # https://docs.snowflake.com/en/sql-reference/functions/date_part.html
 669        # https://docs.snowflake.com/en/sql-reference/functions-date-time.html#label-supported-date-time-parts
 670        def _parse_date_part(self: Snowflake.Parser) -> t.Optional[exp.Expression]:
 671            this = self._parse_var() or self._parse_type()
 672
 673            if not this:
 674                return None
 675
 676            self._match(TokenType.COMMA)
 677            expression = self._parse_bitwise()
 678            this = map_date_part(this)
 679            name = this.name.upper()
 680
 681            if name.startswith("EPOCH"):
 682                if name == "EPOCH_MILLISECOND":
 683                    scale = 10**3
 684                elif name == "EPOCH_MICROSECOND":
 685                    scale = 10**6
 686                elif name == "EPOCH_NANOSECOND":
 687                    scale = 10**9
 688                else:
 689                    scale = None
 690
 691                ts = self.expression(exp.Cast, this=expression, to=exp.DataType.build("TIMESTAMP"))
 692                to_unix: exp.Expression = self.expression(exp.TimeToUnix, this=ts)
 693
 694                if scale:
 695                    to_unix = exp.Mul(this=to_unix, expression=exp.Literal.number(scale))
 696
 697                return to_unix
 698
 699            return self.expression(exp.Extract, this=this, expression=expression)
 700
 701        def _parse_bracket_key_value(self, is_map: bool = False) -> t.Optional[exp.Expression]:
 702            if is_map:
 703                # Keys are strings in Snowflake's objects, see also:
 704                # - https://docs.snowflake.com/en/sql-reference/data-types-semistructured
 705                # - https://docs.snowflake.com/en/sql-reference/functions/object_construct
 706                return self._parse_slice(self._parse_string())
 707
 708            return self._parse_slice(self._parse_alias(self._parse_assignment(), explicit=True))
 709
 710        def _parse_lateral(self) -> t.Optional[exp.Lateral]:
 711            lateral = super()._parse_lateral()
 712            if not lateral:
 713                return lateral
 714
 715            if isinstance(lateral.this, exp.Explode):
 716                table_alias = lateral.args.get("alias")
 717                columns = [exp.to_identifier(col) for col in self.FLATTEN_COLUMNS]
 718                if table_alias and not table_alias.args.get("columns"):
 719                    table_alias.set("columns", columns)
 720                elif not table_alias:
 721                    exp.alias_(lateral, "_flattened", table=columns, copy=False)
 722
 723            return lateral
 724
 725        def _parse_table_parts(
 726            self, schema: bool = False, is_db_reference: bool = False, wildcard: bool = False
 727        ) -> exp.Table:
 728            # https://docs.snowflake.com/en/user-guide/querying-stage
 729            if self._match(TokenType.STRING, advance=False):
 730                table = self._parse_string()
 731            elif self._match_text_seq("@", advance=False):
 732                table = self._parse_location_path()
 733            else:
 734                table = None
 735
 736            if table:
 737                file_format = None
 738                pattern = None
 739
 740                wrapped = self._match(TokenType.L_PAREN)
 741                while self._curr and wrapped and not self._match(TokenType.R_PAREN):
 742                    if self._match_text_seq("FILE_FORMAT", "=>"):
 743                        file_format = self._parse_string() or super()._parse_table_parts(
 744                            is_db_reference=is_db_reference
 745                        )
 746                    elif self._match_text_seq("PATTERN", "=>"):
 747                        pattern = self._parse_string()
 748                    else:
 749                        break
 750
 751                    self._match(TokenType.COMMA)
 752
 753                table = self.expression(exp.Table, this=table, format=file_format, pattern=pattern)
