[ https://issues.apache.org/jira/browse/PHOENIX-3865?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
James Taylor updated PHOENIX-3865: ---------------------------------- Attachment: PHOENIX-3865-wip.patch Patch that fixes the issue. We need to take into account the empty key value column family when the empty key value will be projected. Patch needs a bit more work. > IS NULL does not return results when STORE_NULLS, and ONE_CELL_PER_COLUMN is > set on IMMUTABLE ROWS > -------------------------------------------------------------------------------------------------- > > Key: PHOENIX-3865 > URL: https://issues.apache.org/jira/browse/PHOENIX-3865 > Project: Phoenix > Issue Type: Bug > Reporter: Lars Hofhansl > Attachments: PHOENIX-3865-wip.patch > > > I'm using poor a poor man's method for sampling. In the following table, I > write a non-null value to COLB with a probability of 0.001. > {code} > create table test (mykey integer not null primary key, A.COLA integer, B.COLB > integer) IMMUTABLE_ROWS=true, STORE_NULL=false, IMMUTABLE_STORAGE_SCHEME = > ONE_CELL_PER_COLUMN, DISABLE_WAL=true > {code} > Then: > {code} > 0: jdbc:phoenix:localhost> select count(*) from test; > +-----------+ > | COUNT(1) | > +-----------+ > | 10000000 | > +-----------+ > 1 row selected (8.95 seconds) > 0: jdbc:phoenix:localhost> select count(*) from test where B.COLB is not null; > +-----------+ > | COUNT(1) | > +-----------+ > | 10054 | > +-----------+ > 1 row selected (0.023 seconds) > 0: jdbc:phoenix:localhost> select count(*) from test where B.COLB is null; > +-----------+ > | COUNT(1) | > +-----------+ > | 0 | > +-----------+ > 1 row selected (0.022 seconds) > {code} > Last statement should have returned 10000000-10054 = 9989946 rows. -- This message was sent by Atlassian JIRA (v6.3.15#6346)