sqlglot.dialects.snowflake
1from __future__ import annotations 2 3import typing as t 4 5from sqlglot import exp, generator, parser, tokens 6from sqlglot.dialects.dialect import ( 7 Dialect, 8 datestrtodate_sql, 9 format_time_lambda, 10 inline_array_sql, 11 min_or_least, 12 rename_func, 13 timestrtotime_sql, 14 ts_or_ds_to_date_sql, 15 var_map_sql, 16) 17from sqlglot.expressions import Literal 18from sqlglot.helper import flatten, seq_get 19from sqlglot.tokens import TokenType 20 21 22def _check_int(s): 23 if s[0] in ("-", "+"): 24 return s[1:].isdigit() 25 return s.isdigit() 26 27 28# from https://docs.snowflake.com/en/sql-reference/functions/to_timestamp.html 29def _snowflake_to_timestamp(args): 30 if len(args) == 2: 31 first_arg, second_arg = args 32 if second_arg.is_string: 33 # case: <string_expr> [ , <format> ] 34 return format_time_lambda(exp.StrToTime, "snowflake")(args) 35 36 # case: <numeric_expr> [ , <scale> ] 37 if second_arg.name not in ["0", "3", "9"]: 38 raise ValueError( 39 f"Scale for snowflake numeric timestamp is {second_arg}, but should be 0, 3, or 9" 40 ) 41 42 if second_arg.name == "0": 43 timescale = exp.UnixToTime.SECONDS 44 elif second_arg.name == "3": 45 timescale = exp.UnixToTime.MILLIS 46 elif second_arg.name == "9": 47 timescale = exp.UnixToTime.MICROS 48 49 return exp.UnixToTime(this=first_arg, scale=timescale) 50 51 first_arg = seq_get(args, 0) 52 if not isinstance(first_arg, Literal): 53 # case: <variant_expr> 54 return format_time_lambda(exp.StrToTime, "snowflake", default=True)(args) 55 56 if first_arg.is_string: 57 if _check_int(first_arg.this): 58 # case: <integer> 59 return exp.UnixToTime.from_arg_list(args) 60 61 # case: <date_expr> 62 return format_time_lambda(exp.StrToTime, "snowflake", default=True)(args) 63 64 # case: <numeric_expr> 65 return exp.UnixToTime.from_arg_list(args) 66 67 68def _unix_to_time_sql(self, expression): 69 scale = expression.args.get("scale") 70 timestamp = self.sql(expression, "this") 71 if scale in [None, exp.UnixToTime.SECONDS]: 72 return f"TO_TIMESTAMP({timestamp})" 73 if scale == exp.UnixToTime.MILLIS: 74 return f"TO_TIMESTAMP({timestamp}, 3)" 75 if scale == exp.UnixToTime.MICROS: 76 return f"TO_TIMESTAMP({timestamp}, 9)" 77 78 raise ValueError("Improper scale for timestamp") 79 80 81# https://docs.snowflake.com/en/sql-reference/functions/date_part.html 82# https://docs.snowflake.com/en/sql-reference/functions-date-time.html#label-supported-date-time-parts 83def _parse_date_part(self): 84 this = self._parse_var() or self._parse_type() 85 self._match(TokenType.COMMA) 86 expression = self._parse_bitwise() 87 88 name = this.name.upper() 89 if name.startswith("EPOCH"): 90 if name.startswith("EPOCH_MILLISECOND"): 91 scale = 10**3 92 elif name.startswith("EPOCH_MICROSECOND"): 93 scale = 10**6 94 elif name.startswith("EPOCH_NANOSECOND"): 95 scale = 10**9 96 else: 97 scale = None 98 99 ts = self.expression(exp.Cast, this=expression, to=exp.DataType.build("TIMESTAMP")) 100 to_unix = self.expression(exp.TimeToUnix, this=ts) 101 102 if scale: 103 to_unix = exp.Mul(this=to_unix, expression=exp.Literal.number(scale)) 104 105 return to_unix 106 107 return self.expression(exp.Extract, this=this, expression=expression) 108 109 110# https://docs.snowflake.com/en/sql-reference/functions/div0 111def _div0_to_if(args): 112 cond = exp.EQ(this=seq_get(args, 1), expression=exp.Literal.number(0)) 113 true = exp.Literal.number(0) 114 false = exp.FloatDiv(this=seq_get(args, 0), expression=seq_get(args, 1)) 115 return exp.If(this=cond, true=true, false=false) 116 117 118# https://docs.snowflake.com/en/sql-reference/functions/zeroifnull 119def _zeroifnull_to_if(args): 120 cond = exp.Is(this=seq_get(args, 0), expression=exp.Null()) 121 return exp.If(this=cond, true=exp.Literal.number(0), false=seq_get(args, 0)) 122 123 124# https://docs.snowflake.com/en/sql-reference/functions/zeroifnull 125def _nullifzero_to_if(args): 126 cond = exp.EQ(this=seq_get(args, 0), expression=exp.Literal.number(0)) 127 return exp.If(this=cond, true=exp.Null(), false=seq_get(args, 0)) 128 129 130def _datatype_sql(self, expression): 131 if expression.this == exp.DataType.Type.ARRAY: 132 return "ARRAY" 133 elif expression.this == exp.DataType.Type.MAP: 134 return "OBJECT" 135 return self.datatype_sql(expression) 136 137 138class Snowflake(Dialect): 139 null_ordering = "nulls_are_large" 140 time_format = "'yyyy-mm-dd hh24:mi:ss'" 141 142 time_mapping = { 143 "YYYY": "%Y", 144 "yyyy": "%Y", 145 "YY": "%y", 146 "yy": "%y", 147 "MMMM": "%B", 148 "mmmm": "%B", 149 "MON": "%b", 150 "mon": "%b", 151 "MM": "%m", 152 "mm": "%m", 153 "DD": "%d", 154 "dd": "%d", 155 "d": "%-d", 156 "DY": "%w", 157 "dy": "%w", 158 "HH24": "%H", 