 754            else:
 755                table = super()._parse_table_parts(schema=schema, is_db_reference=is_db_reference)
 756
 757            return table
 758
 759        def _parse_table(
 760            self,
 761            schema: bool = False,
 762            joins: bool = False,
 763            alias_tokens: t.Optional[t.Collection[TokenType]] = None,
 764            parse_bracket: bool = False,
 765            is_db_reference: bool = False,
 766            parse_partition: bool = False,
 767        ) -> t.Optional[exp.Expression]:
 768            table = super()._parse_table(
 769                schema=schema,
 770                joins=joins,
 771                alias_tokens=alias_tokens,
 772                parse_bracket=parse_bracket,
 773                is_db_reference=is_db_reference,
 774                parse_partition=parse_partition,
 775            )
 776            if isinstance(table, exp.Table) and isinstance(table.this, exp.TableFromRows):
 777                table_from_rows = table.this
 778                for arg in exp.TableFromRows.arg_types:
 779                    if arg != "this":
 780                        table_from_rows.set(arg, table.args.get(arg))
 781
 782                table = table_from_rows
 783
 784            return table
 785
 786        def _parse_id_var(
 787            self,
 788            any_token: bool = True,
 789            tokens: t.Optional[t.Collection[TokenType]] = None,
 790        ) -> t.Optional[exp.Expression]:
 791            if self._match_text_seq("IDENTIFIER", "("):
 792                identifier = (
 793                    super()._parse_id_var(any_token=any_token, tokens=tokens)
 794                    or self._parse_string()
 795                )
 796                self._match_r_paren()
 797                return self.expression(exp.Anonymous, this="IDENTIFIER", expressions=[identifier])
 798
 799            return super()._parse_id_var(any_token=any_token, tokens=tokens)
 800
 801        def _parse_show_snowflake(self, this: str) -> exp.Show:
 802            scope = None
 803            scope_kind = None
 804
 805            # will identity SHOW TERSE SCHEMAS but not SHOW TERSE PRIMARY KEYS
 806            # which is syntactically valid but has no effect on the output
 807            terse = self._tokens[self._index - 2].text.upper() == "TERSE"
 808
 809            history = self._match_text_seq("HISTORY")
 810
 811            like = self._parse_string() if self._match(TokenType.LIKE) else None
 812
 813            if self._match(TokenType.IN):
 814                if self._match_text_seq("ACCOUNT"):
 815                    scope_kind = "ACCOUNT"
 816                elif self._match_text_seq("CLASS"):
 817                    scope_kind = "CLASS"
 818                    scope = self._parse_table_parts()
 819                elif self._match_text_seq("APPLICATION"):
 820                    scope_kind = "APPLICATION"
 821                    if self._match_text_seq("PACKAGE"):
 822                        scope_kind += " PACKAGE"
 823                    scope = self._parse_table_parts()
 824                elif self._match_set(self.DB_CREATABLES):
 825                    scope_kind = self._prev.text.upper()
 826                    if self._curr:
 827                        scope = self._parse_table_parts()
 828                elif self._curr:
 829                    scope_kind = "SCHEMA" if this in self.SCHEMA_KINDS else "TABLE"
 830                    scope = self._parse_table_parts()
 831
 832            return self.expression(
 833                exp.Show,
 834                **{
 835                    "terse": terse,
 836                    "this": this,
 837                    "history": history,
 838                    "like": like,
 839                    "scope": scope,
 840                    "scope_kind": scope_kind,
 841                    "starts_with": self._match_text_seq("STARTS", "WITH") and self._parse_string(),
 842                    "limit": self._parse_limit(),
