sqlglot.generators.clickhouse

  1from __future__ import annotations
  2
  3import datetime
  4import typing as t
  5
  6from sqlglot import exp, generator
  7from sqlglot.dialects.dialect import (
  8    arg_max_or_min_no_count,
  9    inline_array_sql,
 10    jarowinkler_similarity,
 11    json_extract_segments,
 12    json_path_key_only_name,
 13    length_or_char_length_sql,
 14    no_pivot_sql,
 15    rename_func,
 16    remove_from_array_using_filter,
 17    sha256_sql,
 18    strposition_sql,
 19    var_map_sql,
 20    weekstart_unit_to_str,
 21    unit_to_var,
 22    trim_sql,
 23    sha2_digest_sql,
 24)
 25from sqlglot.generator import unsupported_args
 26from sqlglot.helper import is_int
 27from collections import defaultdict
 28
 29DATETIME_DELTA = t.Union[exp.DateAdd, exp.DateDiff, exp.DateSub, exp.TimestampSub, exp.TimestampAdd]
 30
 31
 32def _unix_to_time_sql(self: ClickHouseGenerator, expression: exp.UnixToTime) -> str:
 33    scale = expression.args.get("scale")
 34    timestamp = expression.this
 35
 36    if scale in (None, exp.UnixToTime.SECONDS):
 37        return self.func("fromUnixTimestamp", exp.cast(timestamp, exp.DType.BIGINT))
 38    if scale == exp.UnixToTime.MILLIS:
 39        return self.func("fromUnixTimestamp64Milli", exp.cast(timestamp, exp.DType.BIGINT))
 40    if scale == exp.UnixToTime.MICROS:
 41        return self.func("fromUnixTimestamp64Micro", exp.cast(timestamp, exp.DType.BIGINT))
 42    if scale == exp.UnixToTime.NANOS:
 43        return self.func("fromUnixTimestamp64Nano", exp.cast(timestamp, exp.DType.BIGINT))
 44
 45    return self.func(
 46        "fromUnixTimestamp",
 47        exp.cast(exp.Div(this=timestamp, expression=exp.func("POW", 10, scale)), exp.DType.BIGINT),
 48    )
 49
 50
 51def _lower_func(sql: str) -> str:
 52    index = sql.index("(")
 53    return sql[:index].lower() + sql[index:]
 54
 55
 56def _quantile_sql(self: ClickHouseGenerator, expression: exp.Quantile) -> str:
 57    quantile = expression.args["quantile"]
 58    args = f"({self.sql(expression, 'this')})"
 59
 60    if isinstance(quantile, exp.Array):
 61        func = self.func("quantiles", *quantile)
 62    else:
 63        func = self.func("quantile", quantile)
 64
 65    return func + args
 66
 67
 68def _datetime_delta_sql(name: str) -> t.Callable[[generator.Generator, DATETIME_DELTA], str]:
 69    def _delta_sql(self: generator.Generator, expression: DATETIME_DELTA) -> str:
 70        if not expression.unit:
 71            return rename_func(name)(self, expression)
 72
 73        return self.func(
 74            name,
 75            unit_to_var(expression),
 76            expression.expression,
 77            expression.this,
 78            expression.args.get("zone"),
 79        )
 80
 81    return _delta_sql
 82
 83
 84def _timestrtotime_sql(self: ClickHouseGenerator, expression: exp.TimeStrToTime):
 85    ts = expression.this
 86
 87    tz = expression.args.get("zone")
 88    if tz and isinstance(ts, exp.Literal):
 89        # Clickhouse will not accept timestamps that include a UTC offset, so we must remove them.
 90        # The first step to removing is parsing the string with `datetime.datetime.fromisoformat`.
 91        #
 92        # In python <3.11, `fromisoformat()` can only parse timestamps of millisecond (3 digit)
 93        # or microsecond (6 digit) precision. It will error if passed any other number of fractional
 94        # digits, so we extract the fractional seconds and pad to 6 digits before parsing.
 95        ts_string = ts.name.strip()
 96
 97        # separate [date and time] from [fractional seconds and UTC offset]
 98        ts_parts = ts_string.split(".")
