sqlglot.dialects.presto

  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    encode_decode_sql,
 11    format_time_lambda,
 12    if_sql,
 13    left_to_substring_sql,
 14    no_ilike_sql,
 15    no_pivot_sql,
 16    no_safe_divide_sql,
 17    regexp_extract_sql,
 18    rename_func,
 19    right_to_substring_sql,
 20    struct_extract_sql,
 21    timestamptrunc_sql,
 22    timestrtotime_sql,
 23)
 24from sqlglot.dialects.mysql import MySQL
 25from sqlglot.helper import apply_index_offset, seq_get
 26from sqlglot.tokens import TokenType
 27
 28
 29def _approx_distinct_sql(self: generator.Generator, expression: exp.ApproxDistinct) -> str:
 30    accuracy = expression.args.get("accuracy")
 31    accuracy = ", " + self.sql(accuracy) if accuracy else ""
 32    return f"APPROX_DISTINCT({self.sql(expression, 'this')}{accuracy})"
 33
 34
 35def _datatype_sql(self: generator.Generator, expression: exp.DataType) -> str:
 36    sql = self.datatype_sql(expression)
 37    if expression.is_type("timestamptz"):
 38        sql = f"{sql} WITH TIME ZONE"
 39    return sql
 40
 41
 42def _explode_to_unnest_sql(self: generator.Generator, expression: exp.Lateral) -> str:
 43    if isinstance(expression.this, (exp.Explode, exp.Posexplode)):
 44        expression = expression.copy()
 45        return self.sql(
 46            exp.Join(
 47                this=exp.Unnest(
 48                    expressions=[expression.this.this],
 49                    alias=expression.args.get("alias"),
 50                    ordinality=isinstance(expression.this, exp.Posexplode),
 51                ),
 52                kind="cross",
 53            )
 54        )
 55    return self.lateral_sql(expression)
 56
 57
 58def _initcap_sql(self: generator.Generator, expression: exp.Initcap) -> str:
 59    regex = r"(\w)(\w*)"
 60    return f"REGEXP_REPLACE({self.sql(expression, 'this')}, '{regex}', x -> UPPER(x[1]) || LOWER(x[2]))"
 61
 62
 63def _no_sort_array(self: generator.Generator, expression: exp.SortArray) -> str:
 64    if expression.args.get("asc") == exp.false():
 65        comparator = "(a, b) -> CASE WHEN a < b THEN 1 WHEN a > b THEN -1 ELSE 0 END"
 66    else:
 67        comparator = None
 68    return self.func("ARRAY_SORT", expression.this, comparator)
 69
 70
 71def _schema_sql(self: generator.Generator, expression: exp.Schema) -> str:
 72    if isinstance(expression.parent, exp.Property):
 73        columns = ", ".join(f"'{c.name}'" for c in expression.expressions)
 74        return f"ARRAY[{columns}]"
 75
 76    if expression.parent:
 77        for schema in expression.parent.find_all(exp.Schema):
 78            if isinstance(schema.parent, exp.Property):
 79                expression = expression.copy()
 80                expression.expressions.extend(schema.expressions)
 81
 82    return self.schema_sql(expression)
 83
 84
 85def _quantile_sql(self: generator.Generator, expression: exp.Quantile) -> str:
 86    self.unsupported("Presto does not support exact quantiles")
 87    return f"APPROX_PERCENTILE({self.sql(expression, 'this')}, {self.sql(expression, 'quantile')})"
 88
 89
 90def _str_to_time_sql(
 91    self: generator.Generator, expression: exp.StrToDate | exp.StrToTime | exp.TsOrDsToDate
 92) -> str:
 93    return f"DATE_PARSE({self.sql(expression, 'this')}, {self.format_time(expression)})"
 94
 95
 96def _ts_or_ds_to_date_sql(self: generator.Generator, expression: exp.TsOrDsToDate) -> str:
 97    time_format = self.format_time(expression)
 98    if time_format and time_format not in (Presto.TIME_FORMAT, Presto.DATE_FORMAT):
 99        return exp.cast(_str_to_time_sql(self, expression), "DATE").sql(dialect="presto")
100    return exp.cast(exp.cast(expression.this, "TIMESTAMP", copy=True), "DATE").sql(dialect="presto")
101
102
103def _ts_or_ds_add_sql(self: generator.Generator, expression: exp.TsOrDsAdd) -> str:
104    this = expression.this
105
106    if not isinstance(this, exp.CurrentDate):
107        this = exp.cast(exp.cast(expression.this, "TIMESTAMP", copy=True), "DATE")
108
109    return self.func(
110        "DATE_ADD",
111        exp.Literal.string(expression.text("unit") or "day"),
112        expression.expression,
113        this,
114    )
115
116
117def _approx_percentile(args: t.List) -> exp.Expression:
118    if len(args) == 4:
119        return exp.ApproxQuantile(
120            this=seq_get(args, 0),
121            weight=seq_get(args, 1),
122            quantile=seq_get(args, 2),
123            accuracy=seq_get(args, 3),
124        )
125    if len(args) == 3:
126        return exp.ApproxQuantile(
127            this=seq_get(args, 0), quantile=seq_get(args, 1), accuracy=seq_get(args, 2)
128        )
129    return exp.ApproxQuantile.from_arg_list(args)
130
131
132def _from_unixtime(args: t.List) -> exp.Expression:
133    if len(args) == 3:
134        return exp.UnixToTime(
135            this=seq_get(args, 0),
136            hours=seq_get(args, 1),
137            minutes=seq_get(args, 2),
138        )
139    if len(args) == 2:
140        return exp.UnixToTime(this=seq_get(args, 0), zone=seq_get(args, 1))
141
142    return exp.UnixToTime.from_arg_list(args)
143
144
145def _parse_element_at(args: t.List) -> exp.SafeBracket:
146    this = seq_get(args, 0)
147    index = seq_get(args, 1)
148    assert isinstance(this, exp.Expression) and isinstance(index, exp.Expression)
149    return exp.SafeBracket(this=this, expressions=apply_index_offset(this, [index], -1))
150
151
152def _unnest_sequence(expression: exp.Expression) -> exp.Expression:
153    if isinstance(expression, exp.Table):
154        if isinstance(expression.this, exp.GenerateSeries):
155            unnest = exp.Unnest(expressions=[expression.this])
156
157            if expression.alias:
158                return exp.alias_(unnest, alias="_u", table=[expression.alias], copy=False)
159            return unnest
160    return expression
161
162
163class Presto(Dialect):
164    INDEX_OFFSET = 1
165    NULL_ORDERING = "nulls_are_last"
166    TIME_FORMAT = MySQL.TIME_FORMAT
167    TIME_MAPPING = MySQL.TIME_MAPPING
168    STRICT_STRING_CONCAT = True
169
170    # https://github.com/trinodb/trino/issues/17
171    # https://github.com/trinodb/trino/issues/12289
172    # https://github.com/prestodb/presto/issues/2863
173    RESOLVES_IDENTIFIERS_AS_UPPERCASE = None
174
175    class Tokenizer(tokens.Tokenizer):
176        KEYWORDS = {
177            **tokens.Tokenizer.KEYWORDS,
178            "START": TokenType.BEGIN,
179            "MATCH_RECOGNIZE": TokenType.MATCH_RECOGNIZE,
180            "ROW": TokenType.STRUCT,
181            "IPADDRESS": TokenType.IPADDRESS,
182            "IPPREFIX": TokenType.IPPREFIX,
183        }
184
185    class Parser(parser.Parser):
186        FUNCTIONS = {
187            **parser.Parser.FUNCTIONS,
188            "APPROX_DISTINCT": exp.ApproxDistinct.from_arg_list,
