[
https://issues.apache.org/jira/browse/PHOENIX-2714?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15166561#comment-15166561
]
James Taylor commented on PHOENIX-2714:
---------------------------------------
Good catch, [~maryannxue]. We need to get 4.7 out, so I think improvements can
wait until 4.8 (but it's of course open for discussion). Some other
improvements include:
- exposing getEstimatedRowCount() and getEstimatedByteCount() in QueryPlan (or
maybe StatementPlan even). For the union and the join query plans, we could be
smart about the way we add it up.
- taking into account point lookups and only adding a single blocks worth of
bytes (as the entire guidepost bytes would not be scanned).
- taking into account the additional column families that are being scanned as
well. We currently only add the guideposts bytes from the one being used to
parallelize.
[~lhofhansl] may have other ideas too.
> Correct byte estimate in BaseResultIterators and expose as interface
> --------------------------------------------------------------------
>
> Key: PHOENIX-2714
> URL: https://issues.apache.org/jira/browse/PHOENIX-2714
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 4.7.0
> Reporter: Maryann Xue
> Assignee: Maryann Xue
> Labels: statistics
> Fix For: 4.8.0
>
>
> The bytes are accumulated even if the range intersect is empty (produces a
> null scan).
> {code}
> while (guideIndex < gpsSize &&
> (currentGuidePost.compareTo(endKey) <= 0 || endKey.length == 0)) {
> Scan newScan = scanRanges.intersectScan(scan,
> currentKeyBytes, currentGuidePostBytes, keyOffset,
> false);
> estimatedRows += gps.getRowCounts().get(guideIndex);
> estimatedSize += gps.getByteCounts().get(guideIndex);
> scans = addNewScan(parallelScans, scans, newScan,
> currentGuidePostBytes, false, regionLocation);
> currentKeyBytes = currentGuidePost.copyBytes();
> currentGuidePost = PrefixByteCodec.decode(decoder, input);
> currentGuidePostBytes = currentGuidePost.copyBytes();
> guideIndex++;
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