 99        if len(ts_parts) == 2:
100            # separate fractional seconds and UTC offset
101            offset_sep = "+" if "+" in ts_parts[1] else "-"
102            ts_frac_parts = ts_parts[1].split(offset_sep)
103            num_frac_parts = len(ts_frac_parts)
104
105            # pad to 6 digits if fractional seconds present
106            ts_frac_parts[0] = ts_frac_parts[0].ljust(6, "0")
107            ts_string = "".join(
108                [
109                    ts_parts[0],  # date and time
110                    ".",
111                    ts_frac_parts[0],  # fractional seconds
112                    offset_sep if num_frac_parts > 1 else "",
113                    ts_frac_parts[1] if num_frac_parts > 1 else "",  # utc offset (if present)
114                ]
115            )
116
117        # return literal with no timezone, eg turn '2020-01-01 12:13:14-08:00' into '2020-01-01 12:13:14'
118        # this is because Clickhouse encodes the timezone as a data type parameter and throws an error if
119        # it's part of the timestamp string
120        ts_without_tz = (
121            datetime.datetime.fromisoformat(ts_string).replace(tzinfo=None).isoformat(sep=" ")
122        )
123        ts = exp.Literal.string(ts_without_tz)
124
125    # Non-nullable DateTime64 with microsecond precision
126    expressions = [exp.DataTypeParam(this=tz)] if tz else []
127    datatype = exp.DType.DATETIME64.into_expr(
128        expressions=[exp.DataTypeParam(this=exp.Literal.number(6)), *expressions],
129        nullable=False,
130    )
131
132    return self.sql(exp.cast(ts, datatype, dialect=self.dialect))
133
134
135def _map_sql(self: ClickHouseGenerator, expression: exp.Map | exp.VarMap) -> str:
136    if not (expression.parent and expression.parent.arg_key == "settings"):
137        return _lower_func(var_map_sql(self, expression))
138
139    keys = expression.args.get("keys")
140    values = expression.args.get("values")
141
142    if not isinstance(keys, exp.Array) or not isinstance(values, exp.Array):
143        self.unsupported("Cannot convert array columns into map.")
144        return ""
145
146    args = []
147    for key, value in zip(keys.expressions, values.expressions):
148        args.append(f"{self.sql(key)}: {self.sql(value)}")
149
150    csv_args = ", ".join(args)
151
152    return f"{{{csv_args}}}"
153
154
155def _json_cast_sql(self: ClickHouseGenerator, expression: exp.JSONCast) -> str:
156    this = self.sql(expression, "this")
157    to = expression.to
158    to_sql = self.sql(to)
159
160    if to.expressions:
161        to_sql = self.sql(exp.to_identifier(to_sql))
162
163    return f"{this}.:{to_sql}"
164
165
166class ClickHouseGenerator(generator.Generator):
167    SELECT_KINDS: tuple[str, ...] = ()
168    TRY_SUPPORTED = False
169    SUPPORTS_UESCAPE = False
170    SUPPORTS_DECODE_CASE = False
171
172    AFTER_HAVING_MODIFIER_TRANSFORMS = generator.AFTER_HAVING_MODIFIER_TRANSFORMS
173
174    QUERY_HINTS = False
175    STRUCT_DELIMITER = ("(", ")")
176    NVL2_SUPPORTED = False
177    ALTER_SET_TYPE = "TYPE"
178    SUPPORTS_ALTER_COLUMN_IF_EXISTS = True
179    TABLESAMPLE_REQUIRES_PARENS = False
180    TABLESAMPLE_SIZE_IS_ROWS = False
181    TABLESAMPLE_KEYWORDS = "SAMPLE"
182    LAST_DAY_SUPPORTS_DATE_PART = False
183    CAN_IMPLEMENT_ARRAY_ANY = True
184    SUPPORTS_TO_NUMBER = False
185    JOIN_HINTS = False
186    TABLE_HINTS = False
187    GROUPINGS_SEP = ""
188    SET_OP_MODIFIERS = False
189    SET_OP_LIMITS = True
190    ARRAY_SIZE_NAME = "LENGTH"
191    WRAP_DERIVED_VALUES = False
192    AUTO_REFRESH_BARE_INTERVALS = True
193
194    STRING_TYPE_MAPPING: t.ClassVar = {
195        exp.DType.BLOB: "String",
