Tanuj Khurana created PHOENIX-7748:
--------------------------------------
Summary: Issues when scan has both EmptyColumnOnlyFilter and
DistinctPrefixFilter
Key: PHOENIX-7748
URL: https://issues.apache.org/jira/browse/PHOENIX-7748
Project: Phoenix
Issue Type: Improvement
Affects Versions: 5.3.0, 5.2.1, 5.1.3
Reporter: Tanuj Khurana
Assignee: Tanuj Khurana
When the scan only needs to look at row key columns then Phoenix puts a
FirstKeyOnlyFilter if column encoding is enabled or EmptyColumnOnlyFilter if
column encoding is disabled. For distinct queries on PK prefix columns, a
DistinctPrefix filter is also appended at the end of the filter list. The issue
that happens when EmptyColumnOnlyFilter and DistinctPrefixFilter are present
together is that instead of returning the empty column cell we return a regular
column cell. The absence of the empty column cell in the scan causes multiple
issues:
# For indexes, if the row doesn't have an empty column cell we assume that the
row is unverified and unnecessarily repair it.
# For datatable, if the row doesn't have an empty column cell we cannot
determine if the row has expired or not. This means that we can return expired
rows as part of the query result.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
