sqlglot.dialects.snowflake

  1from __future__ import annotations
  2
  3import typing as t
  4
  5from sqlglot import exp, generator, parser, tokens, transforms
  6from sqlglot.dialects.dialect import (
  7    Dialect,
  8    binary_from_function,
  9    date_trunc_to_time,
 10    datestrtodate_sql,
 11    format_time_lambda,
 12    if_sql,
 13    inline_array_sql,
 14    max_or_greatest,
 15    min_or_least,
 16    rename_func,
 17    timestamptrunc_sql,
 18    timestrtotime_sql,
 19    ts_or_ds_to_date_sql,
 20    var_map_sql,
 21)
 22from sqlglot.expressions import Literal
 23from sqlglot.helper import seq_get
 24from sqlglot.parser import binary_range_parser
 25from sqlglot.tokens import TokenType
 26
 27
 28def _check_int(s: str) -> bool:
 29    if s[0] in ("-", "+"):
 30        return s[1:].isdigit()
 31    return s.isdigit()
 32
 33
 34# from https://docs.snowflake.com/en/sql-reference/functions/to_timestamp.html
 35def _parse_to_timestamp(args: t.List) -> t.Union[exp.StrToTime, exp.UnixToTime]:
 36    if len(args) == 2:
 37        first_arg, second_arg = args
 38        if second_arg.is_string:
 39            # case: <string_expr> [ , <format> ]
 40            return format_time_lambda(exp.StrToTime, "snowflake")(args)
 41
 42        # case: <numeric_expr> [ , <scale> ]
 43        if second_arg.name not in ["0", "3", "9"]:
 44            raise ValueError(
 45                f"Scale for snowflake numeric timestamp is {second_arg}, but should be 0, 3, or 9"
 46            )
 47
 48        if second_arg.name == "0":
 49            timescale = exp.UnixToTime.SECONDS
 50        elif second_arg.name == "3":
 51            timescale = exp.UnixToTime.MILLIS
 52        elif second_arg.name == "9":
 53            timescale = exp.UnixToTime.MICROS
 54
 55        return exp.UnixToTime(this=first_arg, scale=timescale)
 56
 57    from sqlglot.optimizer.simplify import simplify_literals
 58
 59    # The first argument might be an expression like 40 * 365 * 86400, so we try to
 60    # reduce it using `simplify_literals` first and then check if it's a Literal.
 61    first_arg = seq_get(args, 0)
 62    if not isinstance(simplify_literals(first_arg, root=True), Literal):
 63        # case: <variant_expr>
 64        return format_time_lambda(exp.StrToTime, "snowflake", default=True)(args)
 65
 66    if first_arg.is_string:
 67        if _check_int(first_arg.this):
 68            # case: <integer>
 69            return exp.UnixToTime.from_arg_list(args)
 70
 71        # case: <date_expr>
 72        return format_time_lambda(exp.StrToTime, "snowflake", default=True)(args)
 73
 74    # case: <numeric_expr>
 75    return exp.UnixToTime.from_arg_list(args)
 76
 77
 78def _parse_object_construct(args: t.List) -> t.Union[exp.StarMap, exp.Struct]:
 79    expression = parser.parse_var_map(args)
 80
 81    if isinstance(expression, exp.StarMap):
 82        return expression
 83
 84    return exp.Struct(
 85        expressions=[
 86            t.cast(exp.Condition, k).eq(v) for k, v in zip(expression.keys, expression.values)
 87        ]
 88    )
 89
 90
 91def _parse_datediff(args: t.List) -> exp.DateDiff:
 92    return exp.DateDiff(this=seq_get(args, 2), expression=seq_get(args, 1), unit=seq_get(args, 0))
 93
 94
 95def _unix_to_time_sql(self: Snowflake.Generator, expression: exp.UnixToTime) -> str:
 96    scale = expression.args.get("scale")
 97    timestamp = self.sql(expression, "this")
 98    if scale in [None, exp.UnixToTime.SECONDS]:
 99        return f"TO_TIMESTAMP({timestamp})"
100    if scale == exp.UnixToTime.MILLIS:
101        return f"TO_TIMESTAMP({timestamp}, 3)"
102    if scale == exp.UnixToTime.MICROS:
103        return f"TO_TIMESTAMP({timestamp}, 9)"
104
