Edit on GitHub

sqlglot.dialects.spark2

  1from __future__ import annotations
  2
  3import typing as t
  4
  5from sqlglot import exp, parser, transforms
  6from sqlglot.dialects.dialect import (
  7    create_with_partitions_sql,
  8    pivot_column_names,
  9    rename_func,
 10    trim_sql,
 11)
 12from sqlglot.dialects.hive import Hive
 13from sqlglot.helper import seq_get
 14
 15
 16def _create_sql(self: Hive.Generator, e: exp.Create) -> str:
 17    kind = e.args["kind"]
 18    properties = e.args.get("properties")
 19
 20    if kind.upper() == "TABLE" and any(
 21        isinstance(prop, exp.TemporaryProperty)
 22        for prop in (properties.expressions if properties else [])
 23    ):
 24        return f"CREATE TEMPORARY VIEW {self.sql(e, 'this')} AS {self.sql(e, 'expression')}"
 25    return create_with_partitions_sql(self, e)
 26
 27
 28def _map_sql(self: Hive.Generator, expression: exp.Map) -> str:
 29    keys = self.sql(expression.args["keys"])
 30    values = self.sql(expression.args["values"])
 31    return f"MAP_FROM_ARRAYS({keys}, {values})"
 32
 33
 34def _parse_as_cast(to_type: str) -> t.Callable[[t.List], exp.Expression]:
 35    return lambda args: exp.Cast(this=seq_get(args, 0), to=exp.DataType.build(to_type))
 36
 37
 38def _str_to_date(self: Hive.Generator, expression: exp.StrToDate) -> str:
 39    this = self.sql(expression, "this")
 40    time_format = self.format_time(expression)
 41    if time_format == Hive.DATE_FORMAT:
 42        return f"TO_DATE({this})"
 43    return f"TO_DATE({this}, {time_format})"
 44
 45
 46def _unix_to_time_sql(self: Hive.Generator, expression: exp.UnixToTime) -> str:
 47    scale = expression.args.get("scale")
 48    timestamp = self.sql(expression, "this")
 49    if scale is None:
 50        return f"CAST(FROM_UNIXTIME({timestamp}) AS TIMESTAMP)"
 51    if scale == exp.UnixToTime.SECONDS:
 52        return f"TIMESTAMP_SECONDS({timestamp})"
 53    if scale == exp.UnixToTime.MILLIS:
 54        return f"TIMESTAMP_MILLIS({timestamp})"
 55    if scale == exp.UnixToTime.MICROS:
 56        return f"TIMESTAMP_MICROS({timestamp})"
 57
 58    raise ValueError("Improper scale for timestamp")
 59
 60
 61def _unalias_pivot(expression: exp.Expression) -> exp.Expression:
 62    """
 63    Spark doesn't allow PIVOT aliases, so we need to remove them and possibly wrap a
 64    pivoted source in a subquery with the same alias to preserve the query's semantics.
 65
 66    Example:
 67        >>> from sqlglot import parse_one
 68        >>> expr = parse_one("SELECT piv.x FROM tbl PIVOT (SUM(a) FOR b IN ('x')) piv")
 69        >>> print(_unalias_pivot(expr).sql(dialect="spark"))
 70        SELECT piv.x FROM (SELECT * FROM tbl PIVOT(SUM(a) FOR b IN ('x'))) AS piv
 71    """
 72    if isinstance(expression, exp.From) and expression.this.args.get("pivots"):
 73        pivot = expression.this.args["pivots"][0]
 74        if pivot.alias:
 75            alias = pivot.args["alias"].pop()
 76            return exp.From(
 77                this=expression.this.replace(
 78                    exp.select("*").from_(expression.this.copy()).subquery(alias=alias)
 79                )
 80            )
 81
 82    return expression
 83
 84
 85def _unqualify_pivot_columns(expression: exp.Expression) -> exp.Expression:
 86    """
 87    Spark doesn't allow the column referenced in the PIVOT's field to be qualified,
 88    so we need to unqualify it.