189            "APPROX_PERCENTILE": _approx_percentile,
190            "BITWISE_AND": binary_from_function(exp.BitwiseAnd),
191            "BITWISE_NOT": lambda args: exp.BitwiseNot(this=seq_get(args, 0)),
192            "BITWISE_OR": binary_from_function(exp.BitwiseOr),
193            "BITWISE_XOR": binary_from_function(exp.BitwiseXor),
194            "CARDINALITY": exp.ArraySize.from_arg_list,
195            "CONTAINS": exp.ArrayContains.from_arg_list,
196            "DATE_ADD": lambda args: exp.DateAdd(
197                this=seq_get(args, 2), expression=seq_get(args, 1), unit=seq_get(args, 0)
198            ),
199            "DATE_DIFF": lambda args: exp.DateDiff(
200                this=seq_get(args, 2), expression=seq_get(args, 1), unit=seq_get(args, 0)
201            ),
202            "DATE_FORMAT": format_time_lambda(exp.TimeToStr, "presto"),
203            "DATE_PARSE": format_time_lambda(exp.StrToTime, "presto"),
204            "DATE_TRUNC": date_trunc_to_time,
205            "ELEMENT_AT": _parse_element_at,
206            "FROM_HEX": exp.Unhex.from_arg_list,
207            "FROM_UNIXTIME": _from_unixtime,
208            "FROM_UTF8": lambda args: exp.Decode(
209                this=seq_get(args, 0), replace=seq_get(args, 1), charset=exp.Literal.string("utf-8")
210            ),
211            "NOW": exp.CurrentTimestamp.from_arg_list,
212            "REGEXP_EXTRACT": lambda args: exp.RegexpExtract(
213                this=seq_get(args, 0), expression=seq_get(args, 1), group=seq_get(args, 2)
214            ),
215            "SEQUENCE": exp.GenerateSeries.from_arg_list,
216            "STRPOS": lambda args: exp.StrPosition(
217                this=seq_get(args, 0), substr=seq_get(args, 1), instance=seq_get(args, 2)
218            ),
219            "TO_UNIXTIME": exp.TimeToUnix.from_arg_list,
220            "TO_HEX": exp.Hex.from_arg_list,
221            "TO_UTF8": lambda args: exp.Encode(
222                this=seq_get(args, 0), charset=exp.Literal.string("utf-8")
223            ),
224        }
225        FUNCTION_PARSERS = parser.Parser.FUNCTION_PARSERS.copy()
226        FUNCTION_PARSERS.pop("TRIM")
227
228    class Generator(generator.Generator):
229        INTERVAL_ALLOWS_PLURAL_FORM = False
230        JOIN_HINTS = False
231        TABLE_HINTS = False
232        QUERY_HINTS = False
233        IS_BOOL_ALLOWED = False
234        STRUCT_DELIMITER = ("(", ")")
235
236        PROPERTIES_LOCATION = {
237            **generator.Generator.PROPERTIES_LOCATION,
238            exp.LocationProperty: exp.Properties.Location.UNSUPPORTED,
239            exp.VolatileProperty: exp.Properties.Location.UNSUPPORTED,
240        }
241
242        TYPE_MAPPING = {
243            **generator.Generator.TYPE_MAPPING,
244            exp.DataType.Type.INT: "INTEGER",
245            exp.DataType.Type.FLOAT: "REAL",
246            exp.DataType.Type.BINARY: "VARBINARY",
247            exp.DataType.Type.TEXT: "VARCHAR",
248            exp.DataType.Type.TIMESTAMPTZ: "TIMESTAMP",
249            exp.DataType.Type.STRUCT: "ROW",
250        }
251
252        TRANSFORMS = {
253            **generator.Generator.TRANSFORMS,
254            exp.ApproxDistinct: _approx_distinct_sql,
255            exp.ApproxQuantile: rename_func("APPROX_PERCENTILE"),
256            exp.Array: lambda self, e: f"ARRAY[{self.expressions(e, flat=True)}]",
257            exp.ArrayConcat: rename_func("CONCAT"),