159 "hh24": "%H", 160 "HH12": "%I", 161 "hh12": "%I", 162 "MI": "%M", 163 "mi": "%M", 164 "SS": "%S", 165 "ss": "%S", 166 "FF": "%f", 167 "ff": "%f", 168 "FF6": "%f", 169 "ff6": "%f", 170 } 171 172 class Parser(parser.Parser): 173 FUNCTIONS = { 174 **parser.Parser.FUNCTIONS, 175 "ARRAYAGG": exp.ArrayAgg.from_arg_list, 176 "ARRAY_TO_STRING": exp.ArrayJoin.from_arg_list, 177 "DATEADD": lambda args: exp.DateAdd( 178 this=seq_get(args, 2), 179 expression=seq_get(args, 1), 180 unit=seq_get(args, 0), 181 ), 182 "DATE_TRUNC": lambda args: exp.DateTrunc( 183 unit=exp.Literal.string(seq_get(args, 0).name), # type: ignore 184 this=seq_get(args, 1), 185 ), 186 "DIV0": _div0_to_if, 187 "IFF": exp.If.from_arg_list, 188 "TO_ARRAY": exp.Array.from_arg_list, 189 "TO_TIMESTAMP": _snowflake_to_timestamp, 190 "ARRAY_CONSTRUCT": exp.Array.from_arg_list, 191 "RLIKE": exp.RegexpLike.from_arg_list, 192 "DECODE": exp.Matches.from_arg_list, 193 "OBJECT_CONSTRUCT": parser.parse_var_map, 194 "ZEROIFNULL": _zeroifnull_to_if, 195 "NULLIFZERO": _nullifzero_to_if, 196 } 197 198 FUNCTION_PARSERS = { 199 **parser.Parser.FUNCTION_PARSERS, 200 "DATE_PART": _parse_date_part, 201 } 202 FUNCTION_PARSERS.pop("TRIM") 203 204 FUNC_TOKENS = { 205 *parser.Parser.FUNC_TOKENS, 206 TokenType.RLIKE, 207 TokenType.TABLE, 208 } 209 210 COLUMN_OPERATORS = { 211 **parser.Parser.COLUMN_OPERATORS, # type: ignore 212 TokenType.COLON: lambda self, this, path: self.expression( 213 exp.Bracket, 214 this=this, 215 expressions=[path], 216 ), 217 } 218 219 RANGE_PARSERS = { 220 **parser.Parser.RANGE_PARSERS, # type: ignore 221 TokenType.LIKE_ANY: lambda self, this: self._parse_escape( 222 self.expression(exp.LikeAny, this=this, expression=self._parse_bitwise()) 223 ), 224 TokenType.ILIKE_ANY: lambda self, this: self._parse_escape( 225 self.expression(exp.ILikeAny, this=this, expression=self._parse_bitwise()) 226 ), 227 } 228 229 ALTER_PARSERS = { 230 **parser.Parser.ALTER_PARSERS, # type: ignore 231 "UNSET": lambda self: self._parse_alter_table_set_tag(unset=True), 232 "SET": lambda self: self._parse_alter_table_set_tag(), 233 } 234 235 INTEGER_DIVISION = False 236 237 def _parse_alter_table_set_tag(self, unset: bool = False) -> exp.Expression: 238 self._match_text_seq("TAG") 239 parser = t.cast(t.Callable, self._parse_id_var if unset else self._parse_conjunction) 240 return self.expression(exp.SetTag, expressions=self._parse_csv(parser), unset=unset) 241 242 class Tokenizer(tokens.Tokenizer): 243 QUOTES = ["'", "$$"] 244 STRING_ESCAPES = ["\\", "'"] 245 246 KEYWORDS = { 247 **tokens.Tokenizer.KEYWORDS, 248 "EXCLUDE": TokenType.EXCEPT, 249 "ILIKE ANY": TokenType.ILIKE_ANY, 250 "LIKE ANY": TokenType.LIKE_ANY, 251 "MATCH_RECOGNIZE": TokenType.MATCH_RECOGNIZE, 252 "PUT": TokenType.COMMAND, 253 "RENAME": TokenType.REPLACE, 254 "TIMESTAMP_LTZ": TokenType.TIMESTAMPLTZ, 255 "TIMESTAMP_NTZ": TokenType.TIMESTAMP, 256 "TIMESTAMP_TZ": TokenType.TIMESTAMPTZ, 257 "TIMESTAMPNTZ": TokenType.TIMESTAMP, 258 "MINUS": TokenType.EXCEPT, 259 "SAMPLE": TokenType.TABLE_SAMPLE, 260 } 261 262 SINGLE_TOKENS = { 263 **tokens.Tokenizer.SINGLE_TOKENS, 264 "$": TokenType.PARAMETER, 265 } 266 267 class Generator(generator.Generator): 268 PARAMETER_TOKEN = "$" 269 INTEGER_DIVISION = False 270 MATCHED_BY_SOURCE = False 271 272 TRANSFORMS = { 273 **generator.Generator.TRANSFORMS, # type: ignore 274 exp.Array: inline_array_sql, 275 exp.ArrayConcat: rename_func("ARRAY_CAT"), 276 exp.ArrayJoin: rename_func("ARRAY_TO_STRING"), 277 exp.DateAdd: lambda self, e: self.func("DATEADD", e.text("unit"), e.expression, e.this), 278 exp.DateStrToDate: datestrtodate_sql, 279 exp.DataType: _datatype_sql, 280 exp.If: rename_func("IFF"), 281 exp.Map: lambda self, e: var_map_sql(self, e, "OBJECT_CONSTRUCT"), 282 exp.VarMap: lambda self, e: var_map_sql(self, e, "OBJECT_CONSTRUCT"), 283 exp.PartitionedByProperty: lambda self, e: f"PARTITION BY {self.sql(e, 'this')}", 284 exp.Matches: rename_func("DECODE"), 285 exp.StrPosition: lambda self, e: self.func( 286 "POSITION", e.args.get("substr"), e.this, e.args.get("position") 287 ), 288 exp.StrToTime: lambda self, e: f"TO_TIMESTAMP({self.sql(e, 'this')}, {self.format_time(e)})", 289 exp.TimeStrToTime: timestrtotime_sql, 290 exp.TimeToUnix: lambda self, e: f"EXTRACT(epoch_second FROM {self.sql(e, 'this')})", 291 exp.Trim: lambda self, e: self.func("TRIM", e.this, e.expression), 292 exp.TsOrDsToDate: ts_or_ds_to_date_sql("snowflake"), 293 exp.UnixToTime: _unix_to_time_sql, 294 exp.DayOfWeek: rename_func("DAYOFWEEK"), 295 exp.Min: min_or_least, 296 } 297 298 TYPE_MAPPING = { 299 **generator.Generator.TYPE_MAPPING, # type: ignore 300 exp.DataType.Type.TIMESTAMP: "TIMESTAMPNTZ", 301 } 302 303 STAR_MAPPING = { 304 "except": "EXCLUDE", 305 "replace": "RENAME", 306 } 307 308 PROPERTIES_LOCATION = { 309 **generator.Generator.PROPERTIES_LOCATION, # type: ignore 310 exp.SetProperty: exp.Properties.Location.UNSUPPORTED, 311 } 312 313 def ilikeany_sql(self, expression: exp.ILikeAny) -> str: 314 return self.binary(expression, "ILIKE ANY") 315 316 def likeany_sql(self, expression: exp.LikeAny) -> str: 317 return self.binary(expression, "LIKE ANY") 318 319 def except_op(self, expression): 320 if not expression.args.get("distinct", False): 321 self.unsupported("EXCEPT with All is not supported in Snowflake") 322 return super().except_op(expression) 323 324 def intersect_op(self, expression): 325 if not expression.args.get("distinct", False): 326 self.unsupported("INTERSECT with All is not supported in Snowflake") 327 return super().intersect_op(expression) 328 329 def values_sql(self, expression: exp.Values) -> str: 330 """Due to a bug in Snowflake we want to make sure that all columns in a VALUES table alias are unquoted. 331 332 We also want to make sure that after we find matches where we need to unquote a column that we prevent users 333 from adding quotes to the column by using the `identify` argument when generating the SQL. 334 """ 335 alias = expression.args.get("alias") 336 if alias and alias.args.get("columns"): 337 expression = expression.transform( 338 lambda node: exp.Identifier(**{**node.args, "quoted": False}) 339 if isinstance(node, exp.Identifier) 340 and isinstance(node.parent, exp.TableAlias) 341 and node.arg_key == "columns" 342 else node, 343 ) 344 return self.no_identify(lambda: super(self.__class__, self).values_sql(expression)) 345 return super().values_sql(expression) 346 347 def settag_sql(self, expression: exp.SetTag) -> str: 348 action = "UNSET" if expression.args.get("unset") else "SET" 349 return f"{action} TAG {self.expressions(expression)}" 350 351 def select_sql(self, expression: exp.Select) -> str: 352 """Due to a bug in Snowflake we want to make sure that all columns in a VALUES table alias are unquoted and also 353 that all columns in a SELECT are unquoted. We also want to make sure that after we find matches where we need 354 to unquote a column that we prevent users from adding quotes to the column by using the `identify` argument when 355 generating the SQL. 356 357 Note: We make an assumption that any columns referenced in a VALUES expression should be unquoted throughout the 358 expression. This might not be true in a case where the same column name can be sourced from another table that can 359 properly quote but should be true in most cases. 360 """ 361 values_identifiers = set( 362 flatten( 363 (v.args.get("alias") or exp.Alias()).args.get("columns", []) 364 for v in expression.find_all(exp.Values) 365 ) 366 ) 367 if values_identifiers: 368 expression = expression.transform( 369 lambda node: exp.Identifier(**{**node.args, "quoted": False}) 370 if isinstance(node, exp.Identifier) and node in values_identifiers 371 else node, 372 ) 373 return self.no_identify(lambda: super(self.__class__, self).select_sql(expression)) 374 return super().select_sql(expression) 375 376 def describe_sql(self, expression: exp.Describe) -> str: 377 # Default to table if kind is unknown 378 kind_value = expression.args.get("kind") or "TABLE" 379 kind = f" {kind_value}" if kind_value else "" 380 this = f" {self.sql(expression, 'this')}" 381 return f"DESCRIBE{kind}{this}" 382 383 def generatedasidentitycolumnconstraint_sql( 384 self, expression: exp.GeneratedAsIdentityColumnConstraint 385 ) -> str: 386 start = expression.args.get("start") 387 start = f" START {start}" if start else "" 388 increment = expression.args.get("increment") 389 increment = f" INCREMENT {increment}" if increment else "" 390 return f"AUTOINCREMENT{start}{increment}"
139class Snowflake(Dialect): 140 null_ordering = "nulls_are_large" 141 time_format = "'yyyy-mm-dd hh24:mi:ss'" 142 143 time_mapping = { 144 "YYYY": "%Y", 145 "yyyy": "%Y", 146 "YY": "%y", 147 "yy": "%y", 148 "MMMM": "%B", 149 "mmmm": "%B", 150 "MON": "%b", 151 "mon": "%b", 152 "MM": "%m", 153 "mm": "%m", 154 "DD": "%d", 155 "dd": "%d", 156 "d": "%-d", 157 "DY": "%w", 158 "dy": "%w", 159 "HH24": "%H", 160 "hh24": "%H", 161 "HH12": "%I", 162 "hh12": "%I", 163 "MI": "%M", 164 "mi": "%M", 165 "SS": "%S", 166 "ss": "%S", 167 "FF": "%f", 168 "ff": "%f", 169 "FF6": "%f", 170 "ff6": "%f", 171 } 172 173 class Parser(parser.Parser): 174 FUNCTIONS = { 175 **parser.Parser.FUNCTIONS, 176 "ARRAYAGG": exp.ArrayAgg.from_arg_list, 177 "ARRAY_TO_STRING": exp.ArrayJoin.from_arg_list, 178 "DATEADD": lambda args: exp.DateAdd( 179 this=seq_get(args, 2), 180 expression=seq_get(args, 1), 181 unit=seq_get(args, 