[
https://issues.apache.org/jira/browse/HBASE-9747?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13792403#comment-13792403
]
Julian Zhou commented on HBASE-9747:
------------------------------------
Is this kinds of logic? -->
In the for loop of FilterList#filterRowKey, there are 2 filters in this case:
PrefixFilter and SingleColumnValueFilter. The 1st row 'a1','f1:q2','111' was
filtered by PrefixFilter, but was passed (survived) via the 2nd filter
SingleColumnValueFilter, since SingleColumnValueFilter is actually not a rowkey
filter but a row/kv filter, but it inherits from
FilterBase#filterAllRemaining() and FilterBase#filterRowKey's returning "false"
value, which is not expected in this case.
Sounds like we need to deny the "non-rowkey filter" in the rowkey filter
execution path to avoid the filtered row survived?
> PrefixFilter with OR condition gives wrong results
> --------------------------------------------------
>
> Key: HBASE-9747
> URL: https://issues.apache.org/jira/browse/HBASE-9747
> Project: HBase
> Issue Type: Bug
> Components: Filters
> Affects Versions: 0.94.9
> Reporter: Deepa Remesh
>
> PrefixFilter when used with a SingleColumnValueFilter with an OR condition
> gives wrong results. In below example, each filter when evaluated separately
> gives 1 row each. The OR condition with the two filters gives 3 rows instead
> of 2. Repro below:
> create 't1', 'f1'
> put 't1','a1','f1:q2','111'
> put 't1','b1','f1:q1','112'
> put 't1','c1','f1:q1','113'
> hbase(main):020:0> scan 't1', {FILTER => "PrefixFilter ('b') OR
> SingleColumnValueFilter('f1', 'q1', =, 'binary:113')"}
> ROW COLUMN+CELL
> a1 column=f1:q2,
> timestamp=1381468905492, value=111
> b1 column=f1:q1,
> timestamp=1381468905518, value=112
> c1 column=f1:q1,
> timestamp=1381468905549, value=113
> 3 row(s) in 0.1020 seconds
> hbase(main):021:0> scan 't1', {FILTER => "PrefixFilter ('b')"}
> ROW COLUMN+CELL
> b1 column=f1:q1,
> timestamp=1381468905518, value=112
> 1 row(s) in 0.0150 seconds
> hbase(main):002:0> scan 't1', {FILTER => "SingleColumnValueFilter('f1', 'q1',
> =, 'binary:113')"}
> ROW COLUMN+CELL
> c1 column=f1:q1,
> timestamp=1381469178679, value=113
> 1 row(s) in 0.0140 seconds
--
This message was sent by Atlassian JIRA
(v6.1#6144)