196        exp.DType.CHAR: "String",
197        exp.DType.LONGBLOB: "String",
198        exp.DType.LONGTEXT: "String",
199        exp.DType.MEDIUMBLOB: "String",
200        exp.DType.MEDIUMTEXT: "String",
201        exp.DType.TINYBLOB: "String",
202        exp.DType.TINYTEXT: "String",
203        exp.DType.TEXT: "String",
204        exp.DType.VARBINARY: "String",
205        exp.DType.VARCHAR: "String",
206    }
207
208    SUPPORTED_JSON_PATH_PARTS = {
209        exp.JSONPathKey,
210        exp.JSONPathRoot,
211        exp.JSONPathSubscript,
212    }
213
214    TYPE_MAPPING = {
215        **generator.Generator.TYPE_MAPPING,
216        exp.DType.BLOB: "String",
217        exp.DType.CHAR: "String",
218        exp.DType.LONGBLOB: "String",
219        exp.DType.LONGTEXT: "String",
220        exp.DType.MEDIUMBLOB: "String",
221        exp.DType.MEDIUMTEXT: "String",
222        exp.DType.TINYBLOB: "String",
223        exp.DType.TINYTEXT: "String",
224        exp.DType.TEXT: "String",
225        exp.DType.VARBINARY: "String",
226        exp.DType.VARCHAR: "String",
227        exp.DType.ARRAY: "Array",
228        exp.DType.BOOLEAN: "Bool",
229        exp.DType.BIGINT: "Int64",
230        exp.DType.DATE32: "Date32",
231        exp.DType.DATETIME: "DateTime",
232        exp.DType.DATETIME2: "DateTime",
233        exp.DType.SMALLDATETIME: "DateTime",
234        exp.DType.DATETIME64: "DateTime64",
235        exp.DType.DECIMAL: "Decimal",
236        exp.DType.DECIMAL32: "Decimal32",
237        exp.DType.DECIMAL64: "Decimal64",
238        exp.DType.DECIMAL128: "Decimal128",
239        exp.DType.DECIMAL256: "Decimal256",
240        exp.DType.TIMESTAMP: "DateTime",
241        exp.DType.TIMESTAMPNTZ: "DateTime",
242        exp.DType.TIMESTAMPTZ: "DateTime",
243        exp.DType.DOUBLE: "Float64",
244        exp.DType.ENUM: "Enum",
245        exp.DType.ENUM8: "Enum8",
246        exp.DType.ENUM16: "Enum16",
247        exp.DType.FIXEDSTRING: "FixedString",
248        exp.DType.FLOAT: "Float32",
249        exp.DType.INT: "Int32",
250        exp.DType.MEDIUMINT: "Int32",
251        exp.DType.INT128: "Int128",
252        exp.DType.INT256: "Int256",
253        exp.DType.LOWCARDINALITY: "LowCardinality",
254        exp.DType.MAP: "Map",
255        exp.DType.NESTED: "Nested",
256        exp.DType.NOTHING: "Nothing",
257        exp.DType.SMALLINT: "Int16",
258        exp.DType.STRUCT: "Tuple",
259        exp.DType.TINYINT: "Int8",
260        exp.DType.UBIGINT: "UInt64",
261        exp.DType.UINT: "UInt32",
262        exp.DType.UINT128: "UInt128",
263        exp.DType.UINT256: "UInt256",
264        exp.DType.USMALLINT: "UInt16",
265        exp.DType.UTINYINT: "UInt8",
266        exp.DType.IPV4: "IPv4",
267        exp.DType.IPV6: "IPv6",
268        exp.DType.POINT: "Point",
269        exp.DType.RING: "Ring",
270        exp.DType.LINESTRING: "LineString",
271        exp.DType.MULTILINESTRING: "MultiLineString",
272        exp.DType.POLYGON: "Polygon",
273        exp.DType.MULTIPOLYGON: "MultiPolygon",
274        exp.DType.AGGREGATEFUNCTION: "AggregateFunction",
275        exp.DType.SIMPLEAGGREGATEFUNCTION: "SimpleAggregateFunction",
276        exp.DType.DYNAMIC: "Dynamic",
277    }
278
279    TRANSFORMS = {
280        **{k: v for k, v in generator.Generator.TRANSFORMS.items() if k != exp.AutoRefreshProperty},
281        exp.AnyValue: rename_func("any"),
282        exp.ApproxDistinct: rename_func("uniq"),
283        exp.ArrayDistinct: rename_func("arrayDistinct"),
284        exp.ArrayConcat: rename_func("arrayConcat"),
285        exp.ArrayContains: rename_func("has"),
286        exp.ArrayFilter: lambda self, e: self.func("arrayFilter", e.expression, e.this),