0), 182 ), 183 "DATE_TRUNC": lambda args: exp.DateTrunc( 184 unit=exp.Literal.string(seq_get(args, 0).name), # type: ignore 185 this=seq_get(args, 1), 186 ), 187 "DIV0": _div0_to_if, 188 "IFF": exp.If.from_arg_list, 189 "TO_ARRAY": exp.Array.from_arg_list, 190 "TO_TIMESTAMP": _snowflake_to_timestamp, 191 "ARRAY_CONSTRUCT": exp.Array.from_arg_list, 192 "RLIKE": exp.RegexpLike.from_arg_list, 193 "DECODE": exp.Matches.from_arg_list, 194 "OBJECT_CONSTRUCT": parser.parse_var_map, 195 "ZEROIFNULL": _zeroifnull_to_if, 196 "NULLIFZERO": _nullifzero_to_if, 197 } 198 199 FUNCTION_PARSERS = { 200 **parser.Parser.FUNCTION_PARSERS, 201 "DATE_PART": _parse_date_part, 202 } 203 FUNCTION_PARSERS.pop("TRIM") 204 205 FUNC_TOKENS = { 206 *parser.Parser.FUNC_TOKENS, 207 TokenType.RLIKE, 208 TokenType.TABLE, 209 } 210 211 COLUMN_OPERATORS = { 212 **parser.Parser.COLUMN_OPERATORS, # type: ignore 213 TokenType.COLON: lambda self, this, path: self.expression( 214 exp.Bracket, 215 this=this, 216 expressions=[path], 217 ), 218 } 219 220 RANGE_PARSERS = { 221 **parser.Parser.RANGE_PARSERS, # type: ignore 222 TokenType.LIKE_ANY: lambda self, this: self._parse_escape( 223 self.expression(exp.LikeAny, this=this, expression=self._parse_bitwise()) 224 ), 225 TokenType.ILIKE_ANY: lambda self, this: self._parse_escape( 226 self.expression(exp.ILikeAny, this=this, expression=self._parse_bitwise()) 227 ), 228 } 229 230 ALTER_PARSERS = { 231 **parser.Parser.ALTER_PARSERS, # type: ignore 232 "UNSET": lambda self: self._parse_alter_table_set_tag(unset=True), 233 "SET": lambda self: self._parse_alter_table_set_tag(), 234 } 235 236 INTEGER_DIVISION = False 237 238 def _parse_alter_table_set_tag(self, unset: bool = False) -> exp.Expression: 239 self._match_text_seq("TAG") 240 parser = t.cast(t.Callable, self._parse_id_var if unset else self._parse_conjunction) 241 return self.expression(exp.SetTag, expressions=self._parse_csv(parser), unset=unset) 242 243 class Tokenizer(tokens.Tokenizer): 244 QUOTES = ["'", "$$"] 245 STRING_ESCAPES = ["\\", "'"] 246 247 KEYWORDS = { 248 **tokens.Tokenizer.KEYWORDS, 249 "EXCLUDE": TokenType.EXCEPT, 250 "ILIKE ANY": TokenType.ILIKE_ANY, 251 "LIKE ANY": TokenType.LIKE_ANY, 252 "MATCH_RECOGNIZE": TokenType.MATCH_RECOGNIZE, 253 "PUT": TokenType.COMMAND, 254 "RENAME": TokenType.REPLACE, 255 "TIMESTAMP_LTZ": TokenType.TIMESTAMPLTZ, 256 "TIMESTAMP_NTZ": TokenType.TIMESTAMP, 257 "TIMESTAMP_TZ": TokenType.TIMESTAMPTZ, 258 "TIMESTAMPNTZ": TokenType.TIMESTAMP, 259 "MINUS": TokenType.EXCEPT, 260 "SAMPLE": TokenType.TABLE_SAMPLE, 261 } 262 263 SINGLE_TOKENS = { 264 **tokens.Tokenizer.SINGLE_TOKENS, 265 "$": TokenType.PARAMETER, 266 } 267 268 class Generator(generator.Generator): 269 PARAMETER_TOKEN = "$" 270 INTEGER_DIVISION = False 271 MATCHED_BY_SOURCE = False 272 273 TRANSFORMS = { 274 **generator.Generator.TRANSFORMS, # type: ignore 275 exp.Array: inline_array_sql, 276 exp.ArrayConcat: rename_func("ARRAY_CAT"), 277 exp.ArrayJoin: rename_func("ARRAY_TO_STRING"), 278 exp.DateAdd: lambda self, e: self.func("DATEADD", e.text("unit"), e.expression, e.this), 279 exp.DateStrToDate: datestrtodate_sql, 280 exp.DataType: _datatype_sql, 281 exp.If: rename_func("IFF"), 282 exp.Map: lambda self, e: var_map_sql(self, e, "OBJECT_CONSTRUCT"), 283 exp.VarMap: lambda self, e: var_map_sql(self, e, "OBJECT_CONSTRUCT"), 284 exp.PartitionedByProperty: lambda self, e: f"PARTITION BY {self.sql(e, 'this')}", 285 exp.Matches: rename_func("DECODE"), 286 exp.StrPosition: lambda self, e: self.func( 287 "POSITION", e.args.get("substr"), e.this, e.args.get("position") 288 ), 289 exp.StrToTime: lambda self, e: f"TO_TIMESTAMP({self.sql(e, 'this')}, {self.format_time(e)})", 290 exp.TimeStrToTime: timestrtotime_sql, 291 exp.TimeToUnix: lambda self, e: f"EXTRACT(epoch_second FROM {self.sql(e, 'this')})", 292 exp.Trim: lambda self, e: self.func("TRIM", e.this, e.expression), 293 exp.TsOrDsToDate: ts_or_ds_to_date_sql("snowflake"), 294 exp.UnixToTime: _unix_to_time_sql, 295 exp.DayOfWeek: rename_func("DAYOFWEEK"), 296 exp.Min: min_or_least, 297 } 298 299 TYPE_MAPPING = { 300 **generator.Generator.TYPE_MAPPING, # type: ignore 301 exp.DataType.Type.TIMESTAMP: "TIMESTAMPNTZ", 302 } 303 304 STAR_MAPPING = { 305 "except": "EXCLUDE", 306 "replace": "RENAME", 307 } 308 309 PROPERTIES_LOCATION = { 310 **generator.Generator.PROPERTIES_LOCATION, # type: ignore 311 exp.SetProperty: exp.Properties.Location.UNSUPPORTED, 312 } 313 314 def ilikeany_sql(self, expression: exp.ILikeAny) -> str: 315 return self.binary(expression, "ILIKE ANY") 316 317 def likeany_sql(self, expression: exp.LikeAny) -> str: 318 return self.binary(expression, "LIKE ANY") 319 320 def except_op(self, expression): 321 if not expression.args.get("distinct", False): 322 self.unsupported("EXCEPT with All is not supported in Snowflake") 323 return super().except_op(expression) 324 325 def intersect_op(self, expression): 326 if not expression.args.get("distinct", False): 327 self.unsupported("INTERSECT with All is not supported in Snowflake") 328 return super().intersect_op(expression) 329 330 def values_sql(self, expression: exp.Values) -> str: 331 """Due to a bug in Snowflake we want to make sure that all columns in a VALUES table alias are unquoted. 