105    raise ValueError("Improper scale for timestamp")
106
107
108# https://docs.snowflake.com/en/sql-reference/functions/date_part.html
109# https://docs.snowflake.com/en/sql-reference/functions-date-time.html#label-supported-date-time-parts
110def _parse_date_part(self: Snowflake.Parser) -> t.Optional[exp.Expression]:
111    this = self._parse_var() or self._parse_type()
112
113    if not this:
114        return None
115
116    self._match(TokenType.COMMA)
117    expression = self._parse_bitwise()
118
119    name = this.name.upper()
120    if name.startswith("EPOCH"):
121        if name.startswith("EPOCH_MILLISECOND"):
122            scale = 10**3
123        elif name.startswith("EPOCH_MICROSECOND"):
124            scale = 10**6
125        elif name.startswith("EPOCH_NANOSECOND"):
126            scale = 10**9
127        else:
128            scale = None
129
130        ts = self.expression(exp.Cast, this=expression, to=exp.DataType.build("TIMESTAMP"))
131        to_unix: exp.Expression = self.expression(exp.TimeToUnix, this=ts)
132
133        if scale:
134            to_unix = exp.Mul(this=to_unix, expression=exp.Literal.number(scale))
135
136        return to_unix
137
138    return self.expression(exp.Extract, this=this, expression=expression)
139
140
141# https://docs.snowflake.com/en/sql-reference/functions/div0
142def _div0_to_if(args: t.List) -> exp.If:
143    cond = exp.EQ(this=seq_get(args, 1), expression=exp.Literal.number(0))
144    true = exp.Literal.number(0)
145    false = exp.Div(this=seq_get(args, 0), expression=seq_get(args, 1))
146    return exp.If(this=cond, true=true, false=false)
147
148
149# https://docs.snowflake.com/en/sql-reference/functions/zeroifnull
150def _zeroifnull_to_if(args: t.List) -> exp.If:
151    cond = exp.Is(this=seq_get(args, 0), expression=exp.Null())
152    return exp.If(this=cond, true=exp.Literal.number(0), false=seq_get(args, 0))
153
154
155# https://docs.snowflake.com/en/sql-reference/functions/zeroifnull
156def _nullifzero_to_if(args: t.List) -> exp.If:
157    cond = exp.EQ(this=seq_get(args, 0), expression=exp.Literal.number(0))
158    return exp.If(this=cond, true=exp.Null(), false=seq_get(args, 0))
159
160
161def _datatype_sql(self: Snowflake.Generator, expression: exp.DataType) -> str:
162    if expression.is_type("array"):
163        return "ARRAY"
164    elif expression.is_type("map"):
165        return "OBJECT"
166    return self.datatype_sql(expression)
167
168
169def _regexpilike_sql(self: Snowflake.Generator, expression: exp.RegexpILike) -> str:
170    flag = expression.text("flag")
171
172    if "i" not in flag:
173        flag += "i"
174
175    return self.func(
176        "REGEXP_LIKE", expression.this, expression.expression, exp.Literal.string(flag)
177    )
178
179
180def _parse_convert_timezone(args: t.List) -> t.Union[exp.Anonymous, exp.AtTimeZone]:
181    if len(args) == 3:
182        return exp.Anonymous(this="CONVERT_TIMEZONE", expressions=args)
183    return exp.AtTimeZone(this=seq_get(args, 1), zone=seq_get(args, 0))
184
185
186def _parse_regexp_replace(args: t.List) -> exp.RegexpReplace:
187    regexp_replace = exp.RegexpReplace.from_arg_list(args)
188
189    if not regexp_replace.args.get("replacement"):
190        regexp_replace.set("replacement", exp.Literal.string(""))
191
192    return regexp_replace
193
194
195def _show_parser(*args: t.Any, **kwargs: t.Any) -> t.Callable[[Snowflake.Parser], exp.Show]:
196    def _parse(self: Snowflake.Parser) -> exp.Show:
197        return self._parse_show_snowflake(*args, **kwargs)
198
199    return _parse
200
201
202class Snowflake(Dialect):
203    # https://docs.snowflake.com/en/sql-reference/identifiers-syntax
204    RESOLVES_IDENTIFIERS_AS_UPPERCASE = True
