Fix(spark)!: avoid redundant casts in FROM/TO_UTC_TIMESTAMP (#4725)

Author: georgesittas

sqlglot/dialects/spark2.py

@@ -185,8 +185,12 @@ class Parser(Hive.Parser):
             "DAYOFYEAR": lambda args: exp.DayOfYear(this=exp.TsOrDsToDate(this=seq_get(args, 0))),
             "DOUBLE": _build_as_cast("double"),
             "FLOAT": _build_as_cast("float"),
-            "FROM_UTC_TIMESTAMP": lambda args: exp.AtTimeZone(
-                this=exp.cast(seq_get(args, 0) or exp.Var(this=""), exp.DataType.Type.TIMESTAMP),
+            "FROM_UTC_TIMESTAMP": lambda args, dialect: exp.AtTimeZone(
+                this=exp.cast(
+                    seq_get(args, 0) or exp.Var(this=""),
+                    exp.DataType.Type.TIMESTAMP,
+                    dialect=dialect,
+                ),
                 zone=seq_get(args, 1),
             ),
             "INT": _build_as_cast("int"),
@@ -202,8 +206,12 @@ class Parser(Hive.Parser):
                 else build_formatted_time(exp.StrToTime, "spark")(args)
             ),
             "TO_UNIX_TIMESTAMP": exp.StrToUnix.from_arg_list,
-            "TO_UTC_TIMESTAMP": lambda args: exp.FromTimeZone(
-                this=exp.cast(seq_get(args, 0) or exp.Var(this=""), exp.DataType.Type.TIMESTAMP),
+            "TO_UTC_TIMESTAMP": lambda args, dialect: exp.FromTimeZone(
+                this=exp.cast(
+                    seq_get(args, 0) or exp.Var(this=""),
+                    exp.DataType.Type.TIMESTAMP,
+                    dialect=dialect,
+                ),
                 zone=seq_get(args, 1),
             ),
             "TRUNC": lambda args: exp.DateTrunc(unit=seq_get(args, 1), this=seq_get(args, 0)),

tests/dialects/test_databricks.py

@@ -32,10 +32,6 @@ def test_databricks(self):
         self.validate_identity(
             "CREATE TABLE IF NOT EXISTS db.table (a TIMESTAMP, b BOOLEAN GENERATED ALWAYS AS (NOT a IS NULL)) USING DELTA"
         )
-        self.validate_identity(
-            "SELECT DATE_FORMAT(CAST(FROM_UTC_TIMESTAMP(foo, 'America/Los_Angeles') AS TIMESTAMP), 'yyyy-MM-dd HH:mm:ss') AS foo FROM t",
-            "SELECT DATE_FORMAT(CAST(FROM_UTC_TIMESTAMP(CAST(foo AS TIMESTAMP), 'America/Los_Angeles') AS TIMESTAMP), 'yyyy-MM-dd HH:mm:ss') AS foo FROM t",
-        )
         self.validate_identity(
             "SELECT * FROM sales UNPIVOT INCLUDE NULLS (sales FOR quarter IN (q1 AS `Jan-Mar`))"
         )
@@ -54,6 +50,10 @@ def test_databricks(self):
         self.validate_identity(
             "COPY INTO target FROM `s3://link` FILEFORMAT = AVRO VALIDATE = ALL FILES = ('file1', 'file2') FORMAT_OPTIONS ('opt1'='true', 'opt2'='test') COPY_OPTIONS ('mergeSchema'='true')"
         )
+        self.validate_identity(
+            "SELECT DATE_FORMAT(CAST(FROM_UTC_TIMESTAMP(foo, 'America/Los_Angeles') AS TIMESTAMP), 'yyyy-MM-dd HH:mm:ss') AS foo FROM t",
+            "SELECT DATE_FORMAT(CAST(FROM_UTC_TIMESTAMP(CAST(foo AS TIMESTAMP), 'America/Los_Angeles') AS TIMESTAMP), 'yyyy-MM-dd HH:mm:ss') AS foo FROM t",
+        )
         self.validate_identity(
             "DATE_DIFF(day, created_at, current_date())",
             "DATEDIFF(DAY, created_at, CURRENT_DATE)",
@@ -62,6 +62,10 @@ def test_databricks(self):
             r'SELECT r"\\foo.bar\"',
             r"SELECT '\\\\foo.bar\\'",
         )
+        self.validate_identity(
+            "FROM_UTC_TIMESTAMP(x::TIMESTAMP, tz)",
+            "FROM_UTC_TIMESTAMP(CAST(x AS TIMESTAMP), tz)",
+        )
 
         self.validate_all(
             "CREATE TABLE foo (x INT GENERATED ALWAYS AS (YEAR(y)))",

tests/dialects/test_presto.py

@@ -406,7 +406,7 @@ def test_time(self):
             },
         )
         self.validate_all(
-            "SELECT AT_TIMEZONE(CAST(CAST('2012-10-31 00:00' AS TIMESTAMP WITH TIME ZONE) AS TIMESTAMP), 'America/Sao_Paulo')",
+            "SELECT AT_TIMEZONE(CAST('2012-10-31 00:00' AS TIMESTAMP WITH TIME ZONE), 'America/Sao_Paulo')",
             read={
                 "spark": "SELECT FROM_UTC_TIMESTAMP(TIMESTAMP '2012-10-31 00:00', 'America/Sao_Paulo')",
             },