287        exp.Transform: lambda self, e: self.func("arrayMap", e.expression, e.this),
288        exp.ArrayRemove: remove_from_array_using_filter,
289        exp.ArrayReverse: rename_func("arrayReverse"),
290        exp.ArraySlice: rename_func("arraySlice"),
291        exp.ArraySum: rename_func("arraySum"),
292        exp.ArrayMax: rename_func("arrayMax"),
293        exp.ArrayMin: rename_func("arrayMin"),
294        exp.ArgMax: arg_max_or_min_no_count("argMax"),
295        exp.ArgMin: arg_max_or_min_no_count("argMin"),
296        exp.Array: inline_array_sql,
297        exp.CityHash64: rename_func("cityHash64"),
298        exp.CastToStrType: rename_func("CAST"),
299        exp.CurrentDatabase: rename_func("CURRENT_DATABASE"),
300        exp.CurrentSchemas: rename_func("CURRENT_SCHEMAS"),
301        exp.CountIf: rename_func("countIf"),
302        exp.CosineDistance: rename_func("cosineDistance"),
303        exp.CompressColumnConstraint: lambda self, e: (
304            f"CODEC({self.expressions(e, key='this', flat=True)})"
305        ),
306        exp.ComputedColumnConstraint: lambda self, e: (
307            f"{'MATERIALIZED' if e.args.get('persisted') else 'ALIAS'} {self.sql(e, 'this')}"
308        ),
309        exp.CurrentDate: lambda self, e: self.func("CURRENT_DATE"),
310        exp.CurrentVersion: rename_func("VERSION"),
311        exp.DateAdd: _datetime_delta_sql("DATE_ADD"),
312        exp.DateDiff: _datetime_delta_sql("DATE_DIFF"),
313        exp.DateStrToDate: rename_func("toDate"),
314        exp.DateSub: _datetime_delta_sql("DATE_SUB"),
315        exp.Explode: rename_func("arrayJoin"),
316        exp.FarmFingerprint: rename_func("farmFingerprint64"),
317        exp.Final: lambda self, e: f"{self.sql(e, 'this')} FINAL",
318        exp.IsNan: rename_func("isNaN"),
319        exp.JarowinklerSimilarity: jarowinkler_similarity("jaroWinklerSimilarity"),
320        exp.JSONCast: _json_cast_sql,
321        exp.JSONExtract: json_extract_segments("JSONExtractString", quoted_index=False),
322        exp.JSONExtractScalar: json_extract_segments("JSONExtractString", quoted_index=False),
323        exp.JSONPathKey: json_path_key_only_name,
324        exp.JSONPathRoot: lambda *_: "",
325        exp.Length: length_or_char_length_sql,
326        exp.Map: _map_sql,
327        exp.Median: rename_func("median"),
328        exp.Nullif: rename_func("nullIf"),
329        exp.PartitionedByProperty: lambda self, e: f"PARTITION BY {self.sql(e, 'this')}",
330        exp.Pivot: no_pivot_sql,
331        exp.Quantile: _quantile_sql,
332        exp.RegexpLike: lambda self, e: self.func("match", e.this, e.expression),
333        exp.Rand: rename_func("randCanonical"),
334        exp.StartsWith: rename_func("startsWith"),
335        exp.Struct: rename_func("tuple"),
336        exp.Trunc: rename_func("trunc"),
337        exp.EndsWith: rename_func("endsWith"),
338        exp.EuclideanDistance: rename_func("L2Distance"),
339        exp.StrPosition: lambda self, e: strposition_sql(
340            self,
341            e,
342            func_name="POSITION",
343            supports_position=True,
344            use_ansi_position=False,
345        ),
346        exp.TimeToStr: lambda self, e: self.func(
347            "formatDateTime",
348            e.this.this if isinstance(e.this, exp.TsOrDsToTimestamp) else e.this,
349            self.format_time(e),
350            e.args.get("zone"),
351        ),
352        exp.TimeStrToTime: _timestrtotime_sql,
353        exp.TimestampAdd: _datetime_delta_sql("TIMESTAMP_ADD"),
354        exp.TimestampSub: _datetime_delta_sql("TIMESTAMP_SUB"),
355        exp.Typeof: rename_func("toTypeName"),
356        exp.VarMap: _map_sql,
357        exp.Xor: lambda self, e: self.func("xor", e.this, e.expression),