258            exp.ArrayContains: rename_func("CONTAINS"),
259            exp.ArraySize: rename_func("CARDINALITY"),
260            exp.BitwiseAnd: lambda self, e: f"BITWISE_AND({self.sql(e, 'this')}, {self.sql(e, 'expression')})",
261            exp.BitwiseLeftShift: lambda self, e: f"BITWISE_ARITHMETIC_SHIFT_LEFT({self.sql(e, 'this')}, {self.sql(e, 'expression')})",
262            exp.BitwiseNot: lambda self, e: f"BITWISE_NOT({self.sql(e, 'this')})",
263            exp.BitwiseOr: lambda self, e: f"BITWISE_OR({self.sql(e, 'this')}, {self.sql(e, 'expression')})",
264            exp.BitwiseRightShift: lambda self, e: f"BITWISE_ARITHMETIC_SHIFT_RIGHT({self.sql(e, 'this')}, {self.sql(e, 'expression')})",
265            exp.BitwiseXor: lambda self, e: f"BITWISE_XOR({self.sql(e, 'this')}, {self.sql(e, 'expression')})",
266            exp.Cast: transforms.preprocess([transforms.epoch_cast_to_ts]),
267            exp.CurrentTimestamp: lambda *_: "CURRENT_TIMESTAMP",
268            exp.DataType: _datatype_sql,
269            exp.DateAdd: lambda self, e: self.func(
270                "DATE_ADD", exp.Literal.string(e.text("unit") or "day"), e.expression, e.this
271            ),
272            exp.DateDiff: lambda self, e: self.func(
273                "DATE_DIFF", exp.Literal.string(e.text("unit") or "day"), e.expression, e.this
274            ),
275            exp.DateStrToDate: lambda self, e: f"CAST(DATE_PARSE({self.sql(e, 'this')}, {Presto.DATE_FORMAT}) AS DATE)",
276            exp.DateToDi: lambda self, e: f"CAST(DATE_FORMAT({self.sql(e, 'this')}, {Presto.DATEINT_FORMAT}) AS INT)",
277            exp.Decode: lambda self, e: encode_decode_sql(self, e, "FROM_UTF8"),
278            exp.DiToDate: lambda self, e: f"CAST(DATE_PARSE(CAST({self.sql(e, 'this')} AS VARCHAR), {Presto.DATEINT_FORMAT}) AS DATE)",
279            exp.Encode: lambda self, e: encode_decode_sql(self, e, "TO_UTF8"),
280            exp.FileFormatProperty: lambda self, e: f"FORMAT='{e.name.upper()}'",
281            exp.Group: transforms.preprocess([transforms.unalias_group]),
282            exp.Hex: rename_func("TO_HEX"),
283            exp.If: if_sql,
284            exp.ILike: no_ilike_sql,
285            exp.Initcap: _initcap_sql,
286            exp.Lateral: _explode_to_unnest_sql,
287            exp.Left: left_to_substring_sql,
288            exp.Levenshtein: rename_func("LEVENSHTEIN_DISTANCE"),
289            exp.LogicalAnd: rename_func("BOOL_AND"),
290            exp.LogicalOr: rename_func("BOOL_OR"),
291            exp.Pivot: no_pivot_sql,
292            exp.Quantile: _quantile_sql,
293            exp.RegexpExtract: regexp_extract_sql,
294            exp.Right: right_to_substring_sql,
295            exp.SafeBracket: lambda self, e: self.func(
296                "ELEMENT_AT", e.this, seq_get(apply_index_offset(e.this, e.expressions, 1), 0)
297            ),
298            exp.SafeDivide: no_safe_divide_sql,
299            exp.Schema: _schema_sql,
300            exp.Select: transforms.preprocess(
301                [
302                    transforms.eliminate_qualify,
303                    transforms.eliminate_distinct_on,
304                    transforms.explode_to_unnest,
305                ]
306            ),
307            exp.SortArray: _no_sort_array,
308            exp.StrPosition: rename_func("STRPOS"),
309            exp.StrToDate: lambda self, e: f"CAST({_str_to_time_sql(self, e)} AS DATE)",
