Adar Dembo has posted comments on this change. ( http://gerrit.cloudera.org:8080/15034 )
Change subject: KUDU-2483 Integrate BlockBloomFilter with ColumnPredicate on server side ...................................................................... Patch Set 13: > In general, the bloom filter is used as run-time filter, especially in a > hash-join case. For a computer engine, such as impala, it will wait 1 > second(default) to produce the bloom filter and then push it down. So, I have > another suggestion: maybe we can push it down in the middle of the scan while > the filter is not arrived within the specified interval. It should be in the > next patch^_^ > > computer engine -> kudu client -> kudu server > [1] [2] [3] > throw away > [0] That's interesting: are you suggesting we need the ability to add a bloom filter predicate to an ongoing scan, rather than just when starting a new scan? There's a lot of existing machinery that treats predicates as just another aspect of scan configuration (along with e.g. projections), to be applied to a new scanner but immutable after that. I think it'd be a fair amount of work to change that. As an alternative, would it be possible to delay the onset of a Kudu scan from Impala until the bloom filter can be constructed? Or will doing that stall the entire query pipeline? -- To view, visit http://gerrit.cloudera.org:8080/15034 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I7ecfd67e9c5fbe459c5b4aed91e0be2a194d433a Gerrit-Change-Number: 15034 Gerrit-PatchSet: 13 Gerrit-Owner: Bankim Bhavsar <[email protected]> Gerrit-Reviewer: Adar Dembo <[email protected]> Gerrit-Reviewer: Bankim Bhavsar <[email protected]> Gerrit-Reviewer: Kudu Jenkins (120) Gerrit-Reviewer: Yao Xu <[email protected]> Gerrit-Reviewer: ZhangYao <[email protected]> Gerrit-Reviewer: helifu <[email protected]> Gerrit-Comment-Date: Sat, 08 Feb 2020 05:53:43 +0000 Gerrit-HasComments: No
