from __future__ import annotations import typing as t from sqlglot import exp, generator, parser, tokens, transforms from sqlglot.dialects.dialect import ( Dialect, NormalizationStrategy, binary_from_function, build_default_decimal_type, build_timestamp_from_parts, date_delta_sql, date_trunc_to_time, datestrtodate_sql, build_formatted_time, if_sql, inline_array_sql, max_or_greatest, min_or_least, rename_func, timestamptrunc_sql, timestrtotime_sql, var_map_sql, map_date_part, no_safe_divide_sql, no_timestamp_sql, timestampdiff_sql, no_make_interval_sql, ) from sqlglot.generator import unsupported_args from sqlglot.helper import flatten, is_float, is_int, seq_get from sqlglot.tokens import TokenType if t.TYPE_CHECKING: from sqlglot._typing import E # from https://docs.snowflake.com/en/sql-reference/functions/to_timestamp.html def _build_datetime( name: str, kind: exp.DataType.Type, safe: bool = False ) -> t.Callable[[t.List], exp.Func]: def _builder(args: t.List) -> exp.Func: value = seq_get(args, 0) scale_or_fmt = seq_get(args, 1) int_value = value is not None and is_int(value.name) int_scale_or_fmt = scale_or_fmt is not None and scale_or_fmt.is_int if isinstance(value, exp.Literal) or (value and scale_or_fmt): # Converts calls like `TO_TIME('01:02:03')` into casts if len(args) == 1 and value.is_string and not int_value: return ( exp.TryCast(this=value, to=exp.DataType.build(kind)) if safe else exp.cast(value, kind) ) # Handles `TO_TIMESTAMP(str, fmt)` and `TO_TIMESTAMP(num, scale)` as special # cases so we can transpile them, since they're relatively common if kind == exp.DataType.Type.TIMESTAMP: if not safe and (int_value or int_scale_or_fmt): # TRY_TO_TIMESTAMP('integer') is not parsed into exp.UnixToTime as # it's not easily transpilable return exp.UnixToTime(this=value, scale=scale_or_fmt) if not int_scale_or_fmt and not is_float(value.name): expr = build_formatted_time(exp.StrToTime, "snowflake")(args) expr.set("safe", safe) return expr if kind == exp.DataType.Type.DATE and not int_value: formatted_exp = build_formatted_time(exp.TsOrDsToDate, "snowflake")(args) formatted_exp.set("safe", safe) return formatted_exp return exp.Anonymous(this=name, expressions=args) return _builder def _build_object_construct(args: t.List) -> t.Union[exp.StarMap, exp.Struct]: expression = parser.build_var_map(args) if isinstance(expression, exp.StarMap): return expression return exp.Struct( expressions=[ exp.PropertyEQ(this=k, expression=v) for k, v in zip(expression.keys, expression.values) ] ) def _build_datediff(args: t.List) -> exp.DateDiff: return exp.DateDiff( this=seq_get(args, 2), expression=seq_get(args, 1), unit=map_date_part(seq_get(args, 0)) ) def _build_date_time_add(expr_type: t.Type[E]) -> t.Callable[[t.List], E]: def _builder(args: t.List) -> E: return expr_type( this=seq_get(args, 2), expression=seq_get(args, 1), unit=map_date_part(seq_get(args, 0)), ) return _builder # https://docs.snowflake.com/en/sql-reference/functions/div0 def _build_if_from_div0(args: t.List) -> exp.If: lhs = exp._wrap(seq_get(args, 0), exp.Binary) rhs = exp._wrap(seq_get(args, 1), exp.Binary) cond = exp.EQ(this=rhs, expression=exp.Literal.number(0)).and_( exp.Is(this=lhs, expression=exp.null()).not_() ) true = exp.Literal.number(0) false = exp.Div(this=lhs, expression=rhs) return exp.If(this=cond, true=true, false=false) # https://docs.snowflake.com/en/sql-reference/functions/zeroifnull def _build_if_from_zeroifnull(args: t.List) -> exp.If: cond = exp.Is(this=seq_get(args, 0), expression=exp.Null()) return exp.If(this=cond, true=exp.Literal.number(0), false=seq_get(args, 0)) # https://docs.snowflake.com/en/sql-reference/functions/zeroifnull def _build_if_from_nullifzero(args: t.List) -> exp.If: cond = exp.EQ(this=seq_get(args, 0), expression=exp.Literal.number(0)) return exp.If(this=cond, true=exp.Null(), false=seq_get(args, 0)) def _regexpilike_sql(self: Snowflake.Generator, expression: exp.RegexpILike) -> str: flag = expression.text("flag") if "i" not in flag: flag += "i" return self.func( "REGEXP_LIKE", expression.this, expression.expression, exp.Literal.string(flag) ) def _build_regexp_replace(args: t.List) -> exp.RegexpReplace: regexp_replace = exp.RegexpReplace.from_arg_list(args) if not regexp_replace.args.get("replacement"): regexp_replace.set("replacement", exp.Literal.string("")) return regexp_replace def _show_parser(*args: t.Any, **kwargs: t.Any) -> t.Callable[[Snowflake.Parser], exp.Show]: def _parse(self: Snowflake.Parser) -> exp.Show: return self._parse_show_snowflake(*args, **kwargs) return _parse def _date_trunc_to_time(args: t.List) -> exp.DateTrunc | exp.TimestampTrunc: trunc = date_trunc_to_time(args) trunc.set("unit", map_date_part(trunc.args["unit"])) return