310            exp.StrToTime: _str_to_time_sql,
311            exp.StrToUnix: lambda self, e: f"TO_UNIXTIME(DATE_PARSE({self.sql(e, 'this')}, {self.format_time(e)}))",
312            exp.StructExtract: struct_extract_sql,
313            exp.Table: transforms.preprocess([_unnest_sequence]),
314            exp.TimestampTrunc: timestamptrunc_sql,
315            exp.TimeStrToDate: timestrtotime_sql,
316            exp.TimeStrToTime: timestrtotime_sql,
317            exp.TimeStrToUnix: lambda self, e: f"TO_UNIXTIME(DATE_PARSE({self.sql(e, 'this')}, {Presto.TIME_FORMAT}))",
318            exp.TimeToStr: lambda self, e: f"DATE_FORMAT({self.sql(e, 'this')}, {self.format_time(e)})",
319            exp.TimeToUnix: rename_func("TO_UNIXTIME"),
320            exp.TryCast: transforms.preprocess([transforms.epoch_cast_to_ts]),
321            exp.TsOrDiToDi: lambda self, e: f"CAST(SUBSTR(REPLACE(CAST({self.sql(e, 'this')} AS VARCHAR), '-', ''), 1, 8) AS INT)",
322            exp.TsOrDsAdd: _ts_or_ds_add_sql,
323            exp.TsOrDsToDate: _ts_or_ds_to_date_sql,
324            exp.Unhex: rename_func("FROM_HEX"),
325            exp.UnixToStr: lambda self, e: f"DATE_FORMAT(FROM_UNIXTIME({self.sql(e, 'this')}), {self.format_time(e)})",
326            exp.UnixToTime: rename_func("FROM_UNIXTIME"),
327            exp.UnixToTimeStr: lambda self, e: f"CAST(FROM_UNIXTIME({self.sql(e, 'this')}) AS VARCHAR)",
328            exp.VariancePop: rename_func("VAR_POP"),
329            exp.With: transforms.preprocess([transforms.add_recursive_cte_column_names]),
330            exp.WithinGroup: transforms.preprocess(
331                [transforms.remove_within_group_for_percentiles]
332            ),
333        }
334
335        def interval_sql(self, expression: exp.Interval) -> str:
336            unit = self.sql(expression, "unit")
337            if expression.this and unit.lower().startswith("week"):
338                return f"({expression.this.name} * INTERVAL '7' day)"
339            return super().interval_sql(expression)
340
341        def transaction_sql(self, expression: exp.Transaction) -> str:
342            modes = expression.args.get("modes")
343            modes = f" {', '.join(modes)}" if modes else ""
344            return f"START TRANSACTION{modes}"
345
346        def generateseries_sql(self, expression: exp.GenerateSeries) -> str:
347            start = expression.args["start"]
348            end = expression.args["end"]
349            step = expression.args.get("step")
350
351            if isinstance(start, exp.Cast):
352                target_type = start.to
353            elif isinstance(end, exp.Cast):
354                target_type = end.to
355            else:
356                target_type = None
357
358            if target_type and target_type.is_type("timestamp"):
359                to = target_type.copy()
360
361                if target_type is start.to:
362                    end = exp.cast(end, to)
363                else:
364                    start = exp.cast(start, to)
365
366            return self.func("SEQUENCE", start, end, step)
367
368        def offset_limit_modifiers(
369            self, expression: exp.Expression, fetch: bool, limit: t.Optional[exp.Fetch | exp.Limit]
370        ) -> t.List[str]:
371            return [
372                self.sql(expression, "offset"),
373                self.sql(limit),
374            ]