[
https://issues.apache.org/jira/browse/HBASE-8930?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13707920#comment-13707920
]
Lars Hofhansl commented on HBASE-8930:
--------------------------------------
This is (somewhat) by design. Filters are evaluated before the column trackers.
The column trackers are responsible for the version counting and TTL processing
(that part is by design - it used to be the other way round, but was changed
because that caused more problems).
As a side-effect the columns are also processed after the filters have been
evaluated.
Disentangling that is tedious and would lead to perf issues in other cases.
The place to change this is ExplicitColumnTracker.java. There would need to be
a method that can check a column ahead of the filtering, but the tracker still
needs to do versions/TTL after the filters ran.
Whatever the refactoring... if columns need to be compare twice that would be
an unacceptable performance hit.
At some point I had experimented with letting a filter decide whether it is
evaluated before or after the column trackers. Maybe I should restart that. See
HBASE-5257.
This seems at odds with HBASE-4364.
In the interim, the filters should do their own column filtering.
> Filter evaluates KVs outside requested columns
> ----------------------------------------------
>
> Key: HBASE-8930
> URL: https://issues.apache.org/jira/browse/HBASE-8930
> Project: HBase
> Issue Type: Bug
> Components: Filters
> Affects Versions: 0.94.7
> Reporter: Federico Gaule
> Priority: Critical
> Labels: filters, hbase, keyvalue
>
> 1- Fill row with some columns
> 2- Get row with some columns less than universe - Use filter to print kvs
> 3- Filter prints not requested columns
> Filter (AllwaysNextColFilter) always return ReturnCode.INCLUDE_AND_NEXT_COL
> and prints KV's qualifier
> SUFFIX_0 = 0
> SUFFIX_1 = 1
> SUFFIX_4 = 4
> SUFFIX_6 = 6
> P= Persisted
> R= Requested
> E= Evaluated
> X= Returned
> | 5580 | 5581 | 5584 | 5586 | 5590 | 5591 | 5594 | 5596 | 5600 | 5601 | 5604
> | 5606 |...
> | | P | P | | | P | P | | | P | P
> | |...
> | | R | R | R | | R | R | R | | |
> | |...
> | | E | E | | | E | E | | |
> {color:red}E{color} | | |...
> | | X | X | | | X | X | | | |
> | |
> {code:title=ExtraColumnTest.java|borderStyle=solid}
> @Test
> public void testFilter() throws Exception {
> Configuration config = HBaseConfiguration.create();
> config.set("hbase.zookeeper.quorum", "myZK");
> HTable hTable = new HTable(config, "testTable");
> byte[] cf = Bytes.toBytes("cf");
> byte[] row = Bytes.toBytes("row");
> byte[] col1 = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 558, (byte) SUFFIX_1));
> byte[] col2 = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 559, (byte) SUFFIX_1));
> byte[] col3 = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 560, (byte) SUFFIX_1));
> byte[] col4 = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 561, (byte) SUFFIX_1));
> byte[] col5 = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 562, (byte) SUFFIX_1));
> byte[] col6 = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 563, (byte) SUFFIX_1));
> byte[] col1g = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 558, (byte) SUFFIX_6));
> byte[] col2g = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 559, (byte) SUFFIX_6));
> byte[] col1v = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 558, (byte) SUFFIX_4));