332 333 We also want to make sure that after we find matches where we need to unquote a column that we prevent users 334 from adding quotes to the column by using the `identify` argument when generating the SQL. 335 """ 336 alias = expression.args.get("alias") 337 if alias and alias.args.get("columns"): 338 expression = expression.transform( 339 lambda node: exp.Identifier(**{**node.args, "quoted": False}) 340 if isinstance(node, exp.Identifier) 341 and isinstance(node.parent, exp.TableAlias) 342 and node.arg_key == "columns" 343 else node, 344 ) 345 return self.no_identify(lambda: super(self.__class__, self).values_sql(expression)) 346 return super().values_sql(expression) 347 348 def settag_sql(self, expression: exp.SetTag) -> str: 349 action = "UNSET" if expression.args.get("unset") else "SET" 350 return f"{action} TAG {self.expressions(expression)}" 351 352 def select_sql(self, expression: exp.Select) -> str: 353 """Due to a bug in Snowflake we want to make sure that all columns in a VALUES table alias are unquoted and also 354 that all columns in a SELECT are unquoted. We also want to make sure that after we find matches where we need 355 to unquote a column that we prevent users from adding quotes to the column by using the `identify` argument when 356 generating the SQL. 357 358 Note: We make an assumption that any columns referenced in a VALUES expression should be unquoted throughout the 359 expression. This might not be true in a case where the same column name can be sourced from another table that can 360 properly quote but should be true in most cases. 361 """ 362 values_identifiers = set( 363 flatten( 364 (v.args.get("alias") or exp.Alias()).args.get("columns", []) 365 for v in expression.find_all(exp.Values) 366 ) 367 ) 368 if values_identifiers: 369 expression = expression.transform( 370 lambda node: exp.Identifier(**{**node.args, "quoted": False}) 371 if isinstance(node, exp.Identifier) and node in values_identifiers 372 else node, 373 ) 374 return self.no_identify(lambda: super(self.__class__, self).select_sql(expression)) 375 return super().select_sql(expression) 376 377 def describe_sql(self, expression: exp.Describe) -> str: 378 # Default to table if kind is unknown 379 kind_value = expression.args.get("kind") or "TABLE" 380 kind = f" {kind_value}" if kind_value else "" 381 this = f" {self.sql(expression, 'this')}" 382 return f"DESCRIBE{kind}{this}" 383 384 def generatedasidentitycolumnconstraint_sql( 385 self, expression: exp.GeneratedAsIdentityColumnConstraint 386 ) -> str: 387 start = expression.args.get("start") 388 start = f" START {start}" if start else "" 389 increment = expression.args.get("increment") 390 increment = f" INCREMENT {increment}" if increment else "" 391 return f"AUTOINCREMENT{start}{increment}"
Inherited Members
173 class Parser(parser.Parser): 174 FUNCTIONS = { 175 **parser.Parser.FUNCTIONS, 176 "ARRAYAGG": exp.ArrayAgg.from_arg_list, 177 "ARRAY_TO_STRING": exp.ArrayJoin.from_arg_list, 178 "DATEADD": lambda args: exp.DateAdd( 179 this=seq_get(args, 2), 180 expression=seq_get(args, 1), 181 unit=seq_get(args, 0), 182 ), 183 "DATE_TRUNC": lambda args: exp.DateTrunc( 184 unit=exp.Literal.string(seq_get(args, 0).name), # type: ignore 185 this=seq_get(args, 1), 186 ), 187 "DIV0": _div0_to_if, 188 "IFF": exp.If.from_arg_list, 189 "TO_ARRAY": exp.Array.from_arg_list, 190 "TO_TIMESTAMP": _snowflake_to_timestamp, 191 "ARRAY_CONSTRUCT": exp.Array.from_arg_list, 192 "RLIKE": exp.RegexpLike.from_arg_list, 193 "DECODE": exp.Matches.from_arg_list, 194 "OBJECT_CONSTRUCT": parser.parse_var_map, 195 "ZEROIFNULL": _zeroifnull_to_if, 196 "NULLIFZERO": _nullifzero_to_if, 197 } 198 199 FUNCTION_PARSERS = { 200 **parser.Parser.FUNCTION_PARSERS, 201 "DATE_PART": _parse_date_part, 202 } 203 FUNCTION_PARSERS.pop("TRIM") 204 205 FUNC_TOKENS = { 206 *parser.Parser.FUNC_TOKENS, 207 TokenType.RLIKE, 208 TokenType.TABLE, 209 } 210 211 COLUMN_OPERATORS = { 212 **parser.Parser.COLUMN_OPERATORS, # type: ignore 213 TokenType.COLON: lambda self, this, path: self.expression( 214 exp.Bracket, 215 this=this, 216 expressions=[path], 217 ), 218 } 219 220 RANGE_PARSERS = { 221 **parser.Parser.RANGE_PARSERS, # type: ignore 222 TokenType.LIKE_ANY: lambda self, this: self._parse_escape( 223 self.expression(exp.LikeAny, this=this, expression=self._parse_bitwise()) 224 ), 225 TokenType.ILIKE_ANY: lambda self, this: self._parse_escape( 226 self.expression(exp.ILikeAny, this=this, expression=self._parse_bitwise()) 227 ), 228 } 229 230 ALTER_PARSERS = { 231 **parser.Parser.ALTER_PARSERS, # type: ignore 232 "UNSET": lambda self: self._parse_alter_table_set_tag(unset=True), 233 "SET": lambda self: self._parse_alter_table_set_tag(), 234 } 235 236 INTEGER_DIVISION = False 237 238 def _parse_alter_table_set_tag(self, unset: bool = False) -> exp.Expression: 239 self._match_text_seq("TAG") 240 parser = t.cast(t.Callable, self._parse_id_var if unset else self._parse_conjunction) 241 return self.expression(exp.SetTag, expressions=self._parse_csv(parser), unset=unset)
Parser consumes a list of tokens produced by the sqlglot.tokens.Tokenizer
and produces
a parsed syntax tree.
Arguments:
- error_level: the desired error level. Default: ErrorLevel.RAISE
- error_message_context: determines the amount of context to capture from a query string when displaying the error message (in number of characters). Default: 50.
- index_offset: Index offset for arrays eg ARRAY[0] vs ARRAY[1] as the head of a list. Default: 0
- alias_post_tablesample: If the table alias comes after tablesample. Default: False
- 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
- null_ordering: Indicates the default null ordering method to use if not explicitly set. Options are "nulls_are_small", "nulls_are_large", "nulls_are_last". Default: "nulls_are_small"
Inherited Members
243 class Tokenizer(tokens.Tokenizer): 244 QUOTES = ["'", "$$"] 245 STRING_ESCAPES = ["\\", "'"] 246 247 KEYWORDS = { 248 **tokens.Tokenizer.KEYWORDS, 249 "EXCLUDE": TokenType.EXCEPT, 250 "ILIKE ANY": TokenType.ILIKE_ANY, 251 "LIKE ANY": TokenType.LIKE_ANY, 252 "MATCH_RECOGNIZE": TokenType.MATCH_RECOGNIZE, 253 "PUT": TokenType.COMMAND, 254 "RENAME": TokenType.REPLACE, 255 "TIMESTAMP_LTZ": TokenType.TIMESTAMPLTZ, 256 "TIMESTAMP_NTZ": TokenType.TIMESTAMP, 257 "TIMESTAMP_TZ": TokenType.TIMESTAMPTZ, 258 "TIMESTAMPNTZ": TokenType.TIMESTAMP, 259 "MINUS": TokenType.EXCEPT, 260 "SAMPLE": TokenType.TABLE_SAMPLE, 261 } 262 263 SINGLE_TOKENS = { 264 **tokens.Tokenizer.SINGLE_TOKENS, 265 "$": TokenType.PARAMETER, 266 }
Inherited Members
268 class Generator(generator.Generator): 269 PARAMETER_TOKEN = "$" 270 INTEGER_DIVISION = False 271 MATCHED_BY_SOURCE = False 272 273 TRANSFORMS = { 274 **generator.Generator.TRANSFORMS, # type: ignore 275 exp.Array: inline_array_sql, 276 exp.ArrayConcat: rename_func("ARRAY_CAT"), 277 exp.ArrayJoin: rename_func("ARRAY_TO_STRING"), 278 exp.DateAdd: lambda self, e: self.func("DATEADD", e.text("unit"), e.expression, e.this), 279 exp.DateStrToDate: datestrtodate_sql, 280 exp.DataType: _datatype_sql, 281 exp.If: rename_func("IFF"), 282 exp.Map: lambda self, e: var_map_sql(self, e, "OBJECT_CONSTRUCT"), 283 exp.VarMap: lambda self, e: var_map_sql(self, e, "OBJECT_CONSTRUCT"), 284 exp.PartitionedByProperty: lambda self, e: f"PARTITION BY {self.sql(e, 'this')}", 285 exp.Matches: rename_func("DECODE"), 286 exp.StrPosition: lambda self, e: self.func( 287 "POSITION", e.args.get("substr"), e.this, e.args.get("position") 288 ), 289 exp.StrToTime: lambda self, e: f"TO_TIMESTAMP({self.sql(e, 'this')}, {self.format_time(e)})", 290 exp.TimeStrToTime: timestrtotime_sql, 291 exp.TimeToUnix: lambda self, e: f"EXTRACT(epoch_second FROM {self.sql(e, 'this')})", 292 exp.Trim: lambda self, e: self.func("TRIM", e.this, e.expression), 293 exp.TsOrDsToDate: ts_or_ds_to_date_sql("snowflake"), 294 exp.UnixToTime: _unix_to_time_sql, 295 exp.DayOfWeek: rename_func("DAYOFWEEK"), 296 exp.Min: min_or_least, 297 } 298 299 TYPE_MAPPING = { 300 **generator.Generator.TYPE_MAPPING, # type: ignore 301 exp.DataType.Type.TIMESTAMP: "TIMESTAMPNTZ", 302 } 303 304 STAR_MAPPING = { 305 "except": "EXCLUDE", 306 "replace": "RENAME", 307 } 308 309 PROPERTIES_LOCATION = { 310 **generator.Generator.PROPERTIES_LOCATION, # type: ignore 311 exp.SetProperty: exp.Properties.Location.UNSUPPORTED, 312 } 313 314 def ilikeany_sql(self, expression: exp.ILikeAny) -> str: 315 return self.binary(expression, "ILIKE ANY") 316 317 def likeany_sql(self, expression: exp.LikeAny) -> str: 318 return self.binary(expression, "LIKE ANY") 319 320 def except_op(self, expression): 321 if not expression.args.get("distinct", False): 322 self.unsupported("EXCEPT with All is not supported in Snowflake") 323 return super().except_op(expression) 324 325 def intersect_op(self, expression): 326 if not expression.args.get("distinct", False): 327 self.unsupported("INTERSECT with All is not supported in Snowflake") 328 return super().intersect_op(expression) 329 330 def values_sql(self, expression: exp.Values) -> str: 331 """Due to a bug in Snowflake we want to make sure that all columns in a VALUES table alias are unquoted. 