 89
 90    Example:
 91        >>> from sqlglot import parse_one
 92        >>> expr = parse_one("SELECT * FROM tbl PIVOT (SUM(tbl.sales) FOR tbl.quarter IN ('Q1', 'Q2'))")
 93        >>> print(_unqualify_pivot_columns(expr).sql(dialect="spark"))
 94        SELECT * FROM tbl PIVOT(SUM(tbl.sales) FOR quarter IN ('Q1', 'Q1'))
 95    """
 96    if isinstance(expression, exp.Pivot):
 97        expression.args["field"].transform(
 98            lambda node: exp.column(node.output_name, quoted=node.this.quoted)
 99            if isinstance(node, exp.Column)
100            else node,
101            copy=False,
102        )
103
104    return expression
105
106
107class Spark2(Hive):
108    class Parser(Hive.Parser):
109        FUNCTIONS = {
110            **Hive.Parser.FUNCTIONS,
111            "MAP_FROM_ARRAYS": exp.Map.from_arg_list,
112            "TO_UNIX_TIMESTAMP": exp.StrToUnix.from_arg_list,
113            "SHIFTLEFT": lambda args: exp.BitwiseLeftShift(
114                this=seq_get(args, 0),
115                expression=seq_get(args, 1),
116            ),
117            "SHIFTRIGHT": lambda args: exp.BitwiseRightShift(
118                this=seq_get(args, 0),
119                expression=seq_get(args, 1),
120            ),
121            "APPROX_PERCENTILE": exp.ApproxQuantile.from_arg_list,
122            "IIF": exp.If.from_arg_list,
123            "AGGREGATE": exp.Reduce.from_arg_list,
124            "DAYOFWEEK": lambda args: exp.DayOfWeek(
125                this=exp.TsOrDsToDate(this=seq_get(args, 0)),
126            ),
127            "DAYOFMONTH": lambda args: exp.DayOfMonth(
128                this=exp.TsOrDsToDate(this=seq_get(args, 0)),
129            ),
130            "DAYOFYEAR": lambda args: exp.DayOfYear(
131                this=exp.TsOrDsToDate(this=seq_get(args, 0)),
132            ),
133            "WEEKOFYEAR": lambda args: exp.WeekOfYear(
134                this=exp.TsOrDsToDate(this=seq_get(args, 0)),
135            ),
136            "DATE_TRUNC": lambda args: exp.TimestampTrunc(
137                this=seq_get(args, 1),
138                unit=exp.var(seq_get(args, 0)),
139            ),
140            "TRUNC": lambda args: exp.DateTrunc(unit=seq_get(args, 1), this=seq_get(args, 0)),
141            "BOOLEAN": _parse_as_cast("boolean"),
142            "DATE": _parse_as_cast("date"),
143            "DOUBLE": _parse_as_cast("double"),
144            "FLOAT": _parse_as_cast("float"),
145            "INT": _parse_as_cast("int"),
146            "STRING": _parse_as_cast("string"),
147            "TIMESTAMP": _parse_as_cast("timestamp"),
148        }
149
150        FUNCTION_PARSERS = {
151            **parser.Parser.FUNCTION_PARSERS,
152            "BROADCAST": lambda self: self._parse_join_hint("BROADCAST"),
153            "BROADCASTJOIN": lambda self: self._parse_join_hint("BROADCASTJOIN"),
154            "MAPJOIN": lambda self: self._parse_join_hint("MAPJOIN"),
155            "MERGE": lambda self: self._parse_join_hint("MERGE"),
156            "SHUFFLEMERGE": lambda self: self._parse_join_hint("SHUFFLEMERGE"),
157            "MERGEJOIN": lambda self: self._parse_join_hint("MERGEJOIN"),
158            "SHUFFLE_HASH": lambda self: self._parse_join_hint("SHUFFLE_HASH"),
159            "SHUFFLE_REPLICATE_NL": lambda self: self._parse_join_hint("SHUFFLE_REPLICATE_NL"),
160        }
161
162        def _parse_add_column(self) -> t.Optional[exp.Expression]:
163            return self._match_text_seq("ADD", "COLUMNS") and self._parse_schema()
164
165        def _parse_drop_column(self) -> t.Optional[exp.Drop | exp.Command]:
166            return self._match_text_seq("DROP", "COLUMNS") and self.expression(
167                exp.Drop, this=self._parse_schema(), kind="COLUMNS"
168            )
169
170        def _pivot_column_names(self, aggregations: t.List[exp.Expression]) -> t.List[str]:
171            if len(aggregations) == 1:
172                return [""]
173            return pivot_column_names(aggregations, dialect="spark")
174
175    class Generator(Hive.Generator):
176        TYPE_MAPPING = {
177            **Hive.Generator.TYPE_MAPPING,
178            exp.DataType.Type.TINYINT: "BYTE",
179            exp.DataType.Type.SMALLINT: "SHORT",
180            exp.DataType.Type.BIGINT: "LONG",
181        }
182
183        PROPERTIES_LOCATION = {
184            **Hive.Generator.PROPERTIES_LOCATION,
185            exp.EngineProperty: exp.Properties.Location.UNSUPPORTED,
