[jira] [Commented] (HBASE-26967) FilterList with FuzzyRowFilter and SingleColumnValueFilter evaluated with operator MUST_PASS_ONE doesn't work as expected

["Duo Zhang (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-26967?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17681557#comment-17681557 ] 

Duo Zhang commented on HBASE-26967:
-----------------------------------

Ah, actually I'm not very familiar with FuzzyRowFilter in the past.

I skimmed the code, seems the problem is what [~Dzuba] described, FuzzyRowFilter should implement filterRow, as it performs the filtering on row key, if the row key does not match we will skip the whole row...

Let me try to fix the problem based on [~chaijunjie]'s PR.

Thanks.

> FilterList with FuzzyRowFilter and SingleColumnValueFilter evaluated with operator MUST_PASS_ONE doesn't work as expected
> -------------------------------------------------------------------------------------------------------------------------
>
>                 Key: HBASE-26967
>                 URL: https://issues.apache.org/jira/browse/HBASE-26967
>             Project: HBase
>          Issue Type: Bug
>          Components: Filters
>    Affects Versions: 2.4.11
>            Reporter: Boris
>            Priority: Blocker
>
> I created test table with two column families by hbase shell:
>  
> {code:java}
> create 'test_table2', 'f1', 'f2'
> put 'test_table2', '1', 'f1:col1', 'a1'
> put 'test_table2', '1', 'f2:col2', 'a2'
> put 'test_table2', '2', 'f1:col1', 'b1'
> put 'test_table2', '2', 'f2:col2', 'b2' {code}
>  
>  
> The table contains of two rows (rowkeys '1' and '2'), tested FuzzyRowFilter selects first row,
> SingleColumnValueFilter selects no rows, combination of both filters evaluated with
> MUST_PASS_ONE operator returns surprisingly whole table. I prepared java examples to
> show this strange behavior.
>  
> Code snippet below doesn't work as expected:
>  
> {code:java}
> try (Table table = connection.getTable(TableName.valueOf("test_table2"))) {
>     Scan scan = new Scan();
>     scan.addFamily(Bytes.toBytes("f1"));
>     scan.addFamily(Bytes.toBytes("f2"));
>     Filter fuzzyRowFilter = new FuzzyRowFilter(List.of(new Pair<>(Bytes.toBytes("1"), new byte[] { 0x00 })));
>     scan.setFilter(fuzzyRowFilter);
>     System.out.println("result of fuzzy filter:");
>     for (Result r : table.getScanner(scan)) {
>         System.out.println(Bytes.toString(r.getRow()));
>     }
>     System.out.println("result of single column value filter:");
>     Filter singleColumnValueFilter = new SingleColumnValueFilter(Bytes.toBytes("f2"), Bytes.toBytes("col2"), CompareOperator.EQUAL,
>             Bytes.toBytes("x"));
>     scan.setFilter(singleColumnValueFilter);
>     for (Result r : table.getScanner(scan)) {
>         System.out.println(Bytes.toString(r.getRow()));
>     }
>     System.out.println("result of fuzzy or single column value filters:");
>     FilterList filterList = new FilterList(Operator.MUST_PASS_ONE);
>     filterList.addFilter(fuzzyRowFilter);
>     filterList.addFilter(singleColumnValueFilter);
>     scan.setFilter(filterList);
>     for (Result r : table.getScanner(scan)) {
>         System.out.println(Bytes.toString(r.getRow()));
>     }
> } {code}
> Expected result in my opinion is:
>  
> {quote}result of fuzzy filter:
> 1
> result of single column value filter:
> result of fuzzy or single column value filters:
> 1
> {quote}
> But i am getting (NOT OK):
> {quote}result of fuzzy filter:
> 1
> result of single column value filter:
> result of fuzzy or single column value filters:
> 1
> 2
> {quote}
>  
> For tables with one column family or commentig out the line _{color:#000000}scan{color}.addFamily({color:#000000}Bytes{color}.toBytes({color:#067d17}"f1"{color}))_ filter list evaluation is working OK. Similar example with PrefixFilter is working like a charm:
>  
> {code:java}
> try (Table table = connection.getTable(TableName.valueOf("test_table2"))) {
>     Scan scan = new Scan();
>     scan.addFamily(Bytes.toBytes("f1"));
>     scan.addFamily(Bytes.toBytes("f2"));
>     Filter prefixFilter = new PrefixFilter(Bytes.toBytes("1"));
>     scan.setFilter(prefixFilter);
>     System.out.println("result of prefix filter:");
>     for (Result r : table.getScanner(scan)) {
>         System.out.println(Bytes.toString(r.getRow()));
>     }
>     System.out.println("result of single column value filter:");
>     Filter singleColumnValueFilter = new SingleColumnValueFilter(Bytes.toBytes("f2"), Bytes.toBytes("col2"), CompareOperator.EQUAL,
>             Bytes.toBytes("x"));
>     scan.setFilter(singleColumnValueFilter);
>     for (Result r : table.getScanner(scan)) {
>         System.out.println(Bytes.toString(r.getRow()));
>     }
>     System.out.println("result of prefix or single column value filters");
>     FilterList filterList = new FilterList(Operator.MUST_PASS_ONE);
>     filterList.addFilter(prefixFilter);
>     filterList.addFilter(singleColumnValueFilter);
>     scan.setFilter(filterList);
>     for (Result r : table.getScanner(scan)) {
>         System.out.println(Bytes.toString(r.getRow()));
>     }
> } {code}
>  
>  
> Result OK{_}:{_}
> {quote}result of prefix filter:
> 1
> result of single column value filter:
> result of prefix or single column value filters
> 1
> {quote}



--
This message was sent by Atlassian Jira
(v8.20.10#820010)