205    NULL_ORDERING = "nulls_are_large"
206    TIME_FORMAT = "'YYYY-MM-DD HH24:MI:SS'"
207    SUPPORTS_USER_DEFINED_TYPES = False
208    SUPPORTS_SEMI_ANTI_JOIN = False
209
210    TIME_MAPPING = {
211        "YYYY": "%Y",
212        "yyyy": "%Y",
213        "YY": "%y",
214        "yy": "%y",
215        "MMMM": "%B",
216        "mmmm": "%B",
217        "MON": "%b",
218        "mon": "%b",
219        "MM": "%m",
220        "mm": "%m",
221        "DD": "%d",
222        "dd": "%-d",
223        "DY": "%a",
224        "dy": "%w",
225        "HH24": "%H",
226        "hh24": "%H",
227        "HH12": "%I",
228        "hh12": "%I",
229        "MI": "%M",
230        "mi": "%M",
231        "SS": "%S",
232        "ss": "%S",
233        "FF": "%f",
234        "ff": "%f",
235        "FF6": "%f",
236        "ff6": "%f",
237    }
238
239    class Parser(parser.Parser):
240        IDENTIFY_PIVOT_STRINGS = True
241
242        FUNCTIONS = {
243            **parser.Parser.FUNCTIONS,
244            "ARRAYAGG": exp.ArrayAgg.from_arg_list,
245            "ARRAY_CONSTRUCT": exp.Array.from_arg_list,
246            "ARRAY_GENERATE_RANGE": lambda args: exp.GenerateSeries(
247                # ARRAY_GENERATE_RANGE has an exlusive end; we normalize it to be inclusive
248                start=seq_get(args, 0),
249                end=exp.Sub(this=seq_get(args, 1), expression=exp.Literal.number(1)),
250                step=seq_get(args, 2),
251            ),
252            "ARRAY_TO_STRING": exp.ArrayJoin.from_arg_list,
253            "BITXOR": binary_from_function(exp.BitwiseXor),
254            "BIT_XOR": binary_from_function(exp.BitwiseXor),
255            "BOOLXOR": binary_from_function(exp.Xor),
256            "CONVERT_TIMEZONE": _parse_convert_timezone,
257            "DATE_TRUNC": date_trunc_to_time,
258            "DATEADD": lambda args: exp.DateAdd(
259                this=seq_get(args, 2), expression=seq_get(args, 1), unit=seq_get(args, 0)
260            ),
261            "DATEDIFF": _parse_datediff,
262            "DIV0": _div0_to_if,
263            "IFF": exp.If.from_arg_list,
264            "LISTAGG": exp.GroupConcat.from_arg_list,
265            "NULLIFZERO": _nullifzero_to_if,
266            "OBJECT_CONSTRUCT": _parse_object_construct,
267            "REGEXP_REPLACE": _parse_regexp_replace,
268            "REGEXP_SUBSTR": exp.RegexpExtract.from_arg_list,
269            "RLIKE": exp.RegexpLike.from_arg_list,
270            "SQUARE": lambda args: exp.Pow(this=seq_get(args, 0), expression=exp.Literal.number(2)),
271            "TIMEDIFF": _parse_datediff,
272            "TIMESTAMPDIFF": _parse_datediff,
273            "TO_ARRAY": exp.Array.from_arg_list,
274            "TO_TIMESTAMP": _parse_to_timestamp,
275            "TO_VARCHAR": exp.ToChar.from_arg_list,
276            "ZEROIFNULL": _zeroifnull_to_if,
277        }
278
279        FUNCTION_PARSERS = {
280            **parser.Parser.FUNCTION_PARSERS,
281            "DATE_PART": _parse_date_part,
282        }
283        FUNCTION_PARSERS.pop("TRIM")
284
285        COLUMN_OPERATORS = {
286            **parser.Parser.COLUMN_OPERATORS,
287            TokenType.COLON: lambda self, this, path: self.expression(
288                exp.Bracket, this=this, expressions=[path]
289            ),
290        }
291
292        TIMESTAMPS = parser.Parser.TIMESTAMPS - {TokenType.TIME}
293
294        RANGE_PARSERS = {
295            **parser.Parser.RANGE_PARSERS,
296            TokenType.LIKE_ANY: binary_range_parser(exp.LikeAny),
297            TokenType.ILIKE_ANY: binary_range_parser(exp.ILikeAny),
298        }
299
300        ALTER_PARSERS = {
301            **parser.Parser.ALTER_PARSERS,
302            "SET": lambda self: self._parse_set(tag=self._match_text_seq("TAG")),