trunc def _unqualify_unpivot_columns(expression: exp.Expression) -> exp.Expression: """ Snowflake doesn't allow columns referenced in UNPIVOT to be qualified, so we need to unqualify them. Example: >>> from sqlglot import parse_one >>> expr = parse_one("SELECT * FROM m_sales UNPIVOT(sales FOR month IN (m_sales.jan, feb, mar, april))") >>> print(_unqualify_unpivot_columns(expr).sql(dialect="snowflake")) SELECT * FROM m_sales UNPIVOT(sales FOR month IN (jan, feb, mar, april)) """ if isinstance(expression, exp.Pivot) and expression.unpivot: expression = transforms.unqualify_columns(expression) return expression def _flatten_structured_types_unless_iceberg(expression: exp.Expression) -> exp.Expression: assert isinstance(expression, exp.Create) def _flatten_structured_type(expression: exp.DataType) -> exp.DataType: if expression.this in exp.DataType.NESTED_TYPES: expression.set("expressions", None) return expression props = expression.args.get("properties") if isinstance(expression.this, exp.Schema) and not (props and props.find(exp.IcebergProperty)): for schema_expression in expression.this.expressions: if isinstance(schema_expression, exp.ColumnDef): column_type = schema_expression.kind if isinstance(column_type, exp.DataType): column_type.transform(_flatten_structured_type, copy=False) return expression def _unnest_generate_date_array(unnest: exp.Unnest) -> None: generate_date_array = unnest.expressions[0] start = generate_date_array.args.get("start") end = generate_date_array.args.get("end") step = generate_date_array.args.get("step") if not start or not end or not isinstance(step, exp.Interval) or step.name != "1": return unit = step.args.get("unit") unnest_alias = unnest.args.get("alias") if unnest_alias: unnest_alias = unnest_alias.copy() sequence_value_name = seq_get(unnest_alias.columns, 0) or "value" else: sequence_value_name = "value" # We'll add the next sequence value to the starting date and project the result date_add = _build_date_time_add(exp.DateAdd)( [unit, exp.cast(sequence_value_name, "int"), exp.cast(start, "date")] ).as_(sequence_value_name) # We use DATEDIFF to compute the number of sequence values needed number_sequence = Snowflake.Parser.FUNCTIONS["ARRAY_GENERATE_RANGE"]( [exp.Literal.number(0), _build_datediff([unit, start, end]) + 1] ) unnest.set("expressions", [number_sequence]) unnest.replace(exp.select(date_add).from_(unnest.copy()).subquery(unnest_alias)) def _transform_generate_date_array(expression: exp.Expression) -> exp.Expression: if isinstance(expression, exp.Select): for generate_date_array in expression.find_all(exp.GenerateDateArray): parent = generate_date_array.parent # If GENERATE_DATE_ARRAY is used directly as an array (e.g passed into ARRAY_LENGTH), the transformed Snowflake # query is the following (it'll be unnested properly on the next iteration due to copy): # SELECT ref(GENERATE_DATE_ARRAY(...)) -> SELECT ref((SELECT ARRAY_AGG(*) FROM UNNEST(GENERATE_DATE_ARRAY(...)))) if not isinstance(parent, exp.Unnest): unnest = exp.Unnest(expressions=[generate_date_array.copy()]) generate_date_array.replace( exp.select(exp.ArrayAgg(this=exp.Star())).from_(unnest).subquery() ) if ( isinstance(parent, exp.Unnest) and isinstance(parent.parent, (exp.From, exp.Join)) and len(parent.expressions) == 1 ): _unnest_generate_date_array(parent) return expression def _build_regexp_extract(expr_type: t.Type[E]) -> t.Callable[[t.List], E]: def _builder(args: t.List) -> E: return expr_type( this=seq_get(args, 0), expression=seq_get(args, 1), position=seq_get(args, 2), occurrence=seq_get(args, 3), parameters=seq_get(args, 4), group=seq_get(args, 5) or exp.Literal.number(0), ) return _builder def _regexpextract_sql(self, expression: exp.RegexpExtract | exp.RegexpExtractAll) -> str: # Other dialects don't support all of the following parameters, so we need to # generate default values as necessary to ensure the transpilation is correct group = expression.args.get("group") # To avoid generating all these default values, we set group to None if # it's 0 (also default value) which doesn't trigger the following chain if group and group.name == "0": group = None parameters = expression.args.get("parameters") or (group and exp.Literal.string("c")) occurrence = expression.args.get("occurrence") or (parameters and exp.Literal.number(1)) position = expression.args.get("position") or (occurrence and exp.Literal.number(1)) return self.func( "REGEXP_SUBSTR" if isinstance(expression, exp.RegexpExtract) else "REGEXP_EXTRACT_ALL", expression.this, expression.expression, position, occurrence, parameters, group, ) def _json_extract_value_array_sql( self: Snowflake.Generator, expression: exp.JSONValueArray | exp.JSONExtractArray ) -> str: json_extract = exp.JSONExtract(this=expression.this, expression=expression.expression) ident = exp.to_identifier("x") if isinstance(expression, exp.JSONValueArray): this: exp.Expression = exp.cast(ident, to=exp.DataType.Type.VARCHAR) else: this = exp.ParseJSON(this=f"TO_JSON({ident})") transform_lambda = exp.Lambda(expressions=[ident], this=this) return self.func("TRANSFORM", json_extract, transform_lambda) class Snowflake(Dialect): # https://docs.snowflake.com/en/sql-reference/identifiers-syntax NORMALIZATION_STRATEGY = NormalizationStrategy.UPPERCASE NULL_ORDERING = "nulls_are_large" TIME_FORMAT = "'YYYY-MM-DD HH24:MI:SS'" SUPPORTS_USER_DEFINED_TYPES = False SUPPORTS_SEMI_ANTI_JOIN = False PREFER_CTE_ALIAS_COLUMN = True TABLESAMPLE_SIZE_IS_PERCENT = True COPY_PARAMS_ARE_CSV = False ARRAY_AGG_INCLUDES_NULLS = None TIME_MAPPING = { "YYYY": "%Y", "yyyy": "%Y", "YY": "%y", "yy": "%y", "MMMM": "%B", "mmmm": "%B", "MON": "%b", "mon": "%b", "MM": "%m", "mm": "%m", "DD": "%d", "dd": "%-d", "DY": "%a", "dy": "%w", "HH24": "%H", "hh24": "%H", "HH12": "%I", "hh12": "%I", "MI": "%M", "mi": "%M", "SS": "%S", "ss": "%S", "FF": "%f", "ff": "%f", "FF6": "%f", "ff6": "%f", } def quote_identifier(self, expression: E, identify: bool = True) -> E: # This disables quoting DUAL in SELECT ... FROM DUAL, because Snowflake treats an # unquoted DUAL keyword in a special way and does not map it to a user-defined table if ( isinstance(expression, exp.Identifier) and isinstance(expression.parent, exp.Table) and expression.name.lower() == "dual" ): return expression # type: ignore return super().quote_identifier(expression, identify=identify) class Parser(parser.Parser): IDENTIFY_PIVOT_STRINGS = True DEFAULT_SAMPLING_METHOD = "BERNOULLI" COLON_IS_VARIANT_EXTRACT = True ID_VAR_TOKENS = { *parser.Parser.ID_VAR_TOKENS, TokenType.MATCH_CONDITION, } TABLE_ALIAS_TOKENS = parser.Parser.TABLE_ALIAS_TOKENS | {TokenType.WINDOW} TABLE_ALIAS_TOKENS.discard(TokenType.MATCH_CONDITION) FUNCTIONS = { **parser.Parser.FUNCTIONS, "APPROX_PERCENTILE": exp.ApproxQuantile.from_arg_list, "ARRAY_CONSTRUCT": lambda args: exp.Array(expressions=args), "ARRAY_CONTAINS": lambda args: exp.ArrayContains( this=seq_get(args, 1), expression=seq_get(args, 0) ), "ARRAY_GENERATE_RANGE": lambda args: exp.GenerateSeries( # ARRAY_GENERATE_RANGE has an exlusive end; we normalize it to be inclusive start=seq_get(args, 0), end=exp.Sub(this=seq_get(args, 1), expression=exp.Literal.number(1)), step=seq_get(args, 2), ), "BITXOR": binary_from_function(exp.BitwiseXor), "BIT_XOR": binary_from_function(exp.BitwiseXor), "BOOLXOR": binary_from_function(exp.Xor), "DATE": _build_datetime("DATE", exp.DataType.Type.DATE), "DATE_TRUNC": _date_trunc_to_time, "DATEADD": _build_date_time_add(exp.DateAdd), "DATEDIFF": _build_datediff, "DIV0": _build_if_from_div0, "EDITDISTANCE": lambda args: exp.Levenshtein( this=seq_get(args, 0), expression=seq_get(args, 1), max_dist=seq_get(args, 2) ), "FLATTEN": exp.Explode.from_arg_list, "GET_PATH": lambda args, dialect: exp.JSONExtract( this=seq_get(args, 0), expression=dialect.to_json_path(seq_get(args, 1)) ), "IFF": exp.If.from_arg_list, "LAST_DAY": lambda args: exp.LastDay( this=seq_get(args, 0), unit=map_date_part(seq_get(args, 1)) ), "LEN": lambda args: exp.Length(this=seq_get(args, 0), binary=True), "LENGTH": lambda args: exp.Length(this=seq_get(args, 0), binary=True), "LISTAGG": exp.GroupConcat.from_arg_list, "NULLIFZERO": _build_if_from_nullifzero, "OBJECT_CONSTRUCT": _build_object_construct, "REGEXP_EXTRACT_ALL": _build_regexp_extract(exp.RegexpExtractAll), "REGEXP_REPLACE": _build_regexp_replace, "REGEXP_SUBSTR": _build_regexp_extract(exp.RegexpExtract), "REGEXP_SUBSTR_ALL": _build_regexp_extract(exp.RegexpExtractAll), "RLIKE": exp.RegexpLike.from_arg_list, "SQUARE": lambda args: exp.Pow(this=seq_get(args, 0), expression=exp.Literal.number(2)), "TIMEADD": _build_date_time_add(exp.TimeAdd), "TIMEDIFF": _build_datediff, "TIMESTAMPADD": _build_date_time_add(exp.DateAdd), "TIMESTAMPDIFF": _build_datediff, "TIMESTAMPFROMPARTS": build_timestamp_from_parts, "TIMESTAMP_FROM_PARTS": build_timestamp_from_parts, "TIMESTAMPNTZFROMPARTS": build_timestamp_from_parts, "TIMESTAMP_NTZ_FROM_PARTS": build_timestamp_from_parts, "TRY_PARSE_JSON": lambda args: exp.ParseJSON(this=seq_get(args, 0), safe=True), "TRY_TO_DATE": _build_datetime("TRY_TO_DATE", exp.DataType.Type.DATE, safe=True), "TRY_TO_TIMESTAMP": _build_datetime( "TRY_TO_TIMESTAMP", exp.DataType.Type.TIMESTAMP, safe=True ), "TO_DATE": _build_datetime("TO_DATE", exp.DataType.Type.DATE), "TO_NUMBER": lambda args: exp.ToNumber( this=seq_get(args, 0), format=seq_get(args, 