332 333 We also want to make sure that after we find matches where we need to unquote a column that we prevent users 334 from adding quotes to the column by using the `identify` argument when generating the SQL. 335 """ 336 alias = expression.args.get("alias") 337 if alias and alias.args.get("columns"): 338 expression = expression.transform( 339 lambda node: exp.Identifier(**{**node.args, "quoted": False}) 340 if isinstance(node, exp.Identifier) 341 and isinstance(node.parent, exp.TableAlias) 342 and node.arg_key == "columns" 343 else node, 344 ) 345 return self.no_identify(lambda: super(self.__class__, self).values_sql(expression)) 346 return super().values_sql(expression) 347 348 def settag_sql(self, expression: exp.SetTag) -> str: 349 action = "UNSET" if expression.args.get("unset") else "SET" 350 return f"{action} TAG {self.expressions(expression)}" 351 352 def select_sql(self, expression: exp.Select) -> str: 353 """Due to a bug in Snowflake we want to make sure that all columns in a VALUES table alias are unquoted and also 354 that all columns in a SELECT are unquoted. We also want to make sure that after we find matches where we need 355 to unquote a column that we prevent users from adding quotes to the column by using the `identify` argument when 356 generating the SQL. 357 358 Note: We make an assumption that any columns referenced in a VALUES expression should be unquoted throughout the 359 expression. This might not be true in a case where the same column name can be sourced from another table that can 360 properly quote but should be true in most cases. 361 """ 362 values_identifiers = set( 363 flatten( 364 (v.args.get("alias") or exp.Alias()).args.get("columns", []) 365 for v in expression.find_all(exp.Values) 366 ) 367 ) 368 if values_identifiers: 369 expression = expression.transform( 370 lambda node: exp.Identifier(**{**node.args, "quoted": False}) 371 if isinstance(node, exp.Identifier) and node in values_identifiers 372 else node, 373 ) 374 return self.no_identify(lambda: super(self.__class__, self).select_sql(expression)) 375 return super().select_sql(expression) 376 377 def describe_sql(self, expression: exp.Describe) -> str: 378 # Default to table if kind is unknown 379 kind_value = expression.args.get("kind") or "TABLE" 380 kind = f" {kind_value}" if kind_value else "" 381 this = f" {self.sql(expression, 'this')}" 382 return f"DESCRIBE{kind}{this}" 383 384 def generatedasidentitycolumnconstraint_sql( 385 self, expression: exp.GeneratedAsIdentityColumnConstraint 386 ) -> str: 387 start = expression.args.get("start") 388 start = f" START {start}" if start else "" 389 increment = expression.args.get("increment") 390 increment = f" INCREMENT {increment}" if increment else "" 391 return f"AUTOINCREMENT{start}{increment}"
Generator interprets the given syntax tree and produces a SQL string as an output.
Arguments:
- time_mapping (dict): the dictionary of custom time mappings in which the key represents a python time format and the output the target time format
- time_trie (trie): a trie of the time_mapping keys
- pretty (bool): if set to True the returned string will be formatted. Default: False.
- quote_start (str): specifies which starting character to use to delimit quotes. Default: '.
- quote_end (str): specifies which ending character to use to delimit quotes. Default: '.
- identifier_start (str): specifies which starting character to use to delimit identifiers. Default: ".
- identifier_end (str): specifies which ending character to use to delimit identifiers. Default: ".
- identify (bool): if set to True all identifiers will be delimited by the corresponding character.
- normalize (bool): if set to True all identifiers will lower cased
- string_escape (str): specifies a string escape character. Default: '.
- identifier_escape (str): specifies an identifier escape character. Default: ".
- pad (int): determines padding in a formatted string. Default: 2.
- indent (int): determines the size of indentation in a formatted string. Default: 4.
- unnest_column_only (bool): if true unnest table aliases are considered only as column aliases
- normalize_functions (str): normalize function names, "upper", "lower", or None Default: "upper"
- alias_post_tablesample (bool): if the table alias comes after tablesample Default: False
- unsupported_level (ErrorLevel): determines the generator's behavior when it encounters unsupported expressions. Default ErrorLevel.WARN.