 843                    "from": self._parse_string() if self._match(TokenType.FROM) else None,
 844                    "privileges": self._match_text_seq("WITH", "PRIVILEGES")
 845                    and self._parse_csv(lambda: self._parse_var(any_token=True, upper=True)),
 846                },
 847            )
 848
 849        def _parse_put(self) -> exp.Put | exp.Command:
 850            if self._curr.token_type != TokenType.STRING:
 851                return self._parse_as_command(self._prev)
 852
 853            return self.expression(
 854                exp.Put,
 855                this=self._parse_string(),
 856                target=self._parse_location_path(),
 857                properties=self._parse_properties(),
 858            )
 859
 860        def _parse_location_property(self) -> exp.LocationProperty:
 861            self._match(TokenType.EQ)
 862            return self.expression(exp.LocationProperty, this=self._parse_location_path())
 863
 864        def _parse_file_location(self) -> t.Optional[exp.Expression]:
 865            # Parse either a subquery or a staged file
 866            return (
 867                self._parse_select(table=True, parse_subquery_alias=False)
 868                if self._match(TokenType.L_PAREN, advance=False)
 869                else self._parse_table_parts()
 870            )
 871
 872        def _parse_location_path(self) -> exp.Var:
 873            start = self._curr
 874            self._advance_any(ignore_reserved=True)
 875
 876            # We avoid consuming a comma token because external tables like @foo and @bar
 877            # can be joined in a query with a comma separator, as well as closing paren
 878            # in case of subqueries
 879            while self._is_connected() and not self._match_set(
 880                (TokenType.COMMA, TokenType.L_PAREN, TokenType.R_PAREN), advance=False
 881            ):
 882                self._advance_any(ignore_reserved=True)
 883
 884            return exp.var(self._find_sql(start, self._prev))
 885
 886        def _parse_lambda_arg(self) -> t.Optional[exp.Expression]:
 887            this = super()._parse_lambda_arg()
 888
 889            if not this:
 890                return this
 891
 892            typ = self._parse_types()
 893
 894            if typ:
 895                return self.expression(exp.Cast, this=this, to=typ)
 896
 897            return this
 898
 899        def _parse_foreign_key(self) -> exp.ForeignKey:
 900            # inlineFK, the REFERENCES columns are implied
 901            if self._match(TokenType.REFERENCES, advance=False):
 902                return self.expression(exp.ForeignKey)
 903
 904            # outoflineFK, explicitly names the columns
 905            return super()._parse_foreign_key()
 906
 907        def _parse_file_format_property(self) -> exp.FileFormatProperty:
 908            self._match(TokenType.EQ)
 909            if self._match(TokenType.L_PAREN, advance=False):
 910                expressions = self._parse_wrapped_options()
 911            else:
 912                expressions = [self._parse_format_name()]
 913
 914            return self.expression(
 915                exp.FileFormatProperty,
 916                expressions=expressions,
 917            )
 918
 919        def _parse_credentials_property(self) -> exp.CredentialsProperty:
 920            return self.expression(
 921                exp.CredentialsProperty,
 922                expressions=self._parse_wrapped_options(),
 923            )
 924
 925    class Tokenizer(tokens.Tokenizer):
 926        STRING_ESCAPES = ["\\", "'"]
 927        HEX_STRINGS = [("x'", "'"), ("X'", "'")]
 928        RAW_STRINGS = ["$$"]
 929        COMMENTS = ["--", "//", ("/*", "*/")]
 930        NESTED_COMMENTS = False
 931
 932        KEYWORDS = {
 933            **tokens.Tokenizer.KEYWORDS,
 934            "FILE://": TokenType.URI_START,
 935            "BYTEINT": TokenType.INT,
 936            "CHAR VARYING": TokenType.VARCHAR,
 937            "CHARACTER VARYING": TokenType.VARCHAR,
 938            "EXCLUDE": TokenType.EXCEPT,