186            exp.AutoIncrementProperty: exp.Properties.Location.UNSUPPORTED,
187            exp.CharacterSetProperty: exp.Properties.Location.UNSUPPORTED,
188            exp.CollateProperty: exp.Properties.Location.UNSUPPORTED,
189        }
190
191        TRANSFORMS = {
192            **Hive.Generator.TRANSFORMS,
193            exp.ApproxDistinct: rename_func("APPROX_COUNT_DISTINCT"),
194            exp.ArraySum: lambda self, e: f"AGGREGATE({self.sql(e, 'this')}, 0, (acc, x) -> acc + x, acc -> acc)",
195            exp.AtTimeZone: lambda self, e: f"FROM_UTC_TIMESTAMP({self.sql(e, 'this')}, {self.sql(e, 'zone')})",
196            exp.BitwiseLeftShift: rename_func("SHIFTLEFT"),
197            exp.BitwiseRightShift: rename_func("SHIFTRIGHT"),
198            exp.Create: _create_sql,
199            exp.DateFromParts: rename_func("MAKE_DATE"),
200            exp.DateTrunc: lambda self, e: self.func("TRUNC", e.this, e.args.get("unit")),
201            exp.DayOfMonth: rename_func("DAYOFMONTH"),
202            exp.DayOfWeek: rename_func("DAYOFWEEK"),
203            exp.DayOfYear: rename_func("DAYOFYEAR"),
204            exp.FileFormatProperty: lambda self, e: f"USING {e.name.upper()}",
205            exp.From: transforms.preprocess([_unalias_pivot]),
206            exp.Hint: lambda self, e: f" /*+ {self.expressions(e).strip()} */",
207            exp.LogicalAnd: rename_func("BOOL_AND"),
208            exp.LogicalOr: rename_func("BOOL_OR"),
209            exp.Map: _map_sql,
210            exp.Pivot: transforms.preprocess([_unqualify_pivot_columns]),
211            exp.Reduce: rename_func("AGGREGATE"),
212            exp.StrToDate: _str_to_date,
213            exp.StrToTime: lambda self, e: f"TO_TIMESTAMP({self.sql(e, 'this')}, {self.format_time(e)})",
214            exp.TimestampTrunc: lambda self, e: self.func(
215                "DATE_TRUNC", exp.Literal.string(e.text("unit")), e.this
216            ),
217            exp.Trim: trim_sql,
218            exp.UnixToTime: _unix_to_time_sql,
219            exp.VariancePop: rename_func("VAR_POP"),
220            exp.WeekOfYear: rename_func("WEEKOFYEAR"),
221            exp.WithinGroup: transforms.preprocess(
222                [transforms.remove_within_group_for_percentiles]
223            ),
224        }
225        TRANSFORMS.pop(exp.ArrayJoin)
226        TRANSFORMS.pop(exp.ArraySort)
227        TRANSFORMS.pop(exp.ILike)
228        TRANSFORMS.pop(exp.Left)
229        TRANSFORMS.pop(exp.Right)
230
231        WRAP_DERIVED_VALUES = False
232        CREATE_FUNCTION_RETURN_AS = False
233
234        def cast_sql(self, expression: exp.Cast) -> str:
235            if isinstance(expression.this, exp.Cast) and expression.this.is_type("json"):
236                schema = f"'{self.sql(expression, 'to')}'"
237                return self.func("FROM_JSON", expression.this.this, schema)
238            if expression.is_type("json"):
239                return self.func("TO_JSON", expression.this)
240
241            return super(Hive.Generator, self).cast_sql(expression)
242
243        def columndef_sql(self, expression: exp.ColumnDef, sep: str = " ") -> str:
244            return super().columndef_sql(
245                expression,
246                sep=": "
247                if isinstance(expression.parent, exp.DataType)
248                and expression.parent.is_type("struct")
249                else sep,
250            )
251
252    class Tokenizer(Hive.Tokenizer):
253        HEX_STRINGS = [("X'", "'")]
class Spark2(sqlglot.dialects.hive.Hive):
108class Spark2(Hive):
109    class Parser(Hive.Parser):
110        FUNCTIONS = {
111            **Hive.Parser.FUNCTIONS,
112            "MAP_FROM_ARRAYS": exp.Map.from_arg_list,
113            "TO_UNIX_TIMESTAMP": exp.StrToUnix.from_arg_list,
114            "SHIFTLEFT": lambda args: exp.BitwiseLeftShift(
115                this=seq_get(args, 0),
116                expression=seq_get(args, 1),
117            ),
118            "SHIFTRIGHT": lambda args: exp.BitwiseRightShift(
119                this=seq_get(args, 0),
120                expression=seq_get(args, 1),
121            ),
122            "APPROX_PERCENTILE": exp.ApproxQuantile.from_arg_list,
123            "IIF": exp.If.from_arg_list,
124            "AGGREGATE": exp.Reduce.from_arg_list,
125            "DAYOFWEEK": lambda args: exp.DayOfWeek(
126                this=exp.TsOrDsToDate(this=seq_get(args, 0)),