1), precision=seq_get(args, 2), scale=seq_get(args, 3), ), "TO_TIME": _build_datetime("TO_TIME", exp.DataType.Type.TIME), "TO_TIMESTAMP": _build_datetime("TO_TIMESTAMP", exp.DataType.Type.TIMESTAMP), "TO_TIMESTAMP_LTZ": _build_datetime("TO_TIMESTAMP_LTZ", exp.DataType.Type.TIMESTAMPLTZ), "TO_TIMESTAMP_NTZ": _build_datetime("TO_TIMESTAMP_NTZ", exp.DataType.Type.TIMESTAMP), "TO_TIMESTAMP_TZ": _build_datetime("TO_TIMESTAMP_TZ", exp.DataType.Type.TIMESTAMPTZ), "TO_VARCHAR": exp.ToChar.from_arg_list, "ZEROIFNULL": _build_if_from_zeroifnull, } FUNCTION_PARSERS = { **parser.Parser.FUNCTION_PARSERS, "DATE_PART": lambda self: self._parse_date_part(), "OBJECT_CONSTRUCT_KEEP_NULL": lambda self: self._parse_json_object(), } FUNCTION_PARSERS.pop("TRIM") TIMESTAMPS = parser.Parser.TIMESTAMPS - {TokenType.TIME} RANGE_PARSERS = { **parser.Parser.RANGE_PARSERS, TokenType.LIKE_ANY: parser.binary_range_parser(exp.LikeAny), TokenType.ILIKE_ANY: parser.binary_range_parser(exp.ILikeAny), } ALTER_PARSERS = { **parser.Parser.ALTER_PARSERS, "UNSET": lambda self: self.expression( exp.Set, tag=self._match_text_seq("TAG"), expressions=self._parse_csv(self._parse_id_var), unset=True, ), } STATEMENT_PARSERS = { **parser.Parser.STATEMENT_PARSERS, TokenType.SHOW: lambda self: self._parse_show(), } PROPERTY_PARSERS = { **parser.Parser.PROPERTY_PARSERS, "LOCATION": lambda self: self._parse_location_property(), "TAG": lambda self: self._parse_tag(), } TYPE_CONVERTERS = { # https://docs.snowflake.com/en/sql-reference/data-types-numeric#number exp.DataType.Type.DECIMAL: build_default_decimal_type(precision=38, scale=0), } SHOW_PARSERS = { "SCHEMAS": _show_parser("SCHEMAS"), "TERSE SCHEMAS": _show_parser("SCHEMAS"), "OBJECTS": _show_parser("OBJECTS"), "TERSE OBJECTS": _show_parser("OBJECTS"), "TABLES": _show_parser("TABLES"), "TERSE TABLES": _show_parser("TABLES"), "VIEWS": _show_parser("VIEWS"), "TERSE VIEWS": _show_parser("VIEWS"), "PRIMARY KEYS": _show_parser("PRIMARY KEYS"), "TERSE PRIMARY KEYS": _show_parser("PRIMARY KEYS"), "IMPORTED KEYS": _show_parser("IMPORTED KEYS"), "TERSE IMPORTED KEYS": _show_parser("IMPORTED KEYS"), "UNIQUE KEYS": _show_parser("UNIQUE KEYS"), "TERSE UNIQUE KEYS": _show_parser("UNIQUE KEYS"), "SEQUENCES": _show_parser("SEQUENCES"), "TERSE SEQUENCES": _show_parser("SEQUENCES"), "COLUMNS": _show_parser("COLUMNS"), "USERS": _show_parser("USERS"), "TERSE USERS": _show_parser("USERS"), } CONSTRAINT_PARSERS = { **parser.Parser.CONSTRAINT_PARSERS, "WITH": lambda self: self._parse_with_constraint(), "MASKING": lambda self: self._parse_with_constraint(), "PROJECTION": lambda self: self._parse_with_constraint(), "TAG": lambda self: self._parse_with_constraint(), } STAGED_FILE_SINGLE_TOKENS = { TokenType.DOT, TokenType.MOD, TokenType.SLASH, } FLATTEN_COLUMNS = ["SEQ", "KEY", "PATH", "INDEX", "VALUE", "THIS"] SCHEMA_KINDS = {"OBJECTS", "TABLES", "VIEWS", "SEQUENCES", "UNIQUE KEYS", "IMPORTED KEYS"} NON_TABLE_CREATABLES = {"STORAGE INTEGRATION", "TAG", "WAREHOUSE", "STREAMLIT"} LAMBDAS = { **parser.Parser.LAMBDAS, TokenType.ARROW: lambda self, expressions: self.expression( exp.Lambda, this=self._replace_lambda( self._parse_assignment(), expressions, ), expressions=[e.this if isinstance(e, exp.Cast) else e for e in expressions], ), } def _negate_range( self, this: t.Optional[exp.Expression] = None ) -> t.Optional[exp.Expression]: if not this: return this query = this.args.get("query") if isinstance(this, exp.In) and isinstance(query, exp.Query): # Snowflake treats `value NOT IN (subquery)` as `VALUE <> ALL (subquery)`, so # we do this conversion here to avoid parsing it into `NOT value IN (subquery)` # which can produce different results (most likely a SnowFlake bug). # # https://docs.snowflake.com/en/sql-reference/functions/in # Context: https://github.com/tobymao/sqlglot/issues/3890 return self.expression( exp.NEQ, this=this.this, expression=exp.All(this=query.unnest()) ) return self.expression(exp.Not, this=this) def _parse_tag(self) -> exp.Tags: return self.expression( exp.Tags, expressions=self._parse_wrapped_csv(self._parse_property), ) def _parse_with_constraint(self) -> t.Optional[exp.Expression]: if self._prev.token_type != TokenType.WITH: self._retreat(self._index - 1) if self._match_text_seq("MASKING", "POLICY"): policy = self._parse_column() return self.expression( exp.MaskingPolicyColumnConstraint, this=policy.to_dot() if isinstance(policy, exp.Column) else policy, expressions=self._match(TokenType.USING) and self._parse_wrapped_csv(self._parse_id_var), ) if self._match_text_seq("PROJECTION", "POLICY"): policy = self._parse_column() return self.expression( exp.ProjectionPolicyColumnConstraint, this=policy.to_dot() if isinstance(policy, exp.Column) else