> byte[] col2v = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 559, (byte) SUFFIX_4));
> byte[] col3v = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 560, (byte) SUFFIX_4));
> byte[] col4v = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 561, (byte) SUFFIX_4));
> byte[] col5v = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 562, (byte) SUFFIX_4));
> byte[] col6v = new QualifierConverter().objectToByteArray(new
> Qualifier((short) 563, (byte) SUFFIX_4));
> // =========== INSERTION =============//
> Put put = new Put(row);
> put.add(cf, col1, Bytes.toBytes((short) 1));
> put.add(cf, col2, Bytes.toBytes((short) 1));
> put.add(cf, col3, Bytes.toBytes((short) 3));
> put.add(cf, col4, Bytes.toBytes((short) 3));
> put.add(cf, col5, Bytes.toBytes((short) 3));
> put.add(cf, col6, Bytes.toBytes((short) 3));
> hTable.put(put);
> put = new Put(row);
> put.add(cf, col1v, Bytes.toBytes((short) 10));
> put.add(cf, col2v, Bytes.toBytes((short) 10));
> put.add(cf, col3v, Bytes.toBytes((short) 10));
> put.add(cf, col4v, Bytes.toBytes((short) 10));
> put.add(cf, col5v, Bytes.toBytes((short) 10));
> put.add(cf, col6v, Bytes.toBytes((short) 10));
> hTable.put(put);
> hTable.flushCommits();
> //==============READING=================//
> Filter allwaysNextColFilter = new AllwaysNextColFilter();
> Get get = new Get(row);
> get.addColumn(cf, col1); //5581
> get.addColumn(cf, col1v); //5584
> get.addColumn(cf, col1g); //5586
> get.addColumn(cf, col2); //5591
> get.addColumn(cf, col2v); //5594
> get.addColumn(cf, col2g); //5596
>
> get.setFilter(allwaysNextColFilter);
> get.setMaxVersions(1);
> System.out.println(get);
> Scan scan = new Scan(get);
> ResultScanner scanner = hTable.getScanner(scan);
> Iterator<Result> iterator = scanner.iterator();
> System.out.println("SCAN");
> while (iterator.hasNext()) {
> Result next = iterator.next();
> for (KeyValue kv : next.list()) {
> System.out.println(new
> QualifierConverter().byteArrayToObject(kv.getQualifier()));
> }
> }
> }
> }
> {code}
> Requested 5581 5584 5586 5591 5594 5596
> NOT REQUESTED: 5561
> Sysout Filter
> {noformat}
> \x00\x00\x1A\xBE\x00\x05^:\x00\x00\xA0X\x00\x00=\x1A/H0:\x02.\x01/1373577819267/Put/vlen=2/ts=2
> Qualifier{date=558, type=SUFFIX_1}
> \x00\x00\x1A\xBE\x00\x05^:\x00\x00\xA0X\x00\x00=\x1A/H0:\x02.\x02/1373577819272/Put/vlen=2/ts=3
> Qualifier{date=558, type=SUFFIX_4}
> \x00\x00\x1A\xBE\x00\x05^:\x00\x00\xA0X\x00\x00=\x1A/H0:\x02/\x01/1373577819267/Put/vlen=2/ts=2
> ualifier{date=559, type=SUFFIX_1}
> \x00\x00\x1A\xBE\x00\x05^:\x00\x00\xA0X\x00\x00=\x1A/H0:\x02/\x02/1373577819272/Put/vlen=2/ts=3
> Qualifier{date=559, type=SUFFIX_4}
>
> \x00\x00\x1A\xBE\x00\x05^:\x00\x00\xA0X\x00\x00=\x1A/H0:\x020\x01/1373577819267/Put/vlen=2/ts=2
> Qualifier{date=560, type=SUFFIX_1} (DATE 5601 NOT REQUESTED BUT EVALUATED)
> {noformat}
> Sysout ExtraColumnTest
> {noformat}
> {"timeRange":[0,9223372036854775807],"totalColumns":6,"cacheBlocks":true,"families":{"H0":["\\x02.\\x01","\\x02.\\x02","\\x02.\\x06","\\x02/\\x01"]},"maxVersions":1,"filter":"AllwaysNextColFilter","row":"\\x00\\x00\\x1A\\xBE\\x00\\x05^:\\x00\\x00\\xA0X\\x00\\x00=\\x1A"}
> SCAN
> Qualifier{date=558, type=SUFFIX_1}
> Qualifier{date=558, type=SUFFIX_4}
> Qualifier{date=559, type=SUFFIX_1}
> Qualifier{date=559, type=SUFFIX_4}
> {noformat}
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