 939            "FILE FORMAT": TokenType.FILE_FORMAT,
 940            "ILIKE ANY": TokenType.ILIKE_ANY,
 941            "LIKE ANY": TokenType.LIKE_ANY,
 942            "MATCH_CONDITION": TokenType.MATCH_CONDITION,
 943            "MATCH_RECOGNIZE": TokenType.MATCH_RECOGNIZE,
 944            "MINUS": TokenType.EXCEPT,
 945            "NCHAR VARYING": TokenType.VARCHAR,
 946            "PUT": TokenType.PUT,
 947            "REMOVE": TokenType.COMMAND,
 948            "RM": TokenType.COMMAND,
 949            "SAMPLE": TokenType.TABLE_SAMPLE,
 950            "SQL_DOUBLE": TokenType.DOUBLE,
 951            "SQL_VARCHAR": TokenType.VARCHAR,
 952            "STORAGE INTEGRATION": TokenType.STORAGE_INTEGRATION,
 953            "TAG": TokenType.TAG,
 954            "TIMESTAMP_TZ": TokenType.TIMESTAMPTZ,
 955            "TOP": TokenType.TOP,
 956            "WAREHOUSE": TokenType.WAREHOUSE,
 957            "STAGE": TokenType.STAGE,
 958            "STREAMLIT": TokenType.STREAMLIT,
 959        }
 960        KEYWORDS.pop("/*+")
 961
 962        SINGLE_TOKENS = {
 963            **tokens.Tokenizer.SINGLE_TOKENS,
 964            "$": TokenType.PARAMETER,
 965        }
 966
 967        VAR_SINGLE_TOKENS = {"$"}
 968
 969        COMMANDS = tokens.Tokenizer.COMMANDS - {TokenType.SHOW}
 970
 971    class Generator(generator.Generator):
 972        PARAMETER_TOKEN = "$"
 973        MATCHED_BY_SOURCE = False
 974        SINGLE_STRING_INTERVAL = True
 975        JOIN_HINTS = False
 976        TABLE_HINTS = False
 977        QUERY_HINTS = False
 978        AGGREGATE_FILTER_SUPPORTED = False
 979        SUPPORTS_TABLE_COPY = False
 980        COLLATE_IS_FUNC = True
 981        LIMIT_ONLY_LITERALS = True
 982        JSON_KEY_VALUE_PAIR_SEP = ","
 983        INSERT_OVERWRITE = " OVERWRITE INTO"
 984        STRUCT_DELIMITER = ("(", ")")
 985        COPY_PARAMS_ARE_WRAPPED = False
 986        COPY_PARAMS_EQ_REQUIRED = True
 987        STAR_EXCEPT = "EXCLUDE"
 988        SUPPORTS_EXPLODING_PROJECTIONS = False
 989        ARRAY_CONCAT_IS_VAR_LEN = False
 990        SUPPORTS_CONVERT_TIMEZONE = True
 991        EXCEPT_INTERSECT_SUPPORT_ALL_CLAUSE = False
 992        SUPPORTS_MEDIAN = True
 993        ARRAY_SIZE_NAME = "ARRAY_SIZE"
 994
 995        TRANSFORMS = {
 996            **generator.Generator.TRANSFORMS,
 997            exp.ApproxDistinct: rename_func("APPROX_COUNT_DISTINCT"),
 998            exp.ArgMax: rename_func("MAX_BY"),
 999            exp.ArgMin: rename_func("MIN_BY"),
1000            exp.Array: inline_array_sql,
1001            exp.ArrayConcat: lambda self, e: self.arrayconcat_sql(e, name="ARRAY_CAT"),
1002            exp.ArrayContains: lambda self, e: self.func("ARRAY_CONTAINS", e.expression, e.this),
1003            exp.AtTimeZone: lambda self, e: self.func(
1004                "CONVERT_TIMEZONE", e.args.get("zone"), e.this
1005            ),
1006            exp.BitwiseOr: rename_func("BITOR"),
1007            exp.BitwiseXor: rename_func("BITXOR"),
1008            exp.BitwiseLeftShift: rename_func("BITSHIFTLEFT"),
1009            exp.BitwiseRightShift: rename_func("BITSHIFTRIGHT"),
1010            exp.Create: transforms.preprocess([_flatten_structured_types_unless_iceberg]),
1011            exp.DateAdd: date_delta_sql("DATEADD"),
1012            exp.DateDiff: date_delta_sql("DATEDIFF"),
1013            exp.DatetimeAdd: date_delta_sql("TIMESTAMPADD"),
1014            exp.DatetimeDiff: timestampdiff_sql,
1015            exp.DateStrToDate: datestrtodate_sql,
1016            exp.DayOfMonth: rename_func("DAYOFMONTH"),
1017            exp.DayOfWeek: rename_func("DAYOFWEEK"),
1018            exp.DayOfYear: rename_func("DAYOFYEAR"),
1019            exp.Explode: rename_func("FLATTEN"),
1020            exp.Extract: rename_func("DATE_PART"),
1021            exp.FileFormatProperty: lambda self,
1022            e: f"FILE_FORMAT=({self.expressions(e, 'expressions', sep=' ')})",
1023            exp.FromTimeZone: lambda self, e: self.func(
1024                "CONVERT_TIMEZONE", e.args.get("zone"), "'UTC'", e.this