358        exp.MD5Digest: rename_func("MD5"),
359        exp.MD5: lambda self, e: self.func("LOWER", self.func("HEX", self.func("MD5", e.this))),
360        exp.SHA: rename_func("SHA1"),
361        exp.SHA1Digest: rename_func("SHA1"),
362        exp.SHA2: sha256_sql,
363        exp.SHA2Digest: sha2_digest_sql,
364        exp.Split: lambda self, e: self.func(
365            "splitByString", e.args.get("expression"), e.this, e.args.get("limit")
366        ),
367        exp.RegexpSplit: lambda self, e: self.func(
368            "splitByRegexp", e.args.get("expression"), e.this, e.args.get("limit")
369        ),
370        exp.UnixToTime: _unix_to_time_sql,
371        exp.Trim: lambda self, e: trim_sql(self, e, default_trim_type="BOTH"),
372        exp.Variance: rename_func("varSamp"),
373        exp.VariancePop: rename_func("varPop"),
374        exp.SchemaCommentProperty: lambda self, e: self.naked_property(e),
375        exp.Stddev: rename_func("stddevSamp"),
376        exp.Chr: rename_func("CHAR"),
377        exp.Lag: rename_func("lag"),
378        exp.Lead: rename_func("lead"),
379        exp.Levenshtein: unsupported_args("ins_cost", "del_cost", "sub_cost", "max_dist")(
380            rename_func("editDistance")
381        ),
382        exp.ParseDatetime: lambda self, e: self.func(
383            "parseDateTime", e.this, e.args.get("format"), e.args.get("zone")
384        ),
385    }
386
387    PROPERTIES_LOCATION = {
388        **generator.Generator.PROPERTIES_LOCATION,
389        exp.AutoRefreshProperty: exp.Properties.Location.POST_NAME,
390        exp.DefinerProperty: exp.Properties.Location.POST_SCHEMA,
391        exp.OnCluster: exp.Properties.Location.POST_NAME,
392        exp.PartitionedByProperty: exp.Properties.Location.POST_SCHEMA,
393        exp.ToTableProperty: exp.Properties.Location.POST_NAME,
394        exp.UuidProperty: exp.Properties.Location.POST_NAME,
395        exp.VolatileProperty: exp.Properties.Location.UNSUPPORTED,
396    }
397
398    # There's no list in docs, but it can be found in Clickhouse code
399    # see `ClickHouse/src/Parsers/ParserCreate*.cpp`
400    ON_CLUSTER_TARGETS: t.ClassVar = {
401        "SCHEMA",  # Transpiled CREATE SCHEMA may have OnCluster property set
402        "DATABASE",
403        "TABLE",
404        "VIEW",
405        "DICTIONARY",
406        "INDEX",
407        "FUNCTION",
408        "NAMED COLLECTION",
409    }
410
411    # https://clickhouse.com/docs/en/sql-reference/data-types/nullable
412    NON_NULLABLE_TYPES: t.ClassVar = {
413        exp.DType.ARRAY,
414        exp.DType.MAP,
415        exp.DType.STRUCT,
416        exp.DType.POINT,
417        exp.DType.RING,
418        exp.DType.LINESTRING,
419        exp.DType.MULTILINESTRING,
420        exp.DType.POLYGON,
421        exp.DType.MULTIPOLYGON,
422    }
423
424    def groupconcat_sql(self, expression: exp.GroupConcat) -> str:
425        this = expression.this
426        separator = expression.args.get("separator")
427
428        if isinstance(this, exp.Limit) and this.this:
429            limit = this
430            this = limit.this.pop()
431            return self.sql(
432                exp.ParameterizedAgg(
433                    this="groupConcat",
434                    params=[this],
435                    expressions=[separator, limit.expression],
436                )
437            )
438
439        if separator:
440            return self.sql(
441                exp.ParameterizedAgg(
442                    this="groupConcat",
443                    params=[this],
444                    expressions=[separator],
445                )
446            )
447
448        return self.func("groupConcat", this)
449
