zabetak commented on code in PR #4591:
URL: https://github.com/apache/calcite/pull/4591#discussion_r2447731048
##########
druid/src/test/java/org/apache/calcite/test/DruidAdapterIT.java:
##########
@@ -3219,7 +3220,7 @@ private void testCountWithApproxDistinct(boolean approx,
String sql, String expe
+ "and '1997-03-01 00:00:00 UTC' order by t limit 2";
final String druidQuery =
"{'queryType':'scan','dataSource':'foodmart','intervals':"
- +
"['1997-01-01T00:00:00.000Z/1997-04-01T00:00:00.000Z'],'virtualColumns':"
+ +
"['1997-01-01T00:00:00.000Z/1997-03-01T00:00:00.001Z'],'virtualColumns':"
Review Comment:
Initially, I had the same impression as you that the old plan/result was
incorrect. However, I just looked the query more carefully and I realized that
in fact the new query plan is wrong. The `WHERE` condition uses
`floor("timestamp" to MONTH)` for the comparison so essentially a timestamp
such as `1997-03-15T00:00:00.000` should become `1997-03-01T00:00:00.000` and
be present in the result.
Changing the query interval to `1997-03-01T00:00:00.001Z` will incorrectly
prune out dates such as `1997-03-15T03:00:10.000`. I will have to revisit the
Druid conversion code to see where the bug is.
--
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]