127            ),
128            "DAYOFMONTH": lambda args: exp.DayOfMonth(
129                this=exp.TsOrDsToDate(this=seq_get(args, 0)),
130            ),
131            "DAYOFYEAR": lambda args: exp.DayOfYear(
132                this=exp.TsOrDsToDate(this=seq_get(args, 0)),
133            ),
134            "WEEKOFYEAR": lambda args: exp.WeekOfYear(
135                this=exp.TsOrDsToDate(this=seq_get(args, 0)),
136            ),
137            "DATE_TRUNC": lambda args: exp.TimestampTrunc(
138                this=seq_get(args, 1),
139                unit=exp.var(seq_get(args, 0)),
140            ),
141            "TRUNC": lambda args: exp.DateTrunc(unit=seq_get(args, 1), this=seq_get(args, 0)),
142            "BOOLEAN": _parse_as_cast("boolean"),
143            "DATE": _parse_as_cast("date"),
144            "DOUBLE": _parse_as_cast("double"),
145            "FLOAT": _parse_as_cast("float"),
146            "INT": _parse_as_cast("int"),
147            "STRING": _parse_as_cast("string"),
148            "TIMESTAMP": _parse_as_cast("timestamp"),
149        }
150
151        FUNCTION_PARSERS = {
152            **parser.Parser.FUNCTION_PARSERS,
153            "BROADCAST": lambda self: self._parse_join_hint("BROADCAST"),
154            "BROADCASTJOIN": lambda self: self._parse_join_hint("BROADCASTJOIN"),
155            "MAPJOIN": lambda self: self._parse_join_hint("MAPJOIN"),
156            "MERGE": lambda self: self._parse_join_hint("MERGE"),
157            "SHUFFLEMERGE": lambda self: self._parse_join_hint("SHUFFLEMERGE"),
158            "MERGEJOIN": lambda self: self._parse_join_hint("MERGEJOIN"),
159            "SHUFFLE_HASH": lambda self: self._parse_join_hint("SHUFFLE_HASH"),
160            "SHUFFLE_REPLICATE_NL": lambda self: self._parse_join_hint("SHUFFLE_REPLICATE_NL"),
161        }
162
163        def _parse_add_column(self) -> t.Optional[exp.Expression]:
164            return self._match_text_seq("ADD", "COLUMNS") and self._parse_schema()
165
166        def _parse_drop_column(self) -> t.Optional[exp.Drop | exp.Command]:
167            return self._match_text_seq("DROP", "COLUMNS") and self.expression(
168                exp.Drop, this=self._parse_schema(), kind="COLUMNS"
169            )
170
171        def _pivot_column_names(self, aggregations: t.List[exp.Expression]) -> t.List[str]:
172            if len(aggregations) == 1:
173                return [""]
174            return pivot_column_names(aggregations, dialect="spark")
175
176    class Generator(Hive.Generator):
177        TYPE_MAPPING = {
178            **Hive.Generator.TYPE_MAPPING,
179            exp.DataType.Type.TINYINT: "BYTE",
180            exp.DataType.Type.SMALLINT: "SHORT",
181            exp.DataType.Type.BIGINT: "LONG",
182        }
183
184        PROPERTIES_LOCATION = {
185            **Hive.Generator.PROPERTIES_LOCATION,
186            exp.EngineProperty: exp.Properties.Location.UNSUPPORTED,
187            exp.AutoIncrementProperty: exp.Properties.Location.UNSUPPORTED,
188            exp.CharacterSetProperty: exp.Properties.Location.UNSUPPORTED,
189            exp.CollateProperty: exp.Properties.Location.UNSUPPORTED,
190        }
191
192        TRANSFORMS = {
193            **Hive.Generator.TRANSFORMS,
194            exp.ApproxDistinct: rename_func("APPROX_COUNT_DISTINCT"),
195            exp.ArraySum: lambda self, e: f"AGGREGATE({self.sql(e, 'this')}, 0, (acc, x) -> acc + x, acc -> acc)",
196            exp.AtTimeZone: lambda self, e: f"FROM_UTC_TIMESTAMP({self.sql(e, 'this')}, {self.sql(e, 'zone')})",
197            exp.BitwiseLeftShift: rename_func("SHIFTLEFT"),
198            exp.BitwiseRightShift: rename_func("SHIFTRIGHT"),
199            exp.Create: _create_sql,
200            exp.DateFromParts: rename_func("MAKE_DATE"),
201            exp.DateTrunc: lambda self, e: self.func("TRUNC", e.this, e.args.get("unit")),
202            exp.DayOfMonth: rename_func("DAYOFMONTH"),
203            exp.DayOfWeek: rename_func("DAYOFWEEK"),
204            exp.DayOfYear: rename_func("DAYOFYEAR"),
205            exp.FileFormatProperty: lambda self, e: f"USING {e.name.upper()}",
206            exp.From: transforms.preprocess([_unalias_pivot]),
207            exp.Hint: lambda self, e: f" /*+ {self.expressions(e).strip()} */",
208            exp.LogicalAnd: rename_func("BOOL_AND"),
209            exp.LogicalOr: rename_func("BOOL_OR"),
210            exp.Map: _map_sql,
