from __future__ import annotations import datetime import re import typing as t from functools import partial, reduce from sqlglot import exp, generator, parser, tokens, transforms from sqlglot.dialects.dialect import ( Dialect, NormalizationStrategy, any_value_to_max_sql, date_delta_sql, datestrtodate_sql, generatedasidentitycolumnconstraint_sql, max_or_greatest, min_or_least, build_date_delta, rename_func, trim_sql, timestrtotime_sql, ) from sqlglot.helper import seq_get from sqlglot.parser import build_coalesce from sqlglot.time import format_time from sqlglot.tokens import TokenType if t.TYPE_CHECKING: from sqlglot._typing import E FULL_FORMAT_TIME_MAPPING = { "weekday": "%A", "dw": "%A", "w": "%A", "month": "%B", "mm": "%B", "m": "%B", } DATE_DELTA_INTERVAL = { "year": "year", "yyyy": "year", "yy": "year", "quarter": "quarter", "qq": "quarter", "q": "quarter", "month": "month", "mm": "month", "m": "month", "week": "week", "ww": "week", "wk": "week", "day": "day", "dd": "day", "d": "day", } DATE_FMT_RE = re.compile("([dD]{1,2})|([mM]{1,2})|([yY]{1,4})|([hH]{1,2})|([sS]{1,2})") # N = Numeric, C=Currency TRANSPILE_SAFE_NUMBER_FMT = {"N", "C"} DEFAULT_START_DATE = datetime.date(1900, 1, 1) BIT_TYPES = {exp.EQ, exp.NEQ, exp.Is, exp.In, exp.Select, exp.Alias} # Unsupported options: # - OPTIMIZE FOR ( @variable_name { UNKNOWN | = } [ , ...n ] ) # - TABLE HINT OPTIONS: parser.OPTIONS_TYPE = { **dict.fromkeys( ( "DISABLE_OPTIMIZED_PLAN_FORCING", "FAST", "IGNORE_NONCLUSTERED_COLUMNSTORE_INDEX", "LABEL", "MAXDOP", "MAXRECURSION", "MAX_GRANT_PERCENT", "MIN_GRANT_PERCENT", "NO_PERFORMANCE_SPOOL", "QUERYTRACEON", "RECOMPILE", ), tuple(), ), "CONCAT": ("UNION",), "DISABLE": ("EXTERNALPUSHDOWN", "SCALEOUTEXECUTION"), "EXPAND": ("VIEWS",), "FORCE": ("EXTERNALPUSHDOWN", "ORDER", "SCALEOUTEXECUTION"), "HASH": ("GROUP", "JOIN", "UNION"), "KEEP": ("PLAN",), "KEEPFIXED": ("PLAN",), "LOOP": ("JOIN",), "MERGE": ("JOIN", "UNION"), "OPTIMIZE": (("FOR", "UNKNOWN"),), "ORDER": ("GROUP",), "PARAMETERIZATION": ("FORCED", "SIMPLE"), "ROBUST": ("PLAN",), "USE": ("PLAN",), } OPTIONS_THAT_REQUIRE_EQUAL = ("MAX_GRANT_PERCENT", "MIN_GRANT_PERCENT", "LABEL") def _build_formatted_time( exp_class: t.Type[E], full_format_mapping: t.Optional[bool] = None ) -> t.Callable[[t.List], E]: def _builder(args: t.List) -> E: assert len(args) == 2 return exp_class( this=exp.cast(args[1], exp.DataType.Type.DATETIME), format=exp.Literal.string( format_time( args[0].name.lower(), ( {**TSQL.TIME_MAPPING, **FULL_FORMAT_TIME_MAPPING} if full_format_mapping else TSQL.TIME_MAPPING ), ) ), ) return _builder def _build_format(args: t.List) -> exp.NumberToStr | exp.TimeToStr: this = seq_get(args, 0) fmt = seq_get(args, 1) culture = seq_get(args, 2) number_fmt = fmt and (fmt.name in TRANSPILE_SAFE_NUMBER_FMT or not DATE_FMT_RE.search(fmt.name)) if number_fmt: return exp.NumberToStr(this=this, format=fmt, culture=culture) if fmt: fmt = exp.Literal.string( format_time(fmt.name, TSQL.FORMAT_TIME_MAPPING) if len(fmt.name) == 1 else format_time(fmt.name, TSQL.TIME_MAPPING) ) return exp.TimeToStr(this=this, format=fmt, culture=culture) def _build_eomonth(args: t.List) -> exp.LastDay: date = exp.TsOrDsToDate(this=seq_get(args, 0)) month_lag = seq_get(args, 1) if month_lag is None: this: exp.Expression = date else: unit = DATE_DELTA_INTERVAL.get("month") this = exp.DateAdd(this=date, expression=month_lag, unit=unit and exp.var(unit)) return exp.LastDay(this=this) def _build_hashbytes(args: t.List) -> exp.Expression: kind, data = args kind = kind.name.upper() if kind.is_string else "" if kind == "MD5": args.pop(0) return exp.MD5(this=data) if kind in ("SHA", "SHA1"): args.pop(0) return exp.SHA(this=data) if kind == "SHA2_256": return exp.SHA2(this=data, length=exp.Literal.number(256)) if kind == "SHA2_512": return exp.SHA2(this=data, length=exp.Literal.number(512)) return exp.func("HASHBYTES", *args) DATEPART_ONLY_FORMATS = {"DW", "WK", "HOUR", "QUARTER"} def _format_sql(self: TSQL.Generator, expression: exp.NumberToStr | exp.TimeToStr) -> str: fmt = expression.args["format"] if not isinstance(expression, exp.NumberToStr): if fmt.is_string: mapped_fmt = format_time(fmt.name, TSQL.INVERSE_TIME_MAPPING) name = (mapped_fmt or "").upper() if name in DATEPART_ONLY_FORMATS: return self.func("DATEPART", name, expression.this) fmt_sql = self.sql(exp.Literal.string(mapped_fmt)) else: fmt_sql = self.format_time(expression) or self.sql(fmt) else: fmt_sql = self.sql(fmt) return self.func("FORMAT", expression.this, fmt_sql, expression.args.get("culture")) def _string_agg_sql(self: TSQL.Generator, expression: exp.GroupConcat) -> str: this = expression.this distinct = expression.find(exp.Distinct) if distinct: # exp.Distinct can appear below an exp.Order or an exp.GroupConcat expression self.unsupported("T-SQL STRING_AGG doesn't support DISTINCT.") this = distinct.pop().expressions[0] order = "" if isinstance(expression.this, exp.Order): if expression.this.this: this = expression.this.this.pop() # Order has a leading space order = f" WITHIN GROUP ({self.sql(expression.this)[1:]})" separator = expression.args.get("separator") or exp.Literal.string(",") return f"STRING_AGG({self.format_args(this, separator)}){order}" def _build_date_delta( exp_class: t.Type[E], unit_mapping: t.Optional[t.Dict[str, str]] = None ) -> t.Callable[[t.List], E]: def _builder(args: t.List) -> E: unit = seq_get(args, 0) if unit and unit_mapping: unit = exp.var(unit_mapping.get(unit.name.lower(), unit.name)) start_date = seq_get(args, 1) if start_date and start_date.is_number: # Numeric types are valid DATETIME values if start_date.is_int: adds = DEFAULT_START_DATE + datetime.timedelta(days=int(start_date.this)) start_date = exp.Literal.string(adds.strftime("%F")) else: # We currently don't handle float values, i.e. they're not converted to equivalent DATETIMEs. # This is not a problem when generating T-SQL code, it is when transpiling to other dialects. return exp_class(this=seq_get(args, 2), expression=start_date, unit=unit) return exp_class( this=exp.TimeStrToTime(this=seq_get(args, 2)), expression=exp.TimeStrToTime(this=start_date), unit=unit, ) return _builder def qualify_derived_table_outputs(expression: exp.Expression) -> exp.Expression: """Ensures all (unnamed) output columns are aliased for CTEs and Subqueries.""" alias = expression.args.get("alias") if ( isinstance(expression, (exp.CTE, exp.Subquery)) and isinstance(alias, exp.TableAlias) and not alias.columns ): from sqlglot.optimizer.qualify_columns import qualify_outputs # We keep track of the unaliased column projection indexes instead of the expressions # themselves, because the latter are going to be replaced by new nodes when the aliases # are added and hence we won't be able to reach these newly added Alias parents query = expression.this unaliased_column_indexes = ( i for i, c in enumerate(query.selects) if isinstance(c, exp.Column) and not c.alias ) qualify_outputs(query) # Preserve the quoting information of columns for newly added Alias nodes query_selects = query.selects for select_index in unaliased_column_indexes: alias = query_selects[select_index] column = alias.this if isinstance(column.this, exp.Identifier): alias.args["alias"].set("quoted", column.this.quoted) return expression # https://learn.microsoft.com/en-us/sql/t-sql/functions/datetimefromparts-transact-sql?view=sql-server-ver16#syntax def _build_datetimefromparts(args: t.List) -> exp.TimestampFromParts: return exp.TimestampFromParts( year=seq_get(args, 0), month=seq_get(args, 1), day=seq_get(args, 2), hour=seq_get(args, 3), min=seq_get(args, 4), sec=seq_get(args, 5), milli=seq_get(args, 6), ) # https://learn.microsoft.com/en-us/sql/t-sql/functions/timefromparts-transact-sql?view=sql-server-ver16#syntax def _build_timefromparts(args: t.List) -> exp.TimeFromParts: return exp.TimeFromParts( hour=seq_get(args, 0), min=seq_get(args, 1), sec=seq_get(args, 2), fractions=seq_get(args, 3), precision=seq_get(args, 4), ) def _build_with_arg_as_text( klass: t.Type[exp.Expression], ) -> t.Callable[[t.List[exp.Expression]], exp.Expression]: def _parse(args: t.List[exp.Expression]) -> exp.Expression: this = seq_get(args, 0) if this and not this.is_string: this = exp.cast(this, exp.DataType.Type.TEXT) expression = seq_get(args, 1) kwargs = {"this": this} if expression: kwargs["expression"] = expression return klass(**kwargs) return _parse # https://learn.microsoft.com/en-us/sql/t-sql/functions/parsename-transact-sql?view=sql-server-ver16 def _build_parsename(args: t.List) -> exp.SplitPart | exp.Anonymous: # PARSENAME(...) will be stored into exp.SplitPart if: # - All args are literals # - The part index (2nd arg) is <= 4 (max valid value, otherwise TSQL returns NULL) if len(args) == 2 and all(isinstance(arg, exp.Literal) for arg in args): this = args[0] part_index = args[1] split_count = len(this.name.split(".")) if split_count <= 4: return exp.SplitPart( this=this, delimiter=exp.Literal.string("."), part_index=exp.Literal.number(split_count + 1 - part_index.to_py()), ) return exp.Anonymous(this="PARSENAME", expressions=args) def _build_json_query(args: t.List, dialect: Dialect) -> exp.JSONExtract: if len(args) == 1: # The default value for path is '$'. As a result, if you don't provide