450    def select_sql(self, expression: exp.Select) -> str:
451        limit = expression.args.get("limit")
452        if isinstance(limit, exp.Fetch) and not expression.args.get("order"):
453            count = limit.args.get("count")
454            expression.set(
455                "limit", exp.Limit(expression=count if count is not None else exp.Literal.number(1))
456            )
457        return super().select_sql(expression)
458
459    def offset_sql(self, expression: exp.Offset) -> str:
460        offset = super().offset_sql(expression)
461
462        # OFFSET ... FETCH syntax requires a "ROW" or "ROWS" keyword
463        # https://clickhouse.com/docs/sql-reference/statements/select/offset
464        parent = expression.parent
465        if isinstance(parent, exp.Select) and isinstance(parent.args.get("limit"), exp.Fetch):
466            offset = f"{offset} ROWS"
467
468        return offset
469
470    def strtodate_sql(self, expression: exp.StrToDate) -> str:
471        strtodate_sql = self.function_fallback_sql(expression)
472
473        if not isinstance(expression.parent, exp.Cast):
474            # StrToDate returns DATEs in other dialects (eg. postgres), so
475            # this branch aims to improve the transpilation to clickhouse
476            return self.cast_sql(exp.cast(expression, "DATE"))
477
478        return strtodate_sql
479
480    def cast_sql(self, expression: exp.Cast, safe_prefix: str | None = None) -> str:
481        this = expression.this
482
483        if isinstance(this, exp.StrToDate) and expression.to == exp.DType.DATETIME.into_expr():
484            return self.sql(this)
485
486        return super().cast_sql(expression, safe_prefix=safe_prefix)
487
488    def trycast_sql(self, expression: exp.TryCast) -> str:
489        dtype = expression.to
490        if not dtype.is_type(*self.NON_NULLABLE_TYPES, check_nullable=True):
491            # Casting x into Nullable(T) appears to behave similarly to TRY_CAST(x AS T)
492            dtype.set("nullable", True)
493
494        return super().cast_sql(expression)
495
496    def _jsonpathsubscript_sql(self, expression: exp.JSONPathSubscript) -> str:
497        this = self.json_path_part(expression.this)
498        return str(int(this) + 1) if is_int(this) else this
499
500    def likeproperty_sql(self, expression: exp.LikeProperty) -> str:
501        return f"AS {self.sql(expression, 'this')}"
502
503    def _any_to_has(
504        self,
505        expression: exp.EQ | exp.NEQ,
506        default: t.Callable[[t.Any], str],
507        prefix: str = "",
508    ) -> str:
509        if isinstance(expression.left, exp.Any):
510            arr = expression.left
511            this = expression.right
512        elif isinstance(expression.right, exp.Any):
513            arr = expression.right
514            this = expression.left
515        else:
516            return default(expression)
517
518        return prefix + self.func("has", arr.this.unnest(), this)
519
520    def eq_sql(self, expression: exp.EQ) -> str:
521        return self._any_to_has(expression, super().eq_sql)
522
523    def neq_sql(self, expression: exp.NEQ) -> str:
524        return self._any_to_has(expression, super().neq_sql, "NOT ")
525
526    def regexpilike_sql(self, expression: exp.RegexpILike) -> str:
527        # Manually add a flag to make the search case-insensitive
528        regex = self.func("CONCAT", "'(?i)'", expression.expression)
529        return self.func("match", expression.this, regex)
530
531    def datatype_sql(self, expression: exp.DataType) -> str:
532        # String is the standard ClickHouse type, every other variant is just an alias.
533        # Additionally, any supplied length parameter will be ignored.