1025            ),
1026            exp.GenerateSeries: lambda self, e: self.func(
1027                "ARRAY_GENERATE_RANGE", e.args["start"], e.args["end"] + 1, e.args.get("step")
1028            ),
1029            exp.GroupConcat: lambda self, e: groupconcat_sql(self, e, sep=""),
1030            exp.If: if_sql(name="IFF", false_value="NULL"),
1031            exp.JSONExtractArray: _json_extract_value_array_sql,
1032            exp.JSONExtractScalar: lambda self, e: self.func(
1033                "JSON_EXTRACT_PATH_TEXT", e.this, e.expression
1034            ),
1035            exp.JSONObject: lambda self, e: self.func("OBJECT_CONSTRUCT_KEEP_NULL", *e.expressions),
1036            exp.JSONPathRoot: lambda *_: "",
1037            exp.JSONValueArray: _json_extract_value_array_sql,
1038            exp.Levenshtein: unsupported_args("ins_cost", "del_cost", "sub_cost")(
1039                rename_func("EDITDISTANCE")
1040            ),
1041            exp.LocationProperty: lambda self, e: f"LOCATION={self.sql(e, 'this')}",
1042            exp.LogicalAnd: rename_func("BOOLAND_AGG"),
1043            exp.LogicalOr: rename_func("BOOLOR_AGG"),
1044            exp.Map: lambda self, e: var_map_sql(self, e, "OBJECT_CONSTRUCT"),
1045            exp.MakeInterval: no_make_interval_sql,
1046            exp.Max: max_or_greatest,
1047            exp.Min: min_or_least,
1048            exp.ParseJSON: lambda self, e: self.func(
1049                "TRY_PARSE_JSON" if e.args.get("safe") else "PARSE_JSON", e.this
1050            ),
1051            exp.PartitionedByProperty: lambda self, e: f"PARTITION BY {self.sql(e, 'this')}",
1052            exp.PercentileCont: transforms.preprocess(
1053                [transforms.add_within_group_for_percentiles]
1054            ),
1055            exp.PercentileDisc: transforms.preprocess(
1056                [transforms.add_within_group_for_percentiles]
1057            ),
1058            exp.Pivot: transforms.preprocess([_unqualify_pivot_columns]),
1059            exp.RegexpExtract: _regexpextract_sql,
1060            exp.RegexpExtractAll: _regexpextract_sql,
1061            exp.RegexpILike: _regexpilike_sql,
1062            exp.Rand: rename_func("RANDOM"),
1063            exp.Select: transforms.preprocess(
1064                [
1065                    transforms.eliminate_distinct_on,
1066                    transforms.explode_projection_to_unnest(),
1067                    transforms.eliminate_semi_and_anti_joins,
1068                    _transform_generate_date_array,
1069                ]
1070            ),
1071            exp.SHA: rename_func("SHA1"),
1072            exp.StarMap: rename_func("OBJECT_CONSTRUCT"),
1073            exp.StartsWith: rename_func("STARTSWITH"),
1074            exp.StrPosition: lambda self, e: strposition_sql(
1075                self, e, func_name="CHARINDEX", supports_position=True
1076            ),
1077            exp.StrToDate: lambda self, e: self.func("DATE", e.this, self.format_time(e)),
1078            exp.Stuff: rename_func("INSERT"),
1079            exp.TimeAdd: date_delta_sql("TIMEADD"),
1080            exp.Timestamp: no_timestamp_sql,
1081            exp.TimestampAdd: date_delta_sql("TIMESTAMPADD"),
1082            exp.TimestampDiff: lambda self, e: self.func(
1083                "TIMESTAMPDIFF", e.unit, e.expression, e.this
1084            ),
1085            exp.TimestampTrunc: timestamptrunc_sql(),
1086            exp.TimeStrToTime: timestrtotime_sql,
1087            exp.TimeToUnix: lambda self, e: f"EXTRACT(epoch_second FROM {self.sql(e, 'this')})",
1088            exp.ToArray: rename_func("TO_ARRAY"),
1089            exp.ToChar: lambda self, e: self.function_fallback_sql(e),
1090            exp.ToDouble: rename_func("TO_DOUBLE"),
1091            exp.TsOrDsAdd: date_delta_sql("DATEADD", cast=True),
1092            exp.TsOrDsDiff: date_delta_sql("DATEDIFF"),
1093            exp.TsOrDsToDate: lambda self, e: self.func(
1094                "TRY_TO_DATE" if e.args.get("safe") else "TO_DATE", e.this, self.format_time(e)
1095            ),
