From 8d36f5966675e23bee7026ba37ae0647fbf47300 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Mon, 8 Apr 2024 10:11:53 +0200 Subject: Merging upstream version 23.7.0. Signed-off-by: Daniel Baumann --- sqlglot/dialects/presto.py | 40 +++++++++++++++++++++++++++++----------- 1 file changed, 29 insertions(+), 11 deletions(-) (limited to 'sqlglot/dialects/presto.py') diff --git a/sqlglot/dialects/presto.py b/sqlglot/dialects/presto.py index 3649bd2..25bba96 100644 --- a/sqlglot/dialects/presto.py +++ b/sqlglot/dialects/presto.py @@ -22,10 +22,13 @@ from sqlglot.dialects.dialect import ( rename_func, right_to_substring_sql, struct_extract_sql, + str_position_sql, timestamptrunc_sql, timestrtotime_sql, ts_or_ds_add_cast, + unit_to_str, ) +from sqlglot.dialects.hive import Hive from sqlglot.dialects.mysql import MySQL from sqlglot.helper import apply_index_offset, seq_get from sqlglot.tokens import TokenType @@ -93,14 +96,14 @@ def _ts_or_ds_to_date_sql(self: Presto.Generator, expression: exp.TsOrDsToDate) def _ts_or_ds_add_sql(self: Presto.Generator, expression: exp.TsOrDsAdd) -> str: expression = ts_or_ds_add_cast(expression) - unit = exp.Literal.string(expression.text("unit") or "DAY") + unit = unit_to_str(expression) return self.func("DATE_ADD", unit, expression.expression, expression.this) def _ts_or_ds_diff_sql(self: Presto.Generator, expression: exp.TsOrDsDiff) -> str: this = exp.cast(expression.this, "TIMESTAMP") expr = exp.cast(expression.expression, "TIMESTAMP") - unit = exp.Literal.string(expression.text("unit") or "DAY") + unit = unit_to_str(expression) return self.func("DATE_DIFF", unit, expr, this) @@ -196,6 +199,7 @@ class Presto(Dialect): SUPPORTS_SEMI_ANTI_JOIN = False TYPED_DIVISION = True TABLESAMPLE_SIZE_IS_PERCENT = True + LOG_BASE_FIRST: t.Optional[bool] = None # https://github.com/trinodb/trino/issues/17 # https://github.com/trinodb/trino/issues/12289 @@ -289,6 +293,7 @@ class Presto(Dialect): SUPPORTS_SINGLE_ARG_CONCAT = False LIKE_PROPERTY_INSIDE_SCHEMA = True MULTI_ARG_DISTINCT = False + SUPPORTS_TO_NUMBER = False PROPERTIES_LOCATION = { **generator.Generator.PROPERTIES_LOCATION, @@ -323,6 +328,7 @@ class Presto(Dialect): exp.ArrayConcat: rename_func("CONCAT"), exp.ArrayContains: rename_func("CONTAINS"), exp.ArraySize: rename_func("CARDINALITY"), + exp.ArrayToString: rename_func("ARRAY_JOIN"), exp.ArrayUniqueAgg: rename_func("SET_AGG"), exp.AtTimeZone: rename_func("AT_TIMEZONE"), exp.BitwiseAnd: lambda self, e: self.func("BITWISE_AND", e.this, e.expression), @@ -339,19 +345,19 @@ class Presto(Dialect): exp.CurrentTimestamp: lambda *_: "CURRENT_TIMESTAMP", exp.DateAdd: lambda self, e: self.func( "DATE_ADD", - exp.Literal.string(e.text("unit") or "DAY"), + unit_to_str(e), _to_int(e.expression), e.this, ), exp.DateDiff: lambda self, e: self.func( - "DATE_DIFF", exp.Literal.string(e.text("unit") or "DAY"), e.expression, e.this + "DATE_DIFF", unit_to_str(e), e.expression, e.this ), exp.DateStrToDate: datestrtodate_sql, exp.DateToDi: lambda self, e: f"CAST(DATE_FORMAT({self.sql(e, 'this')}, {Presto.DATEINT_FORMAT}) AS INT)", exp.DateSub: lambda self, e: self.func( "DATE_ADD", - exp.Literal.string(e.text("unit") or "DAY"), + unit_to_str(e), _to_int(e.expression * -1), e.this, ), @@ -397,13 +403,10 @@ class Presto(Dialect): ] ), exp.SortArray: _no_sort_array, - exp.StrPosition: rename_func("STRPOS"), + exp.StrPosition: lambda self, e: str_position_sql(self, e, generate_instance=True), exp.StrToDate: lambda self, e: f"CAST({_str_to_time_sql(self, e)} AS DATE)", exp.StrToMap: rename_func("SPLIT_TO_MAP"), exp.StrToTime: _str_to_time_sql, - exp.StrToUnix: lambda self, e: self.func( - "TO_UNIXTIME", self.func("DATE_PARSE", e.this, self.format_time(e)) - ), exp.StructExtract: struct_extract_sql, exp.Table: transforms.preprocess([_unnest_sequence]), exp.Timestamp: no_timestamp_sql, @@ -436,6 +439,22 @@ class Presto(Dialect): exp.Xor: bool_xor_sql, } + def strtounix_sql(self, expression: exp.StrToUnix) -> str: + # Since `TO_UNIXTIME` requires a `TIMESTAMP`, we need to parse the argument into one. + # To do this, we first try to `DATE_PARSE` it, but since this can fail when there's a + # timezone involved, we wrap it in a `TRY` call and use `PARSE_DATETIME` as a fallback, + # which seems to be using the same time mapping as Hive, as per: + # https://joda-time.sourceforge.net/apidocs/org/joda/time/format/DateTimeFormat.html + value_as_text = exp.cast(expression.this, "text") + parse_without_tz = self.func("DATE_PARSE", value_as_text, self.format_time(expression)) + parse_with_tz = self.func( + "PARSE_DATETIME", + value_as_text, + self.format_time(expression, Hive.INVERSE_TIME_MAPPING, Hive.INVERSE_TIME_TRIE), + ) + coalesced = self.func("COALESCE", self.func("TRY", parse_without_tz), parse_with_tz) + return self.func("TO_UNIXTIME", coalesced) + def bracket_sql(self, expression: exp.Bracket) -> str: if expression.args.get("safe"): return self.func( @@ -481,8 +500,7 @@ class Presto(Dialect): return f"CAST(ROW({', '.join(values)}) AS ROW({', '.join(schema)}))" def interval_sql(self, expression: exp.Interval) -> str: - unit = self.sql(expression, "unit") - if expression.this and unit.startswith("WEEK"): + if expression.this and expression.text("unit").upper().startswith("WEEK"): return f"({expression.this.name} * INTERVAL '7' DAY)" return super().interval_sql(expression) -- cgit v1.2.3