211            exp.Pivot: transforms.preprocess([_unqualify_pivot_columns]),
212            exp.Reduce: rename_func("AGGREGATE"),
213            exp.StrToDate: _str_to_date,
214            exp.StrToTime: lambda self, e: f"TO_TIMESTAMP({self.sql(e, 'this')}, {self.format_time(e)})",
215            exp.TimestampTrunc: lambda self, e: self.func(
216                "DATE_TRUNC", exp.Literal.string(e.text("unit")), e.this
217            ),
218            exp.Trim: trim_sql,
219            exp.UnixToTime: _unix_to_time_sql,
220            exp.VariancePop: rename_func("VAR_POP"),
221            exp.WeekOfYear: rename_func("WEEKOFYEAR"),
222            exp.WithinGroup: transforms.preprocess(
223                [transforms.remove_within_group_for_percentiles]
224            ),
225        }
226        TRANSFORMS.pop(exp.ArrayJoin)
227        TRANSFORMS.pop(exp.ArraySort)
228        TRANSFORMS.pop(exp.ILike)
229        TRANSFORMS.pop(exp.Left)
230        TRANSFORMS.pop(exp.Right)
231
232        WRAP_DERIVED_VALUES = False
233        CREATE_FUNCTION_RETURN_AS = False
234
235        def cast_sql(self, expression: exp.Cast) -> str:
236            if isinstance(expression.this, exp.Cast) and expression.this.is_type("json"):
237                schema = f"'{self.sql(expression, 'to')}'"
238                return self.func("FROM_JSON", expression.this.this, schema)
239            if expression.is_type("json"):
240                return self.func("TO_JSON", expression.this)
241
242            return super(Hive.Generator, self).cast_sql(expression)
243
244        def columndef_sql(self, expression: exp.ColumnDef, sep: str = " ") -> str:
245            return super().columndef_sql(
246                expression,
247                sep=": "
248                if isinstance(expression.parent, exp.DataType)
249                and expression.parent.is_type("struct")
250                else sep,
251            )
252
253    class Tokenizer(Hive.Tokenizer):
254        HEX_STRINGS = [("X'", "'")]
class Spark2.Parser(sqlglot.dialects.hive.Hive.Parser):
109    class Parser(Hive.Parser):
110        FUNCTIONS = {
111            **Hive.Parser.FUNCTIONS,
112            "MAP_FROM_ARRAYS": exp.Map.from_arg_list,
113            "TO_UNIX_TIMESTAMP": exp.StrToUnix.from_arg_list,
114            "SHIFTLEFT": lambda args: exp.BitwiseLeftShift(
115                this=seq_get(args, 0),
116                expression=seq_get(args, 1),
117            ),
118            "SHIFTRIGHT": lambda args: exp.BitwiseRightShift(
119                this=seq_get(args, 0),
120                expression=seq_get(args, 1),
121            ),
122            "APPROX_PERCENTILE": exp.ApproxQuantile.from_arg_list,
123            "IIF": exp.If.from_arg_list,
124            "AGGREGATE": exp.Reduce.from_arg_list,
125            "DAYOFWEEK": lambda args: exp.DayOfWeek(
126                this=exp.TsOrDsToDate(this=seq_get(args, 0)),
127            ),
128            "DAYOFMONTH": lambda args: exp.DayOfMonth(
129                this=exp.TsOrDsToDate(this=seq_get(args, 0)),
130            ),
131            "DAYOFYEAR": lambda args: exp.DayOfYear(
132                this=exp.TsOrDsToDate(this=seq_get(args, 0)),
133            ),
134            "WEEKOFYEAR": lambda args: exp.WeekOfYear(
135                this=exp.TsOrDsToDate(this=seq_get(args, 0)),
136            ),
137            "DATE_TRUNC": lambda args: exp.TimestampTrunc(
138                this=seq_get(args, 1),
139                unit=exp.var(seq_get(args, 0)),
140            ),
141            "TRUNC": lambda args: exp.DateTrunc(unit=seq_get(args, 1), this=seq_get(args, 0)),
142            "BOOLEAN": _parse_as_cast("boolean"),
143            "DATE": _parse_as_cast("date"),
144            "DOUBLE": _parse_as_cast("double"),
145            "FLOAT": _parse_as_cast("float"),
146            "INT": _parse_as_cast("int"),
147            "STRING": _parse_as_cast("string"),
148            "TIMESTAMP": _parse_as_cast("timestamp"),
149        }
150
151        FUNCTION_PARSERS = {
152            **parser.Parser.FUNCTION_PARSERS,
153            "BROADCAST": lambda self: self._parse_join_hint("BROADCAST"),
154            "BROADCASTJOIN": lambda self: self._parse_join_hint("BROADCASTJOIN"),
155            "MAPJOIN": lambda self: self._parse_join_hint("MAPJOIN"),
156            "MERGE": lambda self: self._parse_join_hint("MERGE"),
157            "SHUFFLEMERGE": lambda self: self._parse_join_hint("SHUFFLEMERGE"),
