julianhyde commented on code in PR #3024:
URL: https://github.com/apache/calcite/pull/3024#discussion_r1090160372
##########
core/src/main/codegen/templates/Parser.jj:
##########
@@ -4677,6 +4677,28 @@ SqlNode DatetimeFunctionCall() :
}
}
+/**
+ * Parses BigQuery's built-in DATETIME_DIFF() function.
Review Comment:
It's not 'BigQuery's'. Just "Parses the DATETIME_DIFF function".
##########
core/src/main/java/org/apache/calcite/sql/fun/SqlLibraryOperators.java:
##########
@@ -764,6 +764,22 @@ private SqlLibraryOperators() {
ReturnTypes.BIGINT_NULLABLE, OperandTypes.TIMESTAMP,
SqlFunctionCategory.TIMEDATE);
+ /** BigQuery's "DATETIME_ADD(timestamp, interval) function; Behaves similarly
Review Comment:
I wouldn't describe it as 'BigQuery's'. Other DBs might support it one day.
The "because in Calcite, datetime is a type alias for timestamp" is
confusing, and I'd remove it. The fact is, both accept TIMESTAMP or TIMESTAMP
WITH LOCAL TIME ZONE arguments, and both return the same type as their first
argument.
##########
testkit/src/main/java/org/apache/calcite/test/SqlOperatorTest.java:
##########
@@ -8122,6 +8122,170 @@ private static void
checkArrayConcatAggFuncFails(SqlOperatorFixture t) {
"2015-01-01 00:00:00", "TIMESTAMP(0) NOT NULL");
}
+ /** Tests BigQuery's DATETIME_ADD(timestamp, interval) function. In Calcite,
+ * datetime is a type alias for timestamp and this function follows the same
+ * behavior as TIMESTAMP_ADD(timestamp, interval). The tests below use
+ * timestamps rather than the datetime alias because the operator fixture
+ * does not currently support type aliases. */
+ @Test void testDatetimeAdd() {
+ final SqlOperatorFixture f0 = fixture()
+ .setFor(SqlLibraryOperators.DATETIME_ADD);
+ f0.checkFails("^datetime_add(timestamp '2008-12-25 15:30:00', "
+ + "interval 5 minute)^",
+ "No match found for function signature "
+ + "DATETIME_ADD\\(<TIMESTAMP>, <INTERVAL_DAY_TIME>\\)", false);
+
+ final SqlOperatorFixture f = f0.withLibrary(SqlLibrary.BIG_QUERY);
+ if (Bug.CALCITE_5422_FIXED) {
+ f.checkScalar("datetime_add(timestamp '2008-12-25 15:30:00', "
+ + "interval 100000000000 microsecond)",
+ "2008-12-26 19:16:40",
+ "TIMESTAMP(3) NOT NULL");
+ f.checkScalar("datetime_add(timestamp '2008-12-25 15:30:00', "
+ + "interval 100000000 millisecond)",
+ "2008-12-26 19:16:40",
+ "TIMESTAMP(3) NOT NULL");
+ }
+
+ f.checkScalar("datetime_add(timestamp '2016-02-24 12:42:25', interval 2
second)",
+ "2016-02-24 12:42:27",
+ "TIMESTAMP(0) NOT NULL");
+ f.checkScalar("datetime_add(timestamp '2016-02-24 12:42:25', interval 2
minute)",
+ "2016-02-24 12:44:25",
+ "TIMESTAMP(0) NOT NULL");
+ f.checkScalar("datetime_add(timestamp '2016-02-24 12:42:25', interval
-2000 hour)",
+ "2015-12-03 04:42:25",
+ "TIMESTAMP(0) NOT NULL");
+ f.checkScalar("datetime_add(timestamp '2016-02-24 12:42:25', interval 1
day)",
+ "2016-02-25 12:42:25",
+ "TIMESTAMP(0) NOT NULL");
+ f.checkScalar("datetime_add(timestamp '2016-02-24 12:42:25', interval 1
month)",
+ "2016-03-24 12:42:25",
+ "TIMESTAMP(0) NOT NULL");
+ f.checkScalar("datetime_add(timestamp '2016-02-24 12:42:25', interval 1
year)",
+ "2017-02-24 12:42:25",
+ "TIMESTAMP(0) NOT NULL");
+ f.checkNull("datetime_add(CAST(NULL AS TIMESTAMP), interval 5 minute)");
+
+ }
+
+ /** Tests BigQuery's DATETIME_DIFF(timestamp, timestamp2, timeUnit)
function. In Calcite,
Review Comment:
same remarks as for DATETIME_ADD
##########
site/_docs/reference.md:
##########
@@ -2620,6 +2622,8 @@ semantics.
| b | DATETIME(date, time) | Returns a datetime
object, given a date and a time.
| b | DATETIME(timestamp) | Converts a timestamp
object to a datetime, assuming UTC.
| b | DATETIME(timestamp, timezone) | Converts a timestamp
object to a datetime, in a given time zone.
+| b | DATETIME_ADD(timestamp, interval) | Synonym for
TIMESTAMP_ADD.
Review Comment:
remove '.'
Not really a synonym. The implementation happens to allow both TIMESTAMP and
TIMESTAMP WITH LOCAL TIME ZONE, and therefore DATETIME_ADD shares an
implementation with TIMESTAMP_ADD. But the user doesn't have to know that.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