policy, ) if self._match(TokenType.TAG): return self._parse_tag() return None def _parse_with_property(self) -> t.Optional[exp.Expression] | t.List[exp.Expression]: if self._match(TokenType.TAG): return self._parse_tag() return super()._parse_with_property() def _parse_create(self) -> exp.Create | exp.Command: expression = super()._parse_create() if isinstance(expression, exp.Create) and expression.kind in self.NON_TABLE_CREATABLES: # Replace the Table node with the enclosed Identifier expression.this.replace(expression.this.this) return expression # https://docs.snowflake.com/en/sql-reference/functions/date_part.html # https://docs.snowflake.com/en/sql-reference/functions-date-time.html#label-supported-date-time-parts def _parse_date_part(self: Snowflake.Parser) -> t.Optional[exp.Expression]: this = self._parse_var() or self._parse_type() if not this: return None self._match(TokenType.COMMA) expression = self._parse_bitwise() this = map_date_part(this) name = this.name.upper() if name.startswith("EPOCH"): if name == "EPOCH_MILLISECOND": scale = 10**3 elif name == "EPOCH_MICROSECOND": scale = 10**6 elif name == "EPOCH_NANOSECOND": scale = 10**9 else: scale = None ts = self.expression(exp.Cast, this=expression, to=exp.DataType.build("TIMESTAMP")) to_unix: exp.Expression = self.expression(exp.TimeToUnix, this=ts) if scale: to_unix = exp.Mul(this=to_unix, expression=exp.Literal.number(scale)) return to_unix return self.expression(exp.Extract, this=this, expression=expression) def _parse_bracket_key_value(self, is_map: bool = False) -> t.Optional[exp.Expression]: if is_map: # Keys are strings in Snowflake's objects, see also: # - https://docs.snowflake.com/en/sql-reference/data-types-semistructured # - https://docs.snowflake.com/en/sql-reference/functions/object_construct return self._parse_slice(self._parse_string()) return self._parse_slice(self._parse_alias(self._parse_assignment(), explicit=True)) def _parse_lateral(self) -> t.Optional[exp.Lateral]: lateral = super()._parse_lateral() if not lateral: return lateral if isinstance(lateral.this, exp.Explode): table_alias = lateral.args.get("alias") columns = [exp.to_identifier(col) for col in self.FLATTEN_COLUMNS] if table_alias and not table_alias.args.get("columns"): table_alias.set("columns", columns) elif not table_alias: exp.alias_(lateral, "_flattened", table=columns, copy=False) return lateral def _parse_table_parts( self, schema: bool = False, is_db_reference: bool = False, wildcard: bool = False ) -> exp.Table: # https://docs.snowflake.com/en/user-guide/querying-stage if self._match(TokenType.STRING, advance=False): table = self._parse_string() elif self._match_text_seq("@", advance=False): table = self._parse_location_path() else: table = None if table: file_format = None pattern = None wrapped = self._match(TokenType.L_PAREN) while self._curr and wrapped and not self._match(TokenType.R_PAREN): if self._match_text_seq("FILE_FORMAT", "=>"): file_format = self._parse_string() or super()._parse_table_parts( is_db_reference=is_db_reference ) elif self._match_text_seq("PATTERN", "=>"): pattern = self._parse_string() else: break self._match(TokenType.COMMA) table = self.expression(exp.Table, this=table, format=file_format, pattern=pattern) else: table = super()._parse_table_parts(schema=schema, is_db_reference=is_db_reference) return table def _parse_id_var( self, any_token: bool = True, tokens: t.Optional[t.Collection[TokenType]] = None, ) -> t.Optional[exp.Expression]: if self._match_text_seq("IDENTIFIER", "("): identifier = ( super()._parse_id_var(any_token=any_token, tokens=tokens) or self._parse_string() ) self._match_r_paren() return self.expression(exp.Anonymous, this="IDENTIFIER", expressions=[identifier]) return super()._parse_id_var(any_token=any_token, tokens=tokens) def _parse_show_snowflake(self, this: str) -> exp.Show: scope = None scope_kind = None # will identity SHOW TERSE SCHEMAS but not SHOW TERSE PRIMARY KEYS # which is syntactically valid but has no effect on the output terse = self._tokens[self._index - 2].text.upper() == "TERSE" history = self._match_text_seq("HISTORY") like = self._parse_string() if self._match(TokenType.LIKE) else None if self._match(TokenType.IN): if self._match_text_seq("ACCOUNT"): scope_kind = "ACCOUNT" elif self._match_set(self.DB_CREATABLES): scope_kind = self._prev.text.upper() if self._curr: scope = self._parse_table_parts() elif self._curr: scope_kind = "SCHEMA" if this in self.SCHEMA_KINDS else "TABLE" scope = self._parse_table_parts() return self.expression( exp.Show, **{ "terse": terse, "this": this, "history": history, "like": like, "scope": scope, "scope_kind": scope_kind, "starts_with": self._match_text_seq("STARTS", "WITH") and