1096            exp.TsOrDsToTime: lambda self, e: self.func(
1097                "TRY_TO_TIME" if e.args.get("safe") else "TO_TIME", e.this, self.format_time(e)
1098            ),
1099            exp.Unhex: rename_func("HEX_DECODE_BINARY"),
1100            exp.UnixToTime: rename_func("TO_TIMESTAMP"),
1101            exp.Uuid: rename_func("UUID_STRING"),
1102            exp.VarMap: lambda self, e: var_map_sql(self, e, "OBJECT_CONSTRUCT"),
1103            exp.WeekOfYear: rename_func("WEEKOFYEAR"),
1104            exp.Xor: rename_func("BOOLXOR"),
1105        }
1106
1107        SUPPORTED_JSON_PATH_PARTS = {
1108            exp.JSONPathKey,
1109            exp.JSONPathRoot,
1110            exp.JSONPathSubscript,
1111        }
1112
1113        TYPE_MAPPING = {
1114            **generator.Generator.TYPE_MAPPING,
1115            exp.DataType.Type.NESTED: "OBJECT",
1116            exp.DataType.Type.STRUCT: "OBJECT",
1117        }
1118
1119        TOKEN_MAPPING = {
1120            TokenType.AUTO_INCREMENT: "AUTOINCREMENT",
1121        }
1122
1123        PROPERTIES_LOCATION = {
1124            **generator.Generator.PROPERTIES_LOCATION,
1125            exp.CredentialsProperty: exp.Properties.Location.POST_WITH,
1126            exp.LocationProperty: exp.Properties.Location.POST_WITH,
1127            exp.PartitionedByProperty: exp.Properties.Location.POST_SCHEMA,
1128            exp.SetProperty: exp.Properties.Location.UNSUPPORTED,
1129            exp.VolatileProperty: exp.Properties.Location.UNSUPPORTED,
1130        }
1131
1132        UNSUPPORTED_VALUES_EXPRESSIONS = {
1133            exp.Map,
1134            exp.StarMap,
1135            exp.Struct,
1136            exp.VarMap,
1137        }
1138
1139        def with_properties(self, properties: exp.Properties) -> str:
1140            return self.properties(properties, wrapped=False, prefix=self.sep(""), sep=" ")
1141
1142        def values_sql(self, expression: exp.Values, values_as_table: bool = True) -> str:
1143            if expression.find(*self.UNSUPPORTED_VALUES_EXPRESSIONS):
1144                values_as_table = False
1145
1146            return super().values_sql(expression, values_as_table=values_as_table)
1147
1148        def datatype_sql(self, expression: exp.DataType) -> str:
1149            expressions = expression.expressions
1150            if (
1151                expressions
1152                and expression.is_type(*exp.DataType.STRUCT_TYPES)
1153                and any(isinstance(field_type, exp.DataType) for field_type in expressions)
1154            ):
1155                # The correct syntax is OBJECT [ (<key> <value_type [NOT NULL] [, ...]) ]
1156                return "OBJECT"
1157
1158            return super().datatype_sql(expression)
1159
1160        def tonumber_sql(self, expression: exp.ToNumber) -> str:
1161            return self.func(
1162                "TO_NUMBER",
1163                expression.this,
1164                expression.args.get("format"),
1165                expression.args.get("precision"),
1166                expression.args.get("scale"),
1167            )
1168
1169        def timestampfromparts_sql(self, expression: exp.TimestampFromParts) -> str:
1170            milli = expression.args.get("milli")
1171            if milli is not None:
1172                milli_to_nano = milli.pop() * exp.Literal.number(1000000)
1173                expression.set("nano", milli_to_nano)
1174
1175            return rename_func("TIMESTAMP_FROM_PARTS")(self, expression)
1176
1177        def cast_sql(self, expression: exp.Cast, safe_prefix: t.Optional[str] = None) -> str:
1178            if expression.is_type(exp.DataType.Type.GEOGRAPHY):
1179                return self.func("TO_GEOGRAPHY", expression.this)
1180            if expression.is_type(exp.DataType.Type.GEOMETRY):
1181                return self.func("TO_GEOMETRY", expression.this)
1182
1183            return super().cast_sql(expression, safe_prefix=safe_prefix)
1184
1185        def trycast_sql(self, expression: exp.TryCast) -> str:
1186            value = expression.this