303            "UNSET": lambda self: self.expression(
304                exp.Set,
305                tag=self._match_text_seq("TAG"),
306                expressions=self._parse_csv(self._parse_id_var),
307                unset=True,
308            ),
309        }
310
311        STATEMENT_PARSERS = {
312            **parser.Parser.STATEMENT_PARSERS,
313            TokenType.SHOW: lambda self: self._parse_show(),
314        }
315
316        SHOW_PARSERS = {
317            "PRIMARY KEYS": _show_parser("PRIMARY KEYS"),
318            "TERSE PRIMARY KEYS": _show_parser("PRIMARY KEYS"),
319        }
320
321        def _parse_id_var(
322            self,
323            any_token: bool = True,
324            tokens: t.Optional[t.Collection[TokenType]] = None,
325        ) -> t.Optional[exp.Expression]:
326            if self._match_text_seq("IDENTIFIER", "("):
327                identifier = (
328                    super()._parse_id_var(any_token=any_token, tokens=tokens)
329                    or self._parse_string()
330                )
331                self._match_r_paren()
332                return self.expression(exp.Anonymous, this="IDENTIFIER", expressions=[identifier])
333
334            return super()._parse_id_var(any_token=any_token, tokens=tokens)
335
336        def _parse_show_snowflake(self, this: str) -> exp.Show:
337            scope = None
338            scope_kind = None
339
340            if self._match(TokenType.IN):
341                if self._match_text_seq("ACCOUNT"):
342                    scope_kind = "ACCOUNT"
343                elif self._match_set(self.DB_CREATABLES):
344                    scope_kind = self._prev.text
345                    if self._curr:
346                        scope = self._parse_table()
347                elif self._curr:
348                    scope_kind = "TABLE"
349                    scope = self._parse_table()
350
351            return self.expression(exp.Show, this=this, scope=scope, scope_kind=scope_kind)
352
353    class Tokenizer(tokens.Tokenizer):
354        STRING_ESCAPES = ["\\", "'"]
355        HEX_STRINGS = [("x'", "'"), ("X'", "'")]
356        RAW_STRINGS = ["$$"]
357        COMMENTS = ["--", "//", ("/*", "*/")]
358
359        KEYWORDS = {
360            **tokens.Tokenizer.KEYWORDS,
361            "BYTEINT": TokenType.INT,
362            "CHAR VARYING": TokenType.VARCHAR,
363            "CHARACTER VARYING": TokenType.VARCHAR,
364            "EXCLUDE": TokenType.EXCEPT,
365            "ILIKE ANY": TokenType.ILIKE_ANY,
366            "LIKE ANY": TokenType.LIKE_ANY,
367            "MATCH_RECOGNIZE": TokenType.MATCH_RECOGNIZE,
368            "MINUS": TokenType.EXCEPT,
369            "NCHAR VARYING": TokenType.VARCHAR,
370            "PUT": TokenType.COMMAND,
371            "RENAME": TokenType.REPLACE,
372            "SAMPLE": TokenType.TABLE_SAMPLE,
373            "TIMESTAMP_LTZ": TokenType.TIMESTAMPLTZ,
374            "TIMESTAMP_NTZ": TokenType.TIMESTAMP,
375            "TIMESTAMP_TZ": TokenType.TIMESTAMPTZ,
376            "TIMESTAMPNTZ": TokenType.TIMESTAMP,
377            "TOP": TokenType.TOP,
378        }
379
380        SINGLE_TOKENS = {
381            **tokens.Tokenizer.SINGLE_TOKENS,
382            "$": TokenType.PARAMETER,
383        }
384
385        VAR_SINGLE_TOKENS = {"$"}
386
387        COMMANDS = tokens.Tokenizer.COMMANDS - {TokenType.SHOW}
388
389    class Generator(generator.Generator):
390        PARAMETER_TOKEN = "$"
391        MATCHED_BY_SOURCE = False
392        SINGLE_STRING_INTERVAL = True
393        JOIN_HINTS = False
394        TABLE_HINTS = False
395        QUERY_HINTS = False
396        AGGREGATE_FILTER_SUPPORTED = False
397
398        TRANSFORMS = {
399            **generator.Generator.TRANSFORMS,