self._parse_string(), "limit": self._parse_limit(), "from": self._parse_string() if self._match(TokenType.FROM) else None, }, ) def _parse_location_property(self) -> exp.LocationProperty: self._match(TokenType.EQ) return self.expression(exp.LocationProperty, this=self._parse_location_path()) def _parse_file_location(self) -> t.Optional[exp.Expression]: # Parse either a subquery or a staged file return ( self._parse_select(table=True, parse_subquery_alias=False) if self._match(TokenType.L_PAREN, advance=False) else self._parse_table_parts() ) def _parse_location_path(self) -> exp.Var: parts = [self._advance_any(ignore_reserved=True)] # We avoid consuming a comma token because external tables like @foo and @bar # can be joined in a query with a comma separator, as well as closing paren # in case of subqueries while self._is_connected() and not self._match_set( (TokenType.COMMA, TokenType.L_PAREN, TokenType.R_PAREN), advance=False ): parts.append(self._advance_any(ignore_reserved=True)) return exp.var("".join(part.text for part in parts if part)) def _parse_lambda_arg(self) -> t.Optional[exp.Expression]: this = super()._parse_lambda_arg() if not this: return this typ = self._parse_types() if typ: return self.expression(exp.Cast, this=this, to=typ) return this class Tokenizer(tokens.Tokenizer): STRING_ESCAPES = ["\\", "'"] HEX_STRINGS = [("x'", "'"), ("X'", "'")] RAW_STRINGS = ["$$"] COMMENTS = ["--", "//", ("/*", "*/")] NESTED_COMMENTS = False KEYWORDS = { **tokens.Tokenizer.KEYWORDS, "BYTEINT": TokenType.INT, "CHAR VARYING": TokenType.VARCHAR, "CHARACTER VARYING": TokenType.VARCHAR, "EXCLUDE": TokenType.EXCEPT, "ILIKE ANY": TokenType.ILIKE_ANY, "LIKE ANY": TokenType.LIKE_ANY, "MATCH_CONDITION": TokenType.MATCH_CONDITION, "MATCH_RECOGNIZE": TokenType.MATCH_RECOGNIZE, "MINUS": TokenType.EXCEPT, "NCHAR VARYING": TokenType.VARCHAR, "PUT": TokenType.COMMAND, "REMOVE": TokenType.COMMAND, "RM": TokenType.COMMAND, "SAMPLE": TokenType.TABLE_SAMPLE, "SQL_DOUBLE": TokenType.DOUBLE, "SQL_VARCHAR": TokenType.VARCHAR, "STORAGE INTEGRATION": TokenType.STORAGE_INTEGRATION, "TAG": TokenType.TAG, "TIMESTAMP_TZ": TokenType.TIMESTAMPTZ, "TOP": TokenType.TOP, "WAREHOUSE": TokenType.WAREHOUSE, "STREAMLIT": TokenType.STREAMLIT, } KEYWORDS.pop("/*+") SINGLE_TOKENS = { **tokens.Tokenizer.SINGLE_TOKENS, "$": TokenType.PARAMETER, } VAR_SINGLE_TOKENS = {"$"} COMMANDS = tokens.Tokenizer.COMMANDS - {TokenType.SHOW} class Generator(generator.Generator): PARAMETER_TOKEN = "$" MATCHED_BY_SOURCE = False SINGLE_STRING_INTERVAL = True JOIN_HINTS = False TABLE_HINTS = False QUERY_HINTS = False AGGREGATE_FILTER_SUPPORTED = False SUPPORTS_TABLE_COPY = False COLLATE_IS_FUNC = True LIMIT_ONLY_LITERALS = True JSON_KEY_VALUE_PAIR_SEP = "," INSERT_OVERWRITE = " OVERWRITE INTO" STRUCT_DELIMITER = ("(", ")") COPY_PARAMS_ARE_WRAPPED = False COPY_PARAMS_EQ_REQUIRED = True STAR_EXCEPT = "EXCLUDE" SUPPORTS_EXPLODING_PROJECTIONS = False ARRAY_CONCAT_IS_VAR_LEN = False SUPPORTS_CONVERT_TIMEZONE = True EXCEPT_INTERSECT_SUPPORT_ALL_CLAUSE = False SUPPORTS_MEDIAN = True ARRAY_SIZE_NAME = "ARRAY_SIZE" TRANSFORMS = { **generator.Generator.TRANSFORMS, exp.ApproxDistinct: rename_func("APPROX_COUNT_DISTINCT"), exp.ArgMax: rename_func("MAX_BY"), exp.ArgMin: rename_func("MIN_BY"), exp.Array: inline_array_sql, exp.ArrayConcat: lambda self, e: self.arrayconcat_sql(e, name="ARRAY_CAT"), exp.ArrayContains: lambda self, e: self.func("ARRAY_CONTAINS", e.expression, e.this), exp.AtTimeZone: lambda self, e: self.func( "CONVERT_TIMEZONE", e.args.get("zone"), e.this ), exp.BitwiseXor: rename_func("BITXOR"), exp.Create: transforms.preprocess([_flatten_structured_types_unless_iceberg]), exp.DateAdd: date_delta_sql("DATEADD"), exp.DateDiff: date_delta_sql("DATEDIFF"), exp.DatetimeAdd: date_delta_sql("TIMESTAMPADD"), exp.DatetimeDiff: timestampdiff_sql, exp.DateStrToDate: datestrtodate_sql, exp.DayOfMonth: rename_func("DAYOFMONTH"), exp.DayOfWeek: rename_func("DAYOFWEEK"), exp.DayOfYear: rename_func("DAYOFYEAR"), exp.Explode: rename_func("FLATTEN"), exp.Extract: rename_func("DATE_PART"), exp.FromTimeZone: lambda self, e: self.func( "CONVERT_TIMEZONE", e.args.get("zone"), "'UTC'", e.this ), exp.GenerateSeries: lambda self, e: self.func( "ARRAY_GENERATE_RANGE", e.args["start"], e.args["end"] + 1, e.args.get("step") ), exp.GroupConcat: rename_func("LISTAGG"), exp.If: if_sql(name="IFF", false_value="NULL"), exp.JSONExtractArray: _json_extract_value_array_sql, exp.JSONExtractScalar: lambda self, e: self.func( "JSON_EXTRACT_PATH_TEXT", e.this, e.expression ), exp.JSONObject: lambda self, e: self.func("OBJECT_CONSTRUCT_KEEP_NULL", *e.expressions), exp.JSONPathRoot: lambda *_: "", exp.JSONValueArray: _json_extract_value_array_sql, exp.LogicalAnd: rename_func("BOOLAND_AGG"), exp.LogicalOr: rename_func("BOOLOR_AGG"), exp.Map: lambda self, e: var_map_sql(self, e, "OBJECT_CONSTRUCT"), exp.MakeInterval: no_make_interval_sql, exp.Max: max_or_greatest, exp.Min: min_or_least, exp.ParseJSON: lambda self, e: self.func( "TRY_PARSE_JSON" if e.args.get("safe") else "PARSE_JSON", e.this ), exp.PartitionedByProperty: lambda self, e: f"PARTITION BY {self.sql(e, 'this')}", exp.PercentileCont: transforms.preprocess( [transforms.add_within_group_for_percentiles] ), exp.PercentileDisc: transforms.preprocess( [transforms.add_within_group_for_percentiles] ), exp.Pivot: transforms.preprocess([_unqualify_unpivot_columns]), exp.RegexpExtract: _regexpextract_sql, exp.RegexpExtractAll: _regexpextract_sql, exp.RegexpILike: _regexpilike_sql, exp.Rand: rename_func("RANDOM"), exp.Select: transforms.preprocess( [ transforms.eliminate_distinct_on, transforms.explode_to_unnest(), transforms.eliminate_semi_and_anti_joins, _transform_generate_date_array, ] ), exp.SafeDivide: lambda self, e: no_safe_divide_sql(self, e, "IFF"), exp.SHA: rename_func("SHA1"), exp.StarMap: rename_func("OBJECT_CONSTRUCT"), exp.StartsWith: rename_func("STARTSWITH"), exp.StrPosition: lambda self, e: self.func( "POSITION", e.args.get("substr"), e.this, e.args.get("position") ), exp.StrToDate: lambda self, e: self.func("DATE", e.this, self.format_time(e)), exp.Stuff: rename_func("INSERT"), exp.TimeAdd: date_delta_sql("TIMEADD"), exp.Timestamp: no_timestamp_sql, exp.TimestampAdd: date_delta_sql("TIMESTAMPADD"), exp.TimestampDiff: lambda self, e: self.func( "TIMESTAMPDIFF", e.unit, e.expression, e.this ), exp.TimestampTrunc: timestamptrunc_sql(), exp.TimeStrToTime: timestrtotime_sql, exp.TimeToUnix: lambda self, e: f"EXTRACT(epoch_second FROM {self.sql(e, 'this')})", exp.ToArray: rename_func("TO_ARRAY"), exp.ToChar: lambda self, e: self.function_fallback_sql(e), exp.ToDouble: rename_func("TO_DOUBLE"), exp.TsOrDsAdd: date_delta_sql("DATEADD", cast=True), exp.TsOrDsDiff: date_delta_sql("DATEDIFF"), exp.TsOrDsToDate: lambda self, e: self.func( "TRY_TO_DATE" if e.args.get("safe") else "TO_DATE", e.this, self.format_time(e) ), exp.UnixToTime: rename_func("TO_TIMESTAMP"), exp.Uuid: rename_func("UUID_STRING"), exp.VarMap: lambda self, e: var_map_sql(self, e, "OBJECT_CONSTRUCT"), exp.WeekOfYear: rename_func("WEEKOFYEAR"), exp.Xor: rename_func("BOOLXOR"), exp.Levenshtein: unsupported_args("ins_cost", "del_cost", "sub_cost")( rename_func("EDITDISTANCE") ), } SUPPORTED_JSON_PATH_PARTS = { exp.JSONPathKey, exp.JSONPathRoot, exp.JSONPathSubscript, } TYPE_MAPPING = { **generator.Generator.TYPE_MAPPING, exp.DataType.Type.NESTED: "OBJECT", exp.DataType.Type.STRUCT: "OBJECT", } PROPERTIES_LOCATION = { **generator.Generator.PROPERTIES_LOCATION, exp.SetProperty: exp.Properties.Location.UNSUPPORTED, exp.VolatileProperty: exp.Properties.Location.UNSUPPORTED, } UNSUPPORTED_VALUES_EXPRESSIONS = { exp.Map, exp.StarMap, exp.Struct, exp.VarMap, } def with_properties(self, properties: exp.Properties) -> str: return self.properties(properties, wrapped=False, prefix=self.sep(""), sep=" ") def values_sql(self, expression: exp.Values, values_as_table: bool = True) -> str: if expression.find(*self.UNSUPPORTED_VALUES_EXPRESSIONS): values_as_table = False return super().values_sql(expression, values_as_table=values_as_table) def datatype_sql(self, expression: exp.DataType) -> str: expressions = expression.expressions if ( expressions and expression.is_type(*exp.DataType.STRUCT_TYPES) and any(isinstance(field_type, exp.DataType) for field_type in expressions) ): # The correct syntax is OBJECT [ ( str: return self.func( "TO_NUMBER", expression.this, expression.args.get("format"), expression.args.get("precision"), expression.args.get("scale"), ) def timestampfromparts_sql(self, expression: exp.TimestampFromParts) -> str: milli = expression.args.get("milli") if milli is not None: milli_to_nano = milli.pop() * exp.Literal.number(1000000) expression.set("nano", milli_to_nano) return rename_func("TIMESTAMP_FROM_PARTS")(self, expression) def cast_sql(self, expression: exp.Cast, safe_prefix: t.Optional[str] = None) -> str: if expression.is_type(exp.DataType.Type.GEOGRAPHY): return self.func("TO_GEOGRAPHY", expression.this) if expression.is_type(exp.DataType.Type.GEOMETRY): return self.func("TO_GEOMETRY", expression.this) return super().cast_sql(expression, safe_prefix=safe_prefix) def trycast_sql(self, expression: exp.TryCast) -> str: value = expression.this if value.type is None: from sqlglot.optimizer.annotate_types import annotate_types value = annotate_types(value) if