a # value for path, JSON_QUERY returns the input expression. args.append(exp.Literal.string("$")) return parser.build_extract_json_with_path(exp.JSONExtract)(args, dialect) def _json_extract_sql( self: TSQL.Generator, expression: exp.JSONExtract | exp.JSONExtractScalar ) -> str: json_query = self.func("JSON_QUERY", expression.this, expression.expression) json_value = self.func("JSON_VALUE", expression.this, expression.expression) return self.func("ISNULL", json_query, json_value) def _timestrtotime_sql(self: TSQL.Generator, expression: exp.TimeStrToTime): sql = timestrtotime_sql(self, expression) if expression.args.get("zone"): # If there is a timezone, produce an expression like: # CAST('2020-01-01 12:13:14-08:00' AS DATETIMEOFFSET) AT TIME ZONE 'UTC' # If you dont have AT TIME ZONE 'UTC', wrapping that expression in another cast back to DATETIME2 just drops the timezone information return self.sql(exp.AtTimeZone(this=sql, zone=exp.Literal.string("UTC"))) return sql class TSQL(Dialect): SUPPORTS_SEMI_ANTI_JOIN = False LOG_BASE_FIRST = False TYPED_DIVISION = True CONCAT_COALESCE = True NORMALIZATION_STRATEGY = NormalizationStrategy.CASE_INSENSITIVE TIME_FORMAT = "'yyyy-mm-dd hh:mm:ss'" TIME_MAPPING = { "year": "%Y", "dayofyear": "%j", "day": "%d", "dy": "%d", "y": "%Y", "week": "%W", "ww": "%W", "wk": "%W", "hour": "%h", "hh": "%I", "minute": "%M", "mi": "%M", "n": "%M", "second": "%S", "ss": "%S", "s": "%-S", "millisecond": "%f", "ms": "%f", "weekday": "%w", "dw": "%w", "month": "%m", "mm": "%M", "m": "%-M", "Y": "%Y", "YYYY": "%Y", "YY": "%y", "MMMM": "%B", "MMM": "%b", "MM": "%m", "M": "%-m", "dddd": "%A", "dd": "%d", "d": "%-d", "HH": "%H", "H": "%-H", "h": "%-I", "ffffff": "%f", "yyyy": "%Y", "yy": "%y", } CONVERT_FORMAT_MAPPING = { "0": "%b %d %Y %-I:%M%p", "1": "%m/%d/%y", "2": "%y.%m.%d", "3": "%d/%m/%y", "4": "%d.%m.%y", "5": "%d-%m-%y", "6": "%d %b %y", "7": "%b %d, %y", "8": "%H:%M:%S", "9": "%b %d %Y %-I:%M:%S:%f%p", "10": "mm-dd-yy", "11": "yy/mm/dd", "12": "yymmdd", "13": "%d %b %Y %H:%M:ss:%f", "14": "%H:%M:%S:%f", "20": "%Y-%m-%d %H:%M:%S", "21": "%Y-%m-%d %H:%M:%S.%f", "22": "%m/%d/%y %-I:%M:%S %p", "23": "%Y-%m-%d", "24": "%H:%M:%S", "25": "%Y-%m-%d %H:%M:%S.%f", "100": "%b %d %Y %-I:%M%p", "101": "%m/%d/%Y", "102": "%Y.%m.%d", "103": "%d/%m/%Y", "104": "%d.%m.%Y", "105": "%d-%m-%Y", "106": "%d %b %Y", "107": "%b %d, %Y", "108": "%H:%M:%S", "109": "%b %d %Y %-I:%M:%S:%f%p", "110": "%m-%d-%Y", "111": "%Y/%m/%d", "112": "%Y%m%d", "113": "%d %b %Y %H:%M:%S:%f", "114": "%H:%M:%S:%f", "120": "%Y-%m-%d %H:%M:%S", "121": "%Y-%m-%d %H:%M:%S.%f", } FORMAT_TIME_MAPPING = { "y": "%B %Y", "d": "%m/%d/%Y", "H": "%-H", "h": "%-I", "s": "%Y-%m-%d %H:%M:%S", "D": "%A,%B,%Y", "f": "%A,%B,%Y %-I:%M %p", "F": "%A,%B,%Y %-I:%M:%S %p", "g": "%m/%d/%Y %-I:%M %p", "G": "%m/%d/%Y %-I:%M:%S %p", "M": "%B %-d", "m": "%B %-d", "O": "%Y-%m-%dT%H:%M:%S", "u": "%Y-%M-%D %H:%M:%S%z", "U": "%A, %B %D, %Y %H:%M:%S%z", "T": "%-I:%M:%S %p", "t": "%-I:%M", "Y": "%a %Y", } class Tokenizer(tokens.Tokenizer): IDENTIFIERS = [("[", "]"), '"'] QUOTES = ["'", '"'] HEX_STRINGS = [("0x", ""), ("0X", "")] VAR_SINGLE_TOKENS = {"@", "$", "#"} KEYWORDS = { **tokens.Tokenizer.KEYWORDS, "CLUSTERED INDEX": TokenType.INDEX, "DATETIME2": TokenType.DATETIME, "DATETIMEOFFSET": TokenType.TIMESTAMPTZ, "DECLARE": TokenType.DECLARE, "EXEC": TokenType.COMMAND, "FOR SYSTEM_TIME": TokenType.TIMESTAMP_SNAPSHOT, "IMAGE": TokenType.IMAGE, "MONEY": TokenType.MONEY, "NONCLUSTERED INDEX": TokenType.INDEX, "NTEXT": TokenType.TEXT, "OPTION": TokenType.OPTION, "OUTPUT": TokenType.RETURNING, "PRINT": TokenType.COMMAND, "PROC": TokenType.PROCEDURE, "REAL": TokenType.FLOAT, "ROWVERSION": TokenType.ROWVERSION, "SMALLDATETIME": TokenType.DATETIME, "SMALLMONEY": TokenType.SMALLMONEY, "SQL_VARIANT": TokenType.VARIANT, "SYSTEM_USER": TokenType.CURRENT_USER, "TOP": TokenType.TOP, "TIMESTAMP": TokenType.ROWVERSION, "TINYINT": TokenType.UTINYINT, "UNIQUEIDENTIFIER": TokenType.UNIQUEIDENTIFIER, "UPDATE STATISTICS": TokenType.COMMAND, "XML": TokenType.XML, } KEYWORDS.pop("/*+") COMMANDS = {*tokens.Tokenizer.COMMANDS, TokenType.END} class Parser(parser.Parser): SET_REQUIRES_ASSIGNMENT_DELIMITER = False LOG_DEFAULTS_TO_LN = True ALTER_TABLE_ADD_REQUIRED_FOR_EACH_COLUMN = False STRING_ALIASES = True NO_PAREN_IF_COMMANDS = False QUERY_MODIFIER_PARSERS = { **parser.Parser.QUERY_MODIFIER_PARSERS, TokenType.OPTION: lambda self: ("options", self._parse_options()), } FUNCTIONS = { **parser.Parser.FUNCTIONS, "CHARINDEX": lambda args: exp.StrPosition( this=seq_get(args, 1), substr=seq_get(args, 0), position=seq_get(args, 