- null_ordering (str): Indicates the default null ordering method to use if not explicitly set. Options are "nulls_are_small", "nulls_are_large", "nulls_are_last". Default: "nulls_are_small"
- max_unsupported (int): 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 (bool): if the the comma is leading or trailing in select statements 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
330 def values_sql(self, expression: exp.Values) -> str: 331 """Due to a bug in Snowflake we want to make sure that all columns in a VALUES table alias are unquoted. 332 333 We also want to make sure that after we find matches where we need to unquote a column that we prevent users 334 from adding quotes to the column by using the `identify` argument when generating the SQL. 335 """ 336 alias = expression.args.get("alias") 337 if alias and alias.args.get("columns"): 338 expression = expression.transform( 339 lambda node: exp.Identifier(**{**node.args, "quoted": False}) 340 if isinstance(node, exp.Identifier) 341 and isinstance(node.parent, exp.TableAlias) 342 and node.arg_key == "columns" 343 else node, 344 ) 345 return self.no_identify(lambda: super(self.__class__, self).values_sql(expression)) 346 return super().values_sql(expression)
Due to a bug in Snowflake we want to make sure that all columns in a VALUES table alias are unquoted.
We also want to make sure that after we find matches where we need to unquote a column that we prevent users
from adding quotes to the column by using the identify
argument when generating the SQL.
352 def select_sql(self, expression: exp.Select) -> str: 353 """Due to a bug in Snowflake we want to make sure that all columns in a VALUES table alias are unquoted and also 354 that all columns in a SELECT are unquoted. We also want to make sure that after we find matches where we need 355 to unquote a column that we prevent users from adding quotes to the column by using the `identify` argument when 356 generating the SQL. 357 358 Note: We make an assumption that any columns referenced in a VALUES expression should be unquoted throughout the 359 expression. This might not be true in a case where the same column name can be sourced from another table that can 360 properly quote but should be true in most cases. 361 """ 362 values_identifiers = set( 363 flatten( 364 (v.args.get("alias") or exp.Alias()).args.get("columns", []) 365 for v in expression.find_all(exp.Values) 366 ) 367 ) 368 if values_identifiers: 369 expression = expression.transform( 370 lambda node: exp.Identifier(**{**node.args, "quoted": False}) 371 if isinstance(node, exp.Identifier) and node in values_identifiers 372 else node, 373 ) 374 return self.no_identify(lambda: super(self.__class__, self).select_sql(expression)) 375 return super().select_sql(expression)
Due to a bug in Snowflake we want to make sure that all columns in a VALUES table alias are unquoted and also
that all columns in a SELECT are unquoted. We also want to make sure that after we find matches where we need
to unquote a column that we prevent users from adding quotes to the column by using the identify
argument when
generating the SQL.
Note: We make an assumption that any columns referenced in a VALUES expression should be unquoted throughout the expression. This might not be true in a case where the same column name can be sourced from another table that can properly quote but should be true in most cases.
384 def generatedasidentitycolumnconstraint_sql( 385 self, expression: exp.GeneratedAsIdentityColumnConstraint 386 ) -> str: 387 start = expression.args.get("start") 388 start = f" START {start}" if start else "" 389 increment = expression.args.get("increment") 390 increment = f" INCREMENT {increment}" if increment else "" 391 return f"AUTOINCREMENT{start}{increment}"
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
- columndef_sql
- columnconstraint_sql
- autoincrementcolumnconstraint_sql
- compresscolumnconstraint_sql
- notnullcolumnconstraint_sql
- primarykeycolumnconstraint_sql
- uniquecolumnconstraint_sql
- create_sql
- prepend_ctes
- with_sql
- cte_sql
- tablealias_sql
- bitstring_sql
- hexstring_sql
- datatype_sql
- directory_sql
- delete_sql
- drop_sql
- except_sql
- fetch_sql
- filter_sql
- hint_sql
- index_sql
- identifier_sql
- national_sql
- partition_sql
- properties_sql
- root_properties
- properties
- with_properties
- locate_properties
- property_sql
- likeproperty_sql
- fallbackproperty_sql
- journalproperty_sql
- freespaceproperty_sql
- afterjournalproperty_sql
- checksumproperty_sql
- mergeblockratioproperty_sql
- datablocksizeproperty_sql
- blockcompressionproperty_sql
- isolatedloadingproperty_sql
- lockingproperty_sql
- withdataproperty_sql
- insert_sql
- intersect_sql
- introducer_sql
- pseudotype_sql
- returning_sql
- rowformatdelimitedproperty_sql
- table_sql
- tablesample_sql
- pivot_sql
- tuple_sql
- update_sql
- var_sql
- into_sql
- from_sql
- group_sql
- having_sql
- join_sql
- lambda_sql
- lateral_sql
- limit_sql
- offset_sql
- lock_sql
- literal_sql
- loaddata_sql
- null_sql
- boolean_sql
- order_sql
- cluster_sql
- distribute_sql
- sort_sql
- ordered_sql
- matchrecognize_sql
- query_modifiers
- schema_sql
- star_sql
- structkwarg_sql
- parameter_sql
- sessionparameter_sql
- placeholder_sql
- subquery_sql
- qualify_sql
- union_sql
- union_op
- unnest_sql
- where_sql
- window_sql
- partition_by_sql
- window_spec_sql
- withingroup_sql
- between_sql
- bracket_sql
- all_sql
- any_sql
- exists_sql
- case_sql
- constraint_sql
- extract_sql
- trim_sql
- concat_sql
- check_sql
- foreignkey_sql
- primarykey_sql
- unique_sql
- if_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
- cast_sql
- currentdate_sql
- collate_sql
- command_sql
- comment_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
- div_sql
- floatdiv_sql
- overlaps_sql
- distance_sql
- dot_sql
- eq_sql
- escape_sql
- glob_sql
- gt_sql
- gte_sql
- ilike_sql
- is_sql
- like_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