534        #
535        # https://clickhouse.com/docs/en/sql-reference/data-types/string
536        if expression.this in self.STRING_TYPE_MAPPING:
537            dtype = "String"
538        else:
539            dtype = super().datatype_sql(expression)
540
541        # This section changes the type to `Nullable(...)` if the following conditions hold:
542        # - It's marked as nullable - this ensures we won't wrap ClickHouse types with `Nullable`
543        #   and change their semantics
544        # - It's not the key type of a `Map`. This is because ClickHouse enforces the following
545        #   constraint: "Type of Map key must be a type, that can be represented by integer or
546        #   String or FixedString (possibly LowCardinality) or UUID or IPv6"
547        # - It's not a composite type, e.g. `Nullable(Array(...))` is not a valid type
548        parent = expression.parent
549        nullable = expression.args.get("nullable")
550        if nullable is True or (
551            nullable is None
552            and not (
553                isinstance(parent, exp.DataType)
554                and parent.is_type(exp.DType.MAP, check_nullable=True)
555                and expression.index in (None, 0)
556            )
557            and not expression.is_type(*self.NON_NULLABLE_TYPES, check_nullable=True)
558        ):
559            dtype = f"Nullable({dtype})"
560
561        return dtype
562
563    def cte_sql(self, expression: exp.CTE) -> str:
564        if expression.args.get("scalar"):
565            this = self.sql(expression, "this")
566            alias = self.sql(expression, "alias")
567            return f"{this} AS {alias}"
568
569        return super().cte_sql(expression)
570
571    def after_limit_modifiers(self, expression: exp.Expr) -> list[str]:
572        return super().after_limit_modifiers(expression) + [
573            (
574                self.seg("SETTINGS ") + self.expressions(expression, key="settings", flat=True)
575                if expression.args.get("settings")
576                else ""
577            ),
578            (
579                self.seg("FORMAT ") + self.sql(expression, "format")
580                if expression.args.get("format")
581                else ""
582            ),
583        ]
584
585    def placeholder_sql(self, expression: exp.Placeholder) -> str:
586        return f"{{{expression.name}: {self.sql(expression, 'kind')}}}"
587
588    def oncluster_sql(self, expression: exp.OnCluster) -> str:
589        return f"ON CLUSTER {self.sql(expression, 'this')}"
590
591    def _refresh_interval_sql(self, expression: exp.Expr) -> str:
592        if isinstance(expression, exp.Add):
593            return f"{self._refresh_interval_sql(expression.this)} {self._refresh_interval_sql(expression.expression)}"
594        return self.sql(expression.assert_is(exp.Interval))
595
596    def autorefreshproperty_sql(self, expression: exp.AutoRefreshProperty) -> str:
597        cadence = self.sql(expression, "cadence")
598        interval = expression.this
599        schedule = (
600            f" {cadence} {self._refresh_interval_sql(interval)}" if cadence and interval else ""
601        )
602        offset = expression.args.get("offset")
603        offset = f" OFFSET {self._refresh_interval_sql(offset)}" if offset else ""
604        randomize = expression.args.get("randomize")
605        randomize = f" RANDOMIZE FOR {self._refresh_interval_sql(randomize)}" if randomize else ""
606        dependencies = self.expressions(expression, flat=True)
607        dependencies = f" DEPENDS ON {dependencies}" if dependencies else ""
608        settings = self.sql(expression, "settings")
609        settings = f" {settings}" if settings else ""
610        append = " APPEND" if expression.args.get("append") else ""
611
612        return f"REFRESH{schedule}{offset}{randomize}{dependencies}{settings}{append}"
613
614    def createable_sql(self, expression: exp.Create, locations: defaultdict) -> str:
615        if expression.kind in self.ON_CLUSTER_TARGETS and locations.get(
616            exp.Properties.Location.POST_NAME
617        ):
618            this_name = self.sql(
619                expression.this if isinstance(expression.this, exp.Schema) else expression,
620                "this",
621            )
622            this_properties = " ".join(
623                [self.sql(prop) for prop in locations[exp.Properties.Location.POST_NAME]]
624            )
625            this_schema = self.schema_columns_sql(expression.this)
626            this_schema = f"{self.sep()}{this_schema}" if this_schema else ""
627
628            return f"{this_name}{self.sep()}{this_properties}{this_schema}"