400            exp.Array: inline_array_sql,
401            exp.ArrayConcat: rename_func("ARRAY_CAT"),
402            exp.ArrayJoin: rename_func("ARRAY_TO_STRING"),
403            exp.AtTimeZone: lambda self, e: self.func(
404                "CONVERT_TIMEZONE", e.args.get("zone"), e.this
405            ),
406            exp.BitwiseXor: rename_func("BITXOR"),
407            exp.DateAdd: lambda self, e: self.func("DATEADD", e.text("unit"), e.expression, e.this),
408            exp.DateDiff: lambda self, e: self.func(
409                "DATEDIFF", e.text("unit"), e.expression, e.this
410            ),
411            exp.DateStrToDate: datestrtodate_sql,
412            exp.DataType: _datatype_sql,
413            exp.DayOfWeek: rename_func("DAYOFWEEK"),
414            exp.Extract: rename_func("DATE_PART"),
415            exp.GenerateSeries: lambda self, e: self.func(
416                "ARRAY_GENERATE_RANGE", e.args["start"], e.args["end"] + 1, e.args.get("step")
417            ),
418            exp.GroupConcat: rename_func("LISTAGG"),
419            exp.If: if_sql(name="IFF", false_value="NULL"),
420            exp.LogicalAnd: rename_func("BOOLAND_AGG"),
421            exp.LogicalOr: rename_func("BOOLOR_AGG"),
422            exp.Map: lambda self, e: var_map_sql(self, e, "OBJECT_CONSTRUCT"),
423            exp.Max: max_or_greatest,
424            exp.Min: min_or_least,
425            exp.PartitionedByProperty: lambda self, e: f"PARTITION BY {self.sql(e, 'this')}",
426            exp.RegexpILike: _regexpilike_sql,
427            exp.Select: transforms.preprocess(
428                [
429                    transforms.eliminate_distinct_on,
430                    transforms.explode_to_unnest(0),
431                    transforms.eliminate_semi_and_anti_joins,
432                ]
433            ),
434            exp.StarMap: rename_func("OBJECT_CONSTRUCT"),
435            exp.StartsWith: rename_func("STARTSWITH"),
436            exp.StrPosition: lambda self, e: self.func(
437                "POSITION", e.args.get("substr"), e.this, e.args.get("position")
438            ),
439            exp.StrToTime: lambda self, e: f"TO_TIMESTAMP({self.sql(e, 'this')}, {self.format_time(e)})",
440            exp.Struct: lambda self, e: self.func(
441                "OBJECT_CONSTRUCT",
442                *(arg for expression in e.expressions for arg in expression.flatten()),
443            ),
444            exp.Stuff: rename_func("INSERT"),
445            exp.TimestampTrunc: timestamptrunc_sql,
446            exp.TimeStrToTime: timestrtotime_sql,
447            exp.TimeToStr: lambda self, e: self.func(
448                "TO_CHAR", exp.cast(e.this, "timestamp"), self.format_time(e)
449            ),
450            exp.TimeToUnix: lambda self, e: f"EXTRACT(epoch_second FROM {self.sql(e, 'this')})",
451            exp.ToChar: lambda self, e: self.function_fallback_sql(e),
452            exp.Trim: lambda self, e: self.func("TRIM", e.this, e.expression),
453            exp.TsOrDsToDate: ts_or_ds_to_date_sql("snowflake"),
454            exp.UnixToTime: _unix_to_time_sql,
455            exp.VarMap: lambda self, e: var_map_sql(self, e, "OBJECT_CONSTRUCT"),
456            exp.WeekOfYear: rename_func("WEEKOFYEAR"),
457            exp.Xor: rename_func("BOOLXOR"),
458        }
459
460        TYPE_MAPPING = {
461            **generator.Generator.TYPE_MAPPING,
462            exp.DataType.Type.TIMESTAMP: "TIMESTAMPNTZ",
463        }
464
465        STAR_MAPPING = {
466            "except": "EXCLUDE",
467            "replace": "RENAME",
468        }
469
470        PROPERTIES_LOCATION = {
471            **generator.Generator.PROPERTIES_LOCATION,
472            exp.SetProperty: exp.Properties.Location.UNSUPPORTED,