1187
1188            if value.type is None:
1189                from sqlglot.optimizer.annotate_types import annotate_types
1190
1191                value = annotate_types(value, dialect=self.dialect)
1192
1193            if value.is_type(*exp.DataType.TEXT_TYPES, exp.DataType.Type.UNKNOWN):
1194                return super().trycast_sql(expression)
1195
1196            # TRY_CAST only works for string values in Snowflake
1197            return self.cast_sql(expression)
1198
1199        def log_sql(self, expression: exp.Log) -> str:
1200            if not expression.expression:
1201                return self.func("LN", expression.this)
1202
1203            return super().log_sql(expression)
1204
1205        def unnest_sql(self, expression: exp.Unnest) -> str:
1206            unnest_alias = expression.args.get("alias")
1207            offset = expression.args.get("offset")
1208
1209            columns = [
1210                exp.to_identifier("seq"),
1211                exp.to_identifier("key"),
1212                exp.to_identifier("path"),
1213                offset.pop() if isinstance(offset, exp.Expression) else exp.to_identifier("index"),
1214                seq_get(unnest_alias.columns if unnest_alias else [], 0)
1215                or exp.to_identifier("value"),
1216                exp.to_identifier("this"),
1217            ]
1218
1219            if unnest_alias:
1220                unnest_alias.set("columns", columns)
1221            else:
1222                unnest_alias = exp.TableAlias(this="_u", columns=columns)
1223
1224            table_input = self.sql(expression.expressions[0])
1225            if not table_input.startswith("INPUT =>"):
1226                table_input = f"INPUT => {table_input}"
1227
1228            explode = f"TABLE(FLATTEN({table_input}))"
1229            alias = self.sql(unnest_alias)
1230            alias = f" AS {alias}" if alias else ""
1231            return f"{explode}{alias}"
1232
1233        def show_sql(self, expression: exp.Show) -> str:
1234            terse = "TERSE " if expression.args.get("terse") else ""
1235            history = " HISTORY" if expression.args.get("history") else ""
1236            like = self.sql(expression, "like")
1237            like = f" LIKE {like}" if like else ""
1238
1239            scope = self.sql(expression, "scope")
1240            scope = f" {scope}" if scope else ""
1241
1242            scope_kind = self.sql(expression, "scope_kind")
1243            if scope_kind:
1244                scope_kind = f" IN {scope_kind}"
1245
1246            starts_with = self.sql(expression, "starts_with")
1247            if starts_with:
1248                starts_with = f" STARTS WITH {starts_with}"
1249
1250            limit = self.sql(expression, "limit")
1251
1252            from_ = self.sql(expression, "from")
1253            if from_:
1254                from_ = f" FROM {from_}"
1255
1256            privileges = self.expressions(expression, key="privileges", flat=True)
1257            privileges = f" WITH PRIVILEGES {privileges}" if privileges else ""
1258
1259            return f"SHOW {terse}{expression.name}{history}{like}{scope_kind}{scope}{starts_with}{limit}{from_}{privileges}"
1260
1261        def describe_sql(self, expression: exp.Describe) -> str:
1262            # Default to table if kind is unknown
1263            kind_value = expression.args.get("kind") or "TABLE"
1264            kind = f" {kind_value}" if kind_value else ""
1265            this = f" {self.sql(expression, 'this')}"
1266            expressions = self.expressions(expression, flat=True)
1267            expressions = f" {expressions}" if expressions else ""
1268            return f"DESCRIBE{kind}{this}{expressions}"
1269
1270        def generatedasidentitycolumnconstraint_sql(
1271            self, expression: exp.GeneratedAsIdentityColumnConstraint
1272        ) -> str:
1273            start = expression.args.get("start")
1274            start = f" START {start}" if start else ""
1275            increment = expression.args.get("increment")