2), ), "COUNT": lambda args: exp.Count( this=seq_get(args, 0), expressions=args[1:], big_int=False ), "COUNT_BIG": lambda args: exp.Count( this=seq_get(args, 0), expressions=args[1:], big_int=True ), "DATEADD": build_date_delta(exp.DateAdd, unit_mapping=DATE_DELTA_INTERVAL), "DATEDIFF": _build_date_delta(exp.DateDiff, unit_mapping=DATE_DELTA_INTERVAL), "DATENAME": _build_formatted_time(exp.TimeToStr, full_format_mapping=True), "DATEPART": _build_formatted_time(exp.TimeToStr), "DATETIMEFROMPARTS": _build_datetimefromparts, "EOMONTH": _build_eomonth, "FORMAT": _build_format, "GETDATE": exp.CurrentTimestamp.from_arg_list, "HASHBYTES": _build_hashbytes, "ISNULL": build_coalesce, "JSON_QUERY": _build_json_query, "JSON_VALUE": parser.build_extract_json_with_path(exp.JSONExtractScalar), "LEN": _build_with_arg_as_text(exp.Length), "LEFT": _build_with_arg_as_text(exp.Left), "RIGHT": _build_with_arg_as_text(exp.Right), "PARSENAME": _build_parsename, "REPLICATE": exp.Repeat.from_arg_list, "SQUARE": lambda args: exp.Pow(this=seq_get(args, 0), expression=exp.Literal.number(2)), "SYSDATETIME": exp.CurrentTimestamp.from_arg_list, "SUSER_NAME": exp.CurrentUser.from_arg_list, "SUSER_SNAME": exp.CurrentUser.from_arg_list, "SYSTEM_USER": exp.CurrentUser.from_arg_list, "TIMEFROMPARTS": _build_timefromparts, } JOIN_HINTS = {"LOOP", "HASH", "MERGE", "REMOTE"} PROCEDURE_OPTIONS = dict.fromkeys( ("ENCRYPTION", "RECOMPILE", "SCHEMABINDING", "NATIVE_COMPILATION", "EXECUTE"), tuple() ) RETURNS_TABLE_TOKENS = parser.Parser.ID_VAR_TOKENS - { TokenType.TABLE, *parser.Parser.TYPE_TOKENS, } STATEMENT_PARSERS = { **parser.Parser.STATEMENT_PARSERS, TokenType.DECLARE: lambda self: self._parse_declare(), } RANGE_PARSERS = { **parser.Parser.RANGE_PARSERS, TokenType.DCOLON: lambda self, this: self.expression( exp.ScopeResolution, this=this, expression=self._parse_function() or self._parse_var(any_token=True), ), } NO_PAREN_FUNCTION_PARSERS = { **parser.Parser.NO_PAREN_FUNCTION_PARSERS, "NEXT": lambda self: self._parse_next_value_for(), } # The DCOLON (::) operator serves as a scope resolution (exp.ScopeResolution) operator in T-SQL COLUMN_OPERATORS = { **parser.Parser.COLUMN_OPERATORS, TokenType.DCOLON: lambda self, this, to: self.expression(exp.Cast, this=this, to=to) if isinstance(to, exp.DataType) and to.this != exp.DataType.Type.USERDEFINED else self.expression(exp.ScopeResolution, this=this, expression=to), } def _parse_dcolon(self) -> t.Optional[exp.Expression]: # We want to use _parse_types() if the first token after :: is a known type, # otherwise we could parse something like x::varchar(max) into a function if self._match_set(self.TYPE_TOKENS, advance=False): return self._parse_types() return self._parse_function() or self._parse_types() def _parse_options(self) -> t.Optional[t.List[exp.Expression]]: if not self._match(TokenType.OPTION): return None def _parse_option() -> t.Optional[exp.Expression]: option = self._parse_var_from_options(OPTIONS) if not option: return None self._match(TokenType.EQ) return self.expression( exp.QueryOption, this=option, expression=self._parse_primary_or_var() ) return self._parse_wrapped_csv(_parse_option) def _parse_projections(self) -> t.List[exp.Expression]: """ T-SQL supports the syntax alias = expression in the SELECT's projection list, so we transform all parsed Selects to convert their EQ projections into Aliases. See: https://learn.microsoft.com/en-us/sql/t-sql/queries/select-clause-transact-sql?view=sql-server-ver16#syntax """ return [ ( exp.alias_(projection.expression, projection.this.this, copy=False) if isinstance(projection, exp.EQ) and isinstance(projection.this, exp.Column) else projection ) for projection in super()._parse_projections() ] def _parse_commit_or_rollback(self) -> exp.Commit | exp.Rollback: """Applies to SQL Server and Azure SQL Database COMMIT [ { TRAN | TRANSACTION } [ transaction_name | @tran_name_variable ] ] [ WITH ( DELAYED_DURABILITY = { OFF | ON } ) ] ROLLBACK { TRAN | TRANSACTION } [ transaction_name | @tran_name_variable | savepoint_name | @savepoint_variable ] """ rollback = self._prev.token_type == TokenType.ROLLBACK self._match_texts(("TRAN", "TRANSACTION")) this = self._parse_id_var() if rollback: return self.expression(exp.Rollback, this=this) durability = None if self._match_pair(TokenType.WITH, TokenType.L_PAREN): self._match_text_seq("DELAYED_DURABILITY") self._match(TokenType.EQ) if self._match_text_seq("OFF"): durability = False else: self._match(TokenType.ON) durability = True self._match_r_paren() return self.expression(exp.Commit, this=this, durability=durability) def _parse_transaction(self) -> exp.Transaction | exp.Command: """Applies to SQL Server and Azure SQL Database BEGIN { TRAN | TRANSACTION } [ { transaction_name | @tran_name_variable } [ WITH MARK [ 'description' ] ] ] """ if self._match_texts(("TRAN", "TRANSACTION")): transaction = self.expression(exp.Transaction, this=self._parse_id_var()) if self._match_text_seq("WITH", "MARK"): transaction.set("mark", self._parse_string()) return transaction return self._parse_as_command(self._prev) def _parse_returns(self) -> exp.ReturnsProperty: table = self._parse_id_var(any_token=False, tokens=self.RETURNS_TABLE_TOKENS) returns = super()._parse_returns() returns.set("table", table) return returns def _parse_convert( self, strict: bool, safe: t.Optional[bool] = None ) -> t.Optional[exp.Expression]: this = self._parse_types() self._match(TokenType.COMMA) args = [this, *self._parse_csv(self._parse_assignment)] convert = exp.Convert.from_arg_list(args) convert.set("safe", safe) convert.set("strict", strict) return convert def _parse_user_defined_function( self, kind: t.Optional[TokenType] = None ) -> t.Optional[exp.Expression]: this = super()._parse_user_defined_function(kind=kind) if ( kind == TokenType.FUNCTION or isinstance(this, exp.UserDefinedFunction) or self._match(TokenType.ALIAS, advance=False) ): return this if not self._match(TokenType.WITH, advance=False): expressions = self._parse_csv(self._parse_function_parameter) else: expressions = None return self.expression(exp.UserDefinedFunction, this=this, expressions=expressions) def _parse_id_var( self, any_token: bool = True, tokens: t.Optional[t.Collection[TokenType]] = None, ) -> t.Optional[exp.Expression]: is_temporary = self._match(TokenType.HASH) is_global = is_temporary and self._match(TokenType.HASH) this = super()._parse_id_var(any_token=any_token, tokens=tokens) if this: if is_global: this.set("global", True) elif is_temporary: this.set("temporary", True) return this def _parse_create(self) -> exp.Create | exp.Command: create = super()._parse_create() if isinstance(create, exp.Create): table = create.this.this if isinstance(create.this, exp.Schema) else create.this if isinstance(table, exp.Table) and table.this.args.get("temporary"): if not create.args.get("properties"): create.set("properties", exp.Properties(expressions=[])) create.args["properties"].append("expressions", exp.TemporaryProperty()) return create def _parse_if(self) -> t.Optional[exp.Expression]: index = self._index if self._match_text_seq("OBJECT_ID"): self._parse_wrapped_csv(self._parse_string) if self._match_text_seq("IS", "NOT", "NULL") and self._match(TokenType.DROP): return self._parse_drop(exists=True) self._retreat(index) return super()._parse_if() def _parse_unique(self) -> exp.UniqueColumnConstraint: if self._match_texts(("CLUSTERED", "NONCLUSTERED")): this = self.CONSTRAINT_PARSERS[self._prev.text.upper()](self) else: this = self._parse_schema(self._parse_id_var(any_token=False)) return self.expression(exp.UniqueColumnConstraint, this=this) def _parse_partition(self) -> t.Optional[exp.Partition]: if not self._match_text_seq("WITH", "(", "PARTITIONS"): return None def parse_range(): low = self._parse_bitwise() high = self._parse_bitwise() if self._match_text_seq("TO") else None return ( self.expression(exp.PartitionRange, this=low, expression=high) if high else low ) partition = self.expression( exp.Partition, expressions=self._parse_wrapped_csv(parse_range) ) self._match_r_paren() return partition def _parse_declare(self) -> exp.Declare | exp.Command: index = self._index expressions = self._try_parse(partial(self._parse_csv, self._parse_declareitem)) if not expressions or self._curr: self._retreat(index) return self._parse_as_command(self._prev) return self.expression(exp.Declare, expressions=expressions) def _parse_declareitem(self) -> t.Optional[exp.DeclareItem]: var = self._parse_id_var() if not var: return None value = None self._match(TokenType.ALIAS) if self._match(TokenType.TABLE): data_type = self._parse_schema() else: data_type = self._parse_types() if self._match(TokenType.EQ): value = self._parse_bitwise() return self.expression(exp.DeclareItem, this=var, kind=data_type, default=value) class Generator(generator.Generator): LIMIT_IS_TOP = True QUERY_HINTS = False RETURNING_END = False NVL2_SUPPORTED = False ALTER_TABLE_INCLUDE_COLUMN_KEYWORD = False LIMIT_FETCH = "FETCH" COMPUTED_COLUMN_WITH_TYPE = False CTE_RECURSIVE_KEYWORD_REQUIRED = False ENSURE_BOOLS = True NULL_ORDERING_SUPPORTED = None SUPPORTS_SINGLE_ARG_CONCAT = False TABLESAMPLE_SEED_KEYWORD = "REPEATABLE" SUPPORTS_SELECT_INTO = True JSON_PATH_BRACKETED_KEY_SUPPORTED = False SUPPORTS_TO_NUMBER = False SET_OP_MODIFIERS = False COPY_PARAMS_EQ_REQUIRED = True PARSE_JSON_NAME = None EXCEPT_INTERSECT_SUPPORT_ALL_CLAUSE = False EXPRESSIONS_WITHOUT_NESTED_CTES = { exp.Create, exp.Delete, exp.Insert, exp.Intersect, exp.Except, exp.Merge, exp.Select, exp.Subquery, exp.Union, exp.Update, } SUPPORTED_JSON_PATH_PARTS = { exp.JSONPathKey, exp.JSONPathRoot, exp.JSONPathSubscript, } TYPE_MAPPING = { **generator.Generator.TYPE_MAPPING, exp.DataType.Type.BOOLEAN: "BIT", exp.DataType.Type.DECIMAL: "NUMERIC", exp.DataType.Type.DATETIME: "DATETIME2", exp.DataType.Type.DOUBLE: "FLOAT", exp.DataType.Type.INT: "INTEGER", exp.DataType.Type.ROWVERSION: "ROWVERSION", exp.DataType.Type.TEXT: "VARCHAR(MAX)", exp.DataType.Type.TIMESTAMP: "DATETIME2", exp.DataType.Type.TIMESTAMPTZ: "DATETIMEOFFSET", exp.DataType.Type.UTINYINT: "TINYINT", exp.DataType.Type.VARIANT: "SQL_VARIANT", } TYPE_MAPPING.pop(exp.DataType.Type.NCHAR) TYPE_MAPPING.pop(exp.DataType.Type.NVARCHAR) TRANSFORMS = { **generator.Generator.TRANSFORMS, exp.AnyValue: any_value_to_max_sql, exp.ArrayToString: rename_func("STRING_AGG"), exp.AutoIncrementColumnConstraint: lambda *_: "IDENTITY", exp.DateAdd: date_delta_sql("DATEADD"), exp.DateDiff: date_delta_sql("DATEDIFF"), exp.CTE: transforms.preprocess([qualify_derived_table_outputs]), exp.CurrentDate: rename_func("GETDATE"), exp.CurrentTimestamp: rename_func("GETDATE"), exp.DateStrToDate: datestrtodate_sql, exp.Extract: rename_func("DATEPART"), exp.GeneratedAsIdentityColumnConstraint: generatedasidentitycolumnconstraint_sql, exp.GroupConcat: _string_agg_sql, exp.If: rename_func("IIF"), exp.JSONExtract: _json_extract_sql, exp.JSONExtractScalar: _json_extract_sql, exp.LastDay: lambda self, e: self.func("EOMONTH", e.this), exp.Ln: rename_func("LOG"), exp.Max: max_or_greatest, exp.MD5: lambda self, e: self.func("HASHBYTES", exp.Literal.string("MD5"), e.this), exp.Min: min_or_least, exp.NumberToStr: _format_sql, exp.Repeat: rename_func("REPLICATE"), exp.Select: transforms.preprocess( [ transforms.eliminate_distinct_on, transforms.eliminate_semi_and_anti_joins, transforms.eliminate_qualify, transforms.unnest_generate_date_array_using_recursive_cte, ] ), exp.Stddev: rename_func("STDEV"), exp.StrPosition: lambda self, e: self.func( "CHARINDEX", e.args.get("substr"), e.this, e.args.get("position") ), exp.Subquery: transforms.preprocess([qualify_derived_table_outputs]), exp.SHA: lambda self, e: self.func("HASHBYTES", exp.Literal.string("SHA1"), e.this), exp.SHA2: lambda self, e: self.func( "HASHBYTES", exp.Literal.string(f"SHA2_{e.args.get('length', 256)}"), e.this ), exp.TemporaryProperty: lambda self, e: "", exp.TimeStrToTime: _timestrtotime_sql, exp.TimeToStr: _format_sql, exp.Trim: trim_sql, exp.TsOrDsAdd: date_delta_sql("DATEADD", cast=True), exp.TsOrDsDiff: date_delta_sql("DATEDIFF"), } TRANSFORMS.pop(exp.ReturnsProperty) PROPERTIES_LOCATION = { **generator.Generator.PROPERTIES_LOCATION, exp.VolatileProperty: exp.Properties.Location.UNSUPPORTED, } def scope_resolution(self, rhs: str, scope_name: str) -> str: return f"{scope_name}::{rhs}" def select_sql(self, expression: exp.Select) -> str: if expression.args.get("offset"): if not expression.args.get("order"): # ORDER BY is required in order to use OFFSET in a query, so we use # a noop order by, since we don't really care about the order. # See: https://www.microsoftpressstore.com/articles/article.aspx?p=2314819 expression.order_by(exp.select(exp.null()).subquery(), copy=False) limit = expression.args.get("limit") if isinstance(limit, exp.Limit): # TOP and OFFSET can't be combined, we need use FETCH instead of TOP # we replace here because otherwise TOP would be generated in select_sql limit.replace(exp.Fetch(direction="FIRST", count=limit.expression)) return super().select_sql(expression) def convert_sql(self, expression: exp.Convert) -> str: name = "TRY_CONVERT" if expression.args.get("safe") else "CONVERT" return self.func( name, expression.this, expression.expression, expression.args.get("style") ) def queryoption_sql(self, expression: exp.QueryOption) -> str: option = self.sql(expression, "this") value = self.sql(expression, "expression") if value: optional_equal_sign = "= " if option in OPTIONS_THAT_REQUIRE_EQUAL else "" return f"{option} {optional_equal_sign}{value}" return option def lateral_op(self, expression: exp.Lateral) -> str: cross_apply = expression.args.get("cross_apply") if cross_apply is True: return "CROSS APPLY" if cross_apply is False: return "OUTER APPLY" # TODO: perhaps we can check if the parent is a Join and transpile it appropriately self.unsupported("LATERAL clause is not supported.") return "LATERAL" def splitpart_sql(self: TSQL.Generator, expression: exp.SplitPart) -> str: this = expression.this split_count = len(this.name.split(".")) delimiter = expression.args.get("delimiter") part_index = expression.args.get("part_index") if ( not all(isinstance(arg, exp.Literal) for arg in (this, delimiter, part_index)) or (delimiter and delimiter.name != ".") or not part_index or split_count > 4 ): self.unsupported( "SPLIT_PART can be transpiled to PARSENAME only for '.' delimiter and literal values" ) return "" return self.func( "PARSENAME", this, exp.Literal.number(split_count + 1 - part_index.to_py()) ) def timefromparts_sql(self, expression: exp.TimeFromParts) -> str: nano = expression.args.get("nano") if nano is not None: nano.pop() self.unsupported("Specifying nanoseconds is not supported in TIMEFROMPARTS.") if expression.args.get("fractions") is None: expression.set("fractions", exp.Literal.number(0)) if expression.args.get("precision") is None: expression.set("precision", exp.Literal.number(0)) return rename_func("TIMEFROMPARTS")(self, expression) def timestampfromparts_sql(self, expression: exp.TimestampFromParts) -> str: zone = expression.args.get("zone") if zone is not None: zone.pop() self.unsupported("Time zone is not supported in DATETIMEFROMPARTS.") nano = expression.args.get("nano") if nano is not None: nano.pop() self.unsupported("Specifying nanoseconds is not supported in DATETIMEFROMPARTS.") if expression.args.get("milli") is None: expression.set("milli", exp.Literal.number(0)) return rename_func("DATETIMEFROMPARTS")(self, expression) def setitem_sql(self, expression: exp.SetItem) -> str: this = expression.this if isinstance(this, exp.EQ) and not isinstance(this.left, exp.Parameter): # T-SQL does not use '=' in SET command, except when the LHS is a variable. return f"{self.sql(this.left)} {self.sql(this.right)}" return super().setitem_sql(expression) def boolean_sql(self, expression: exp.Boolean) -> str: if type(expression.parent) in BIT_TYPES or isinstance( expression.find_ancestor(exp.Values, exp.Select), exp.Values ): return "1" if expression.this else "0" return "(1 = 1)" if expression.this else "(1 = 0)" def is_sql(self, expression: exp.Is) -> str: if isinstance(expression.expression, exp.Boolean): return self.binary(expression, "=") return self.binary(expression, "IS") def createable_sql(self, expression: exp.Create, locations: t.DefaultDict) -> str: sql = self.sql(expression, "this") properties = expression.args.get("properties") if sql[:1] != "#" and any( isinstance(prop, exp.TemporaryProperty) for prop in (properties.expressions if properties else []) ): sql = f"[#{sql[1:]}" if sql.startswith("[") else f"#{sql}" return sql def create_sql(self, expression: exp.Create) -> str: kind = expression.kind exists = expression.args.pop("exists", None) like_property = expression.find(exp.LikeProperty) if like_property: ctas_expression = like_property.this else: ctas_expression = expression.expression if kind == "VIEW": expression.this.set("catalog", None) with_ = expression.args.get("with") if ctas_expression and with_: # We've already preprocessed the Create expression to bubble up any nested CTEs, # but CREATE VIEW actually requires the WITH clause to come after it so we need # to amend the AST by moving the CTEs to the CREATE VIEW statement's query. ctas_expression.set("with", with_.pop()) sql = super().create_sql(expression) table = expression.find(exp.Table) # Convert CTAS statement to SELECT .. INTO .. if kind == "TABLE" and ctas_expression: if isinstance(ctas_expression, exp.UNWRAPPED_QUERIES): ctas_expression = ctas_expression.subquery() select_into = exp.select("*").from_(exp.alias_(ctas_expression, "temp", table=True)) select_into.set("into", exp.Into(this=table)) if like_property: select_into.limit(0, copy=False) sql = self.sql(select_into) if exists: identifier = self.sql(exp.Literal.string(exp.table_name(table) if table else "")) sql_with_ctes = self.prepend_ctes(expression, sql) sql_literal = self.sql(exp.Literal.string(sql_with_ctes)) if kind == "SCHEMA": return f"""IF NOT EXISTS (SELECT * FROM information_schema.schemata WHERE schema_name = {identifier}) EXEC({sql_literal})""" elif kind == "TABLE": assert table where = exp.and_( exp.column("table_name").eq(table.name), exp.column("table_schema").eq(table.db) if table.db else None, exp.column("table_catalog").eq(table.catalog) if table.catalog else None, ) return f"""IF NOT EXISTS (SELECT * FROM information_schema.tables WHERE {where}) EXEC({sql_literal})""" elif kind == "INDEX": index = self.sql(exp.Literal.string(expression.this.text("this"))) return f"""IF NOT