473            exp.VolatileProperty: exp.Properties.Location.UNSUPPORTED,
474        }
475
476        def unnest_sql(self, expression: exp.Unnest) -> str:
477            selects = ["value"]
478            unnest_alias = expression.args.get("alias")
479
480            offset = expression.args.get("offset")
481            if offset:
482                if unnest_alias:
483                    expression = expression.copy()
484                    unnest_alias.append("columns", offset.pop())
485
486                selects.append("index")
487
488            subquery = exp.Subquery(
489                this=exp.select(*selects).from_(
490                    f"TABLE(FLATTEN(INPUT => {self.sql(expression.expressions[0])}))"
491                ),
492            )
493            alias = self.sql(unnest_alias)
494            alias = f" AS {alias}" if alias else ""
495            return f"{self.sql(subquery)}{alias}"
496
497        def show_sql(self, expression: exp.Show) -> str:
498            scope = self.sql(expression, "scope")
499            scope = f" {scope}" if scope else ""
500
501            scope_kind = self.sql(expression, "scope_kind")
502            if scope_kind:
503                scope_kind = f" IN {scope_kind}"
504
505            return f"SHOW {expression.name}{scope_kind}{scope}"
506
507        def regexpextract_sql(self, expression: exp.RegexpExtract) -> str:
508            # Other dialects don't support all of the following parameters, so we need to
509            # generate default values as necessary to ensure the transpilation is correct
510            group = expression.args.get("group")
511            parameters = expression.args.get("parameters") or (group and exp.Literal.string("c"))
512            occurrence = expression.args.get("occurrence") or (parameters and exp.Literal.number(1))
513            position = expression.args.get("position") or (occurrence and exp.Literal.number(1))
514
515            return self.func(
516                "REGEXP_SUBSTR",
517                expression.this,
518                expression.expression,
519                position,
520                occurrence,
521                parameters,
522                group,
523            )
524
525        def except_op(self, expression: exp.Except) -> str:
526            if not expression.args.get("distinct", False):
527                self.unsupported("EXCEPT with All is not supported in Snowflake")
528            return super().except_op(expression)
529
530        def intersect_op(self, expression: exp.Intersect) -> str:
531            if not expression.args.get("distinct", False):
532                self.unsupported("INTERSECT with All is not supported in Snowflake")
533            return super().intersect_op(expression)
534
535        def describe_sql(self, expression: exp.Describe) -> str:
536            # Default to table if kind is unknown
537            kind_value = expression.args.get("kind") or "TABLE"
538            kind = f" {kind_value}" if kind_value else ""
539            this = f" {self.sql(expression, 'this')}"
540            expressions = self.expressions(expression, flat=True)
541            expressions = f" {expressions}" if expressions else ""
542            return f"DESCRIBE{kind}{this}{expressions}"
543
544        def generatedasidentitycolumnconstraint_sql(
545            self, expression: exp.GeneratedAsIdentityColumnConstraint
546        ) -> str:
547            start = expression.args.get("start")
548            start = f" START {start}" if start else ""
549            increment = expression.args.get("increment")
550            increment = f" INCREMENT {increment}" if increment else ""
551            return f"AUTOINCREMENT{start}{increment}"