value.is_type(*exp.DataType.TEXT_TYPES, exp.DataType.Type.UNKNOWN): return super().trycast_sql(expression) # TRY_CAST only works for string values in Snowflake return self.cast_sql(expression) def log_sql(self, expression: exp.Log) -> str: if not expression.expression: return self.func("LN", expression.this) return super().log_sql(expression) def unnest_sql(self, expression: exp.Unnest) -> str: unnest_alias = expression.args.get("alias") offset = expression.args.get("offset") columns = [ exp.to_identifier("seq"), exp.to_identifier("key"), exp.to_identifier("path"), offset.pop() if isinstance(offset, exp.Expression) else exp.to_identifier("index"), seq_get(unnest_alias.columns if unnest_alias else [], 0) or exp.to_identifier("value"), exp.to_identifier("this"), ] if unnest_alias: unnest_alias.set("columns", columns) else: unnest_alias = exp.TableAlias(this="_u", columns=columns) explode = f"TABLE(FLATTEN(INPUT => {self.sql(expression.expressions[0])}))" alias = self.sql(unnest_alias) alias = f" AS {alias}" if alias else "" return f"{explode}{alias}" def show_sql(self, expression: exp.Show) -> str: terse = "TERSE " if expression.args.get("terse") else "" history = " HISTORY" if expression.args.get("history") else "" like = self.sql(expression, "like") like = f" LIKE {like}" if like else "" scope = self.sql(expression, "scope") scope = f" {scope}" if scope else "" scope_kind = self.sql(expression, "scope_kind") if scope_kind: scope_kind = f" IN {scope_kind}" starts_with = self.sql(expression, "starts_with") if starts_with: starts_with = f" STARTS WITH {starts_with}" limit = self.sql(expression, "limit") from_ = self.sql(expression, "from") if from_: from_ = f" FROM {from_}" return f"SHOW {terse}{expression.name}{history}{like}{scope_kind}{scope}{starts_with}{limit}{from_}" def describe_sql(self, expression: exp.Describe) -> str: # Default to table if kind is unknown kind_value = expression.args.get("kind") or "TABLE" kind = f" {kind_value}" if kind_value else "" this = f" {self.sql(expression, 'this')}" expressions = self.expressions(expression, flat=True) expressions = f" {expressions}" if expressions else "" return f"DESCRIBE{kind}{this}{expressions}" def generatedasidentitycolumnconstraint_sql( self, expression: exp.GeneratedAsIdentityColumnConstraint ) -> str: start = expression.args.get("start") start = f" START {start}" if start else "" increment = expression.args.get("increment") increment = f" INCREMENT {increment}" if increment else "" return f"AUTOINCREMENT{start}{increment}" def cluster_sql(self, expression: exp.Cluster) -> str: return f"CLUSTER BY ({self.expressions(expression, flat=True)})" def struct_sql(self, expression: exp.Struct) -> str: keys = [] values = [] for i, e in enumerate(expression.expressions): if isinstance(e, exp.PropertyEQ): keys.append( exp.Literal.string(e.name) if isinstance(e.this, exp.Identifier) else e.this ) values.append(e.expression) else: keys.append(exp.Literal.string(f"_{i}")) values.append(e) return self.func("OBJECT_CONSTRUCT", *flatten(zip(keys, values))) @unsupported_args("weight", "accuracy") def approxquantile_sql(self, expression: exp.ApproxQuantile) -> str: return self.func("APPROX_PERCENTILE", expression.this, expression.args.get("quantile")) def alterset_sql(self, expression: exp.AlterSet) -> str: exprs = self.expressions(expression, flat=True) exprs = f" {exprs}" if exprs else "" file_format = self.expressions(expression, key="file_format", flat=True, sep=" ") file_format = f" STAGE_FILE_FORMAT = ({file_format})" if file_format else "" copy_options = self.expressions(expression, key="copy_options", flat=True, sep=" ") copy_options = f" STAGE_COPY_OPTIONS = ({copy_options})" if copy_options else "" tag = self.expressions(expression, key="tag", flat=True) tag = f" TAG {tag}" if tag else "" return f"SET{exprs}{file_format}{copy_options}{tag}" def strtotime_sql(self, expression: exp.StrToTime): safe_prefix = "TRY_" if expression.args.get("safe") else "" return self.func( f"{safe_prefix}TO_TIMESTAMP", expression.this, self.format_time(expression) ) def timestampsub_sql(self, expression: exp.TimestampSub): return self.sql( exp.TimestampAdd( this=expression.this, expression=expression.expression * -1, unit=expression.unit, ) ) def jsonextract_sql(self, expression: exp.JSONExtract): this = expression.this # JSON strings are valid coming from other dialects such as BQ return self.func( "GET_PATH", exp.ParseJSON(this=this) if this.is_string else this, expression.expression, ) def timetostr_sql(self, expression: exp.TimeToStr) -> str: this = expression.this if not isinstance(this, exp.TsOrDsToTimestamp): this = exp.cast(this, exp.DataType.Type.TIMESTAMP) return self.func("TO_CHAR", this, self.format_time(expression))