629
630        return super().createable_sql(expression, locations)
631
632    def create_sql(self, expression: exp.Create) -> str:
633        # The comment property comes last in CTAS statements, i.e. after the query
634        query = expression.expression
635        if isinstance(query, exp.Query):
636            comment_prop = expression.find(exp.SchemaCommentProperty)
637            if comment_prop:
638                comment_prop.pop()
639                query.replace(exp.paren(query))
640        else:
641            comment_prop = None
642
643        create_sql = super().create_sql(expression)
644
645        comment_sql = self.sql(comment_prop)
646        comment_sql = f" {comment_sql}" if comment_sql else ""
647
648        return f"{create_sql}{comment_sql}"
649
650    def prewhere_sql(self, expression: exp.PreWhere) -> str:
651        this = self.indent(self.sql(expression, "this"))
652        return f"{self.seg('PREWHERE')}{self.sep()}{this}"
653
654    def indexcolumnconstraint_sql(self, expression: exp.IndexColumnConstraint) -> str:
655        this = self.sql(expression, "this")
656        this = f" {this}" if this else ""
657        expr = self.sql(expression, "expression")
658        expr = f" {expr}" if expr else ""
659        index_type = self.sql(expression, "index_type")
660        index_type = f" TYPE {index_type}" if index_type else ""
661        granularity = self.sql(expression, "granularity")
662        granularity = f" GRANULARITY {granularity}" if granularity else ""
663
664        return f"INDEX{this}{expr}{index_type}{granularity}"
665
666    def partition_sql(self, expression: exp.Partition) -> str:
667        return f"PARTITION {self.expressions(expression, flat=True)}"
668
669    def partitionid_sql(self, expression: exp.PartitionId) -> str:
670        return f"ID {self.sql(expression.this)}"
671
672    def replacepartition_sql(self, expression: exp.ReplacePartition) -> str:
673        return f"REPLACE {self.sql(expression.expression)} FROM {self.sql(expression, 'source')}"
674
675    def projectiondef_sql(self, expression: exp.ProjectionDef) -> str:
676        return f"PROJECTION {self.sql(expression.this)} {self.wrap(expression.expression)}"
677
678    def nestedjsonselect_sql(self, expression: exp.NestedJSONSelect) -> str:
679        return f"{self.sql(expression, 'this')}.^{self.sql(expression, 'expression')}"
680
681    def is_sql(self, expression: exp.Is) -> str:
682        is_sql = super().is_sql(expression)
683
684        if isinstance(expression.parent, exp.Not):
685            # value IS NOT NULL -> NOT (value IS NULL)
686            is_sql = self.wrap(is_sql)
687
688        return is_sql
689
690    def in_sql(self, expression: exp.In) -> str:
691        in_sql = super().in_sql(expression)
692
693        if isinstance(expression.parent, exp.Not) and expression.args.get("is_global"):
694            in_sql = in_sql.replace("GLOBAL IN", "GLOBAL NOT IN", 1)
695
696        return in_sql
697
698    def not_sql(self, expression: exp.Not) -> str:
699        if isinstance(expression.this, exp.In):
700            if expression.this.args.get("is_global"):
701                # let `GLOBAL IN` child interpose `NOT`
702                return self.sql(expression, "this")
703
704            expression.set("this", exp.paren(expression.this, copy=False))
705
706        return super().not_sql(expression)
707
708    def values_sql(self, expression: exp.Values, values_as_table: bool = True) -> str:
709        # If the VALUES clause contains tuples of expressions, we need to treat it
710        # as a table since Clickhouse will automatically alias it as such.
711        alias = expression.args.get("alias")
712
713        if alias and alias.args.get("columns") and expression.expressions:
714            values = expression.expressions[0].expressions
715            values_as_table = any(isinstance(value, exp.Tuple) for value in values)
716        else:
717            values_as_table = True
718
719        return super().values_sql(expression, values_as_table=values_as_table)
720
721    def timestamptrunc_sql(self, expression: exp.DateTrunc | exp.TimestampTrunc) -> str:
722        unit = weekstart_unit_to_str(self, expression)
723        # https://clickhouse.com/docs/whats-new/changelog/2023#improvement
724        if self.dialect.version < (23, 12) and unit and unit.is_string:
725            unit = exp.Literal.string(unit.name.lower())
726        return self.func("dateTrunc", unit, expression.this, expression.args.get("zone"))
727
728    def datetrunc_sql(self, expression: exp.DateTrunc) -> str:
729        return self.timestamptrunc_sql(expression)