1276            increment = f" INCREMENT {increment}" if increment else ""
1277            return f"AUTOINCREMENT{start}{increment}"
1278
1279        def cluster_sql(self, expression: exp.Cluster) -> str:
1280            return f"CLUSTER BY ({self.expressions(expression, flat=True)})"
1281
1282        def struct_sql(self, expression: exp.Struct) -> str:
1283            keys = []
1284            values = []
1285
1286            for i, e in enumerate(expression.expressions):
1287                if isinstance(e, exp.PropertyEQ):
1288                    keys.append(
1289                        exp.Literal.string(e.name) if isinstance(e.this, exp.Identifier) else e.this
1290                    )
1291                    values.append(e.expression)
1292                else:
1293                    keys.append(exp.Literal.string(f"_{i}"))
1294                    values.append(e)
1295
1296            return self.func("OBJECT_CONSTRUCT", *flatten(zip(keys, values)))
1297
1298        @unsupported_args("weight", "accuracy")
1299        def approxquantile_sql(self, expression: exp.ApproxQuantile) -> str:
1300            return self.func("APPROX_PERCENTILE", expression.this, expression.args.get("quantile"))
1301
1302        def alterset_sql(self, expression: exp.AlterSet) -> str:
1303            exprs = self.expressions(expression, flat=True)
1304            exprs = f" {exprs}" if exprs else ""
1305            file_format = self.expressions(expression, key="file_format", flat=True, sep=" ")
1306            file_format = f" STAGE_FILE_FORMAT = ({file_format})" if file_format else ""
1307            copy_options = self.expressions(expression, key="copy_options", flat=True, sep=" ")
1308            copy_options = f" STAGE_COPY_OPTIONS = ({copy_options})" if copy_options else ""
1309            tag = self.expressions(expression, key="tag", flat=True)
1310            tag = f" TAG {tag}" if tag else ""
1311
1312            return f"SET{exprs}{file_format}{copy_options}{tag}"
1313
1314        def strtotime_sql(self, expression: exp.StrToTime):
1315            safe_prefix = "TRY_" if expression.args.get("safe") else ""
1316            return self.func(
1317                f"{safe_prefix}TO_TIMESTAMP", expression.this, self.format_time(expression)
1318            )
1319
1320        def timestampsub_sql(self, expression: exp.TimestampSub):
1321            return self.sql(
1322                exp.TimestampAdd(
1323                    this=expression.this,
1324                    expression=expression.expression * -1,
1325                    unit=expression.unit,
1326                )
1327            )
1328
1329        def jsonextract_sql(self, expression: exp.JSONExtract):
1330            this = expression.this
1331
1332            # JSON strings are valid coming from other dialects such as BQ
1333            return self.func(
1334                "GET_PATH",
1335                exp.ParseJSON(this=this) if this.is_string else this,
1336                expression.expression,
1337            )
1338
1339        def timetostr_sql(self, expression: exp.TimeToStr) -> str:
1340            this = expression.this
1341            if not isinstance(this, exp.TsOrDsToTimestamp):
1342                this = exp.cast(this, exp.DataType.Type.TIMESTAMP)
1343
1344            return self.func("TO_CHAR", this, self.format_time(expression))
1345
1346        def datesub_sql(self, expression: exp.DateSub) -> str:
1347            value = expression.expression
1348            if value:
1349                value.replace(value * (-1))
1350            else:
1351                self.unsupported("DateSub cannot be transpiled if the subtracted count is unknown")
1352
1353            return date_delta_sql("DATEADD")(self, expression)
1354
1355        def select_sql(self, expression: exp.Select) -> str:
1356            limit = expression.args.get("limit")
1357            offset = expression.args.get("offset")
1358            if offset and not limit:
1359                expression.limit(exp.Null(), copy=False)
1360            return super().select_sql(expression)