158            "MERGEJOIN": lambda self: self._parse_join_hint("MERGEJOIN"),
159            "SHUFFLE_HASH": lambda self: self._parse_join_hint("SHUFFLE_HASH"),
160            "SHUFFLE_REPLICATE_NL": lambda self: self._parse_join_hint("SHUFFLE_REPLICATE_NL"),
161        }
162
163        def _parse_add_column(self) -> t.Optional[exp.Expression]:
164            return self._match_text_seq("ADD", "COLUMNS") and self._parse_schema()
165
166        def _parse_drop_column(self) -> t.Optional[exp.Drop | exp.Command]:
167            return self._match_text_seq("DROP", "COLUMNS") and self.expression(
168                exp.Drop, this=self._parse_schema(), kind="COLUMNS"
169            )
170
171        def _pivot_column_names(self, aggregations: t.List[exp.Expression]) -> t.List[str]:
172            if len(aggregations) == 1:
173                return [""]
174            return pivot_column_names(aggregations, dialect="spark")

Parser consumes a list of tokens produced by the Tokenizer and produces a parsed syntax tree.

Arguments:
  • error_level: The desired error level. Default: ErrorLevel.IMMEDIATE
  • error_message_context: Determines the amount of context to capture from a query string when displaying the error message (in number of characters). Default: 100
  • max_errors: Maximum number of error messages to include in a raised ParseError. This is only relevant if error_level is ErrorLevel.RAISE. Default: 3
class Spark2.Generator(sqlglot.dialects.hive.Hive.Generator):
176    class Generator(Hive.Generator):
177        TYPE_MAPPING = {
178            **Hive.Generator.TYPE_MAPPING,
179            exp.DataType.Type.TINYINT: "BYTE",
180            exp.DataType.Type.SMALLINT: "SHORT",
181            exp.DataType.Type.BIGINT: "LONG",
182        }
183
184        PROPERTIES_LOCATION = {
185            **Hive.Generator.PROPERTIES_LOCATION,
186            exp.EngineProperty: exp.Properties.Location.UNSUPPORTED,
187            exp.AutoIncrementProperty: exp.Properties.Location.UNSUPPORTED,
188            exp.CharacterSetProperty: exp.Properties.Location.UNSUPPORTED,
189            exp.CollateProperty: exp.Properties.Location.UNSUPPORTED,
190        }
191
192        TRANSFORMS = {
193            **Hive.Generator.TRANSFORMS,
194            exp.ApproxDistinct: rename_func("APPROX_COUNT_DISTINCT"),
195            exp.ArraySum: lambda self, e: f"AGGREGATE({self.sql(e, 'this')}, 0, (acc, x) -> acc + x, acc -> acc)",
196            exp.AtTimeZone: lambda self, e: f"FROM_UTC_TIMESTAMP({self.sql(e, 'this')}, {self.sql(e, 'zone')})",
197            exp.BitwiseLeftShift: rename_func("SHIFTLEFT"),
198            exp.BitwiseRightShift: rename_func("SHIFTRIGHT"),
199            exp.Create: _create_sql,
200            exp.DateFromParts: rename_func("MAKE_DATE"),
201            exp.DateTrunc: lambda self, e: self.func("TRUNC", e.this, e.args.get("unit")),
202            exp.DayOfMonth: rename_func("DAYOFMONTH"),
203            exp.DayOfWeek: rename_func("DAYOFWEEK"),
204            exp.DayOfYear: rename_func("DAYOFYEAR"),
205            exp.FileFormatProperty: lambda self, e: f"USING {e.name.upper()}",
206            exp.From: transforms.preprocess([_unalias_pivot]),
207            exp.Hint: lambda self, e: f" /*+ {self.expressions(e).strip()} */",
208            exp.LogicalAnd: rename_func("BOOL_AND"),
209            exp.LogicalOr: rename_func("BOOL_OR"),
210            exp.Map: _map_sql,
211            exp.Pivot: transforms.preprocess([_unqualify_pivot_columns]),
212            exp.Reduce: rename_func("AGGREGATE"),
213            exp.StrToDate: _str_to_date,
214            exp.StrToTime: lambda self, e: f"TO_TIMESTAMP({self.sql(e, 'this')}, {self.format_time(e)})",
215            exp.TimestampTrunc: lambda self, e: self.func(
216                "DATE_TRUNC", exp.Literal.string(e.text("unit")), e.this
217            ),
218            exp.Trim: trim_sql,
219            exp.UnixToTime: _unix_to_time_sql,
220            exp.VariancePop: rename_func("VAR_POP"),
221            exp.WeekOfYear: rename_func("WEEKOFYEAR"),
222            exp.WithinGroup: transforms.preprocess(
223                [transforms.remove_within_group_for_percentiles]
224            ),
225        }
226        TRANSFORMS.pop(exp.ArrayJoin)
227        TRANSFORMS.pop(exp.ArraySort)
228        TRANSFORMS.pop(exp.ILike)
229        TRANSFORMS.pop(exp.Left)
230        TRANSFORMS.pop(exp.Right)
231
232        WRAP_DERIVED_VALUES = False
