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