EXISTS (SELECT * FROM sys.indexes WHERE object_id = object_id({identifier}) AND name = {index}) EXEC({sql_literal})""" elif expression.args.get("replace"): sql = sql.replace("CREATE OR REPLACE ", "CREATE OR ALTER ", 1) return self.prepend_ctes(expression, sql) def count_sql(self, expression: exp.Count) -> str: func_name = "COUNT_BIG" if expression.args.get("big_int") else "COUNT" return rename_func(func_name)(self, expression) def offset_sql(self, expression: exp.Offset) -> str: return f"{super().offset_sql(expression)} ROWS" def version_sql(self, expression: exp.Version) -> str: name = "SYSTEM_TIME" if expression.name == "TIMESTAMP" else expression.name this = f"FOR {name}" expr = expression.expression kind = expression.text("kind") if kind in ("FROM", "BETWEEN"): args = expr.expressions sep = "TO" if kind == "FROM" else "AND" expr_sql = f"{self.sql(seq_get(args, 0))} {sep} {self.sql(seq_get(args, 1))}" else: expr_sql = self.sql(expr) expr_sql = f" {expr_sql}" if expr_sql else "" return f"{this} {kind}{expr_sql}" def returnsproperty_sql(self, expression: exp.ReturnsProperty) -> str: table = expression.args.get("table") table = f"{table} " if table else "" return f"RETURNS {table}{self.sql(expression, 'this')}" def returning_sql(self, expression: exp.Returning) -> str: into = self.sql(expression, "into") into = self.seg(f"INTO {into}") if into else "" return f"{self.seg('OUTPUT')} {self.expressions(expression, flat=True)}{into}" def transaction_sql(self, expression: exp.Transaction) -> str: this = self.sql(expression, "this") this = f" {this}" if this else "" mark = self.sql(expression, "mark") mark = f" WITH MARK {mark}" if mark else "" return f"BEGIN TRANSACTION{this}{mark}" def commit_sql(self, expression: exp.Commit) -> str: this = self.sql(expression, "this") this = f" {this}" if this else "" durability = expression.args.get("durability") durability = ( f" WITH (DELAYED_DURABILITY = {'ON' if durability else 'OFF'})" if durability is not None else "" ) return f"COMMIT TRANSACTION{this}{durability}" def rollback_sql(self, expression: exp.Rollback) -> str: this = self.sql(expression, "this") this = f" {this}" if this else "" return f"ROLLBACK TRANSACTION{this}" def identifier_sql(self, expression: exp.Identifier) -> str: identifier = super().identifier_sql(expression) if expression.args.get("global"): identifier = f"##{identifier}" elif expression.args.get("temporary"): identifier = f"#{identifier}" return identifier def constraint_sql(self, expression: exp.Constraint) -> str: this = self.sql(expression, "this") expressions = self.expressions(expression, flat=True, sep=" ") return f"CONSTRAINT {this} {expressions}" def length_sql(self, expression: exp.Length) -> str: return self._uncast_text(expression, "LEN") def right_sql(self, expression: exp.Right) -> str: return self._uncast_text(expression, "RIGHT") def left_sql(self, expression: exp.Left) -> str: return self._uncast_text(expression, "LEFT") def _uncast_text(self, expression: exp.Expression, name: str) -> str: this = expression.this if isinstance(this, exp.Cast) and this.is_type(exp.DataType.Type.TEXT): this_sql = self.sql(this, "this") else: this_sql = self.sql(this) expression_sql = self.sql(expression, "expression") return self.func(name, this_sql, expression_sql if expression_sql else None) def partition_sql(self, expression: exp.Partition) -> str: return f"WITH (PARTITIONS({self.expressions(expression, flat=True)}))" def alter_sql(self, expression: exp.Alter) -> str: action = seq_get(expression.args.get("actions") or [], 0) if isinstance(action, exp.AlterRename): return f"EXEC sp_rename '{self.sql(expression.this)}', '{action.this.name}'" return super().alter_sql(expression) def drop_sql(self, expression: exp.Drop) -> str: if expression.args["kind"] == "VIEW": expression.this.set("catalog", None) return super().drop_sql(expression) def declare_sql(self, expression: exp.Declare) -> str: return f"DECLARE {self.expressions(expression, flat=True)}" def declareitem_sql(self, expression: exp.DeclareItem) -> str: variable = self.sql(expression, "this") default = self.sql(expression, "default") default = f" = {default}" if default else "" kind = self.sql(expression, "kind") if isinstance(expression.args.get("kind"), exp.Schema): kind = f"TABLE {kind}" return f"{variable} AS {kind}{default}" def options_modifier(self, expression: exp.Expression) -> str: options = self.expressions(expression, key="options") return f" OPTION{self.wrap(options)}" if options else "" def dpipe_sql(self, expression: exp.DPipe) -> str: return self.sql( reduce(lambda x, y: exp.Add(this=x, expression=y), expression.flatten()) )