233        CREATE_FUNCTION_RETURN_AS = False
234
235        def cast_sql(self, expression: exp.Cast) -> str:
236            if isinstance(expression.this, exp.Cast) and expression.this.is_type("json"):
237                schema = f"'{self.sql(expression, 'to')}'"
238                return self.func("FROM_JSON", expression.this.this, schema)
239            if expression.is_type("json"):
240                return self.func("TO_JSON", expression.this)
241
242            return super(Hive.Generator, self).cast_sql(expression)
243
244        def columndef_sql(self, expression: exp.ColumnDef, sep: str = " ") -> str:
245            return super().columndef_sql(
246                expression,
247                sep=": "
248                if isinstance(expression.parent, exp.DataType)
249                and expression.parent.is_type("struct")
250                else sep,
251            )

Generator converts a given syntax tree to the corresponding SQL string.

Arguments:
  • pretty: Whether or not to format the produced SQL string. Default: False.
  • identify: Determines when an identifier should be quoted. Possible values are: False (default): Never quote, except in cases where it's mandatory by the dialect. True or 'always': Always quote. 'safe': Only quote identifiers that are case insensitive.
  • normalize: Whether or not to normalize identifiers to lowercase. Default: False.
  • pad: Determines the pad size in a formatted string. Default: 2.
  • indent: Determines the indentation size in a formatted string. Default: 2.
  • normalize_functions: Whether or not to normalize all function names. Possible values are: "upper" or True (default): Convert names to uppercase. "lower": Convert names to lowercase. False: Disables function name normalization.
  • unsupported_level: Determines the generator's behavior when it encounters unsupported expressions. Default ErrorLevel.WARN.
  • max_unsupported: Maximum number of unsupported messages to include in a raised UnsupportedError. This is only relevant if unsupported_level is ErrorLevel.RAISE. Default: 3
  • leading_comma: Determines whether or not the comma is leading or trailing in select expressions. This is only relevant when generating in pretty mode. Default: False
  • max_text_width: The max number of characters in a segment before creating new lines in pretty mode. The default is on the smaller end because the length only represents a segment and not the true line length. Default: 80
  • comments: Whether or not to preserve comments in the output SQL code. Default: True
def cast_sql(self, expression: sqlglot.expressions.Cast) -> str:
235        def cast_sql(self, expression: exp.Cast) -> str:
236            if isinstance(expression.this, exp.Cast) and expression.this.is_type("json"):
237                schema = f"'{self.sql(expression, 'to')}'"
238                return self.func("FROM_JSON", expression.this.this, schema)
239            if expression.is_type("json"):
240                return self.func("TO_JSON", expression.this)
241
242            return super(Hive.Generator, self).cast_sql(expression)
def columndef_sql(self, expression: sqlglot.expressions.ColumnDef, sep: str = ' ') -> str:
244        def columndef_sql(self, expression: exp.ColumnDef, sep: str = " ") -> str:
245            return super().columndef_sql(
246                expression,
247                sep=": "
248                if isinstance(expression.parent, exp.DataType)
249                and expression.parent.is_type("struct")
250                else sep,
251            )
@classmethod
def can_identify(text: str, identify: str | bool = 'safe') -> bool:
247    @classmethod
248    def can_identify(cls, text: str, identify: str | bool = "safe") -> bool:
249        """Checks if text can be identified given an identify option.
250
251        Args:
252            text: The text to check.
253            identify:
254                "always" or `True`: Always returns true.
255                "safe": True if the identifier is case-insensitive.
256
257        Returns:
258            Whether or not the given text can be identified.
259        """
260        if identify is True or identify == "always":
261            return True
262
263        if identify == "safe":
264            return not cls.case_sensitive(text)
265
266        return False

Checks if text can be identified given an identify option.

Arguments:
  • text: The text to check.
  • identify: "always" or True: Always returns true. "safe": True if the identifier is case-insensitive.
Returns:

Whether or not the given text can be identified.

Inherited Members
sqlglot.generator.Generator
Generator
generate
unsupported
sep
seg
pad_comment
maybe_comment
wrap
no_identify
normalize_func
indent
sql
uncache_sql
cache_sql
characterset_sql
column_sql
columnposition_sql
columnconstraint_sql
autoincrementcolumnconstraint_sql
compresscolumnconstraint_sql
generatedasidentitycolumnconstraint_sql
notnullcolumnconstraint_sql
primarykeycolumnconstraint_sql
uniquecolumnconstraint_sql
createable_sql
create_sql
clone_sql
describe_sql
prepend_ctes
with_sql
cte_sql
tablealias_sql
bitstring_sql
hexstring_sql
bytestring_sql
rawstring_sql
datatypesize_sql
directory_sql
delete_sql
drop_sql
except_sql
except_op
fetch_sql
filter_sql
hint_sql
index_sql
identifier_sql
inputoutputformat_sql
national_sql
partition_sql
properties_sql
root_properties
properties
locate_properties
property_sql
likeproperty_sql
fallbackproperty_sql
journalproperty_sql
freespaceproperty_sql
checksumproperty_sql
mergeblockratioproperty_sql
datablocksizeproperty_sql
blockcompressionproperty_sql
isolatedloadingproperty_sql
lockingproperty_sql
withdataproperty_sql
insert_sql
intersect_sql
intersect_op
introducer_sql
pseudotype_sql
onconflict_sql
returning_sql
rowformatdelimitedproperty_sql
table_sql
tablesample_sql
pivot_sql
tuple_sql
update_sql
values_sql
var_sql
into_sql
from_sql
group_sql
having_sql
join_sql
lambda_sql
lateral_sql
limit_sql
offset_sql
setitem_sql
set_sql
pragma_sql
lock_sql
literal_sql
escape_str
loaddata_sql
null_sql
boolean_sql
order_sql
cluster_sql
distribute_sql
sort_sql
ordered_sql
matchrecognize_sql
query_modifiers
offset_limit_modifiers
after_limit_modifiers
select_sql
schema_sql
schema_columns_sql
star_sql
parameter_sql
sessionparameter_sql
placeholder_sql
subquery_sql
qualify_sql
union_sql
union_op
unnest_sql
where_sql
window_sql
partition_by_sql
windowspec_sql
withingroup_sql
between_sql
bracket_sql
all_sql
any_sql
exists_sql
case_sql
constraint_sql
nextvaluefor_sql
extract_sql
trim_sql
safeconcat_sql
check_sql
foreignkey_sql
primarykey_sql
if_sql
matchagainst_sql
jsonkeyvalue_sql
jsonobject_sql
openjsoncolumndef_sql
openjson_sql
in_sql
in_unnest_op
interval_sql
return_sql
reference_sql
anonymous_sql
paren_sql
neg_sql
not_sql
alias_sql
aliases_sql
attimezone_sql
add_sql
and_sql
connector_sql
bitwiseand_sql
bitwiseleftshift_sql
bitwisenot_sql
bitwiseor_sql
bitwiserightshift_sql
bitwisexor_sql
currentdate_sql
collate_sql
command_sql
comment_sql
mergetreettlaction_sql
mergetreettl_sql
transaction_sql
commit_sql
rollback_sql
altercolumn_sql
renametable_sql
altertable_sql
droppartition_sql
addconstraint_sql
distinct_sql
ignorenulls_sql
respectnulls_sql
intdiv_sql
dpipe_sql
safedpipe_sql
div_sql
overlaps_sql
distance_sql
dot_sql
eq_sql
escape_sql
glob_sql
gt_sql
gte_sql
ilike_sql
ilikeany_sql
is_sql
like_sql
likeany_sql
similarto_sql
lt_sql
lte_sql
mod_sql
mul_sql
neq_sql
nullsafeeq_sql
nullsafeneq_sql
or_sql
slice_sql
sub_sql
trycast_sql
use_sql
binary
function_fallback_sql
func
format_args
text_width
format_time
expressions
op_expressions
naked_property
set_operation
tag_sql
token_sql
userdefinedfunction_sql
joinhint_sql
kwarg_sql
when_sql
merge_sql
tochar_sql
dictproperty_sql
dictrange_sql
dictsubproperty_sql
oncluster_sql
sqlglot.dialects.hive.Hive.Generator
arrayagg_sql
with_properties
datatype_sql
after_having_modifiers
class Spark2.Tokenizer(sqlglot.dialects.hive.Hive.Tokenizer):
253    class Tokenizer(Hive.Tokenizer):
254        HEX_STRINGS = [("X'", "'")]