Ildar Absalyamov has posted comments on this change.

Change subject: [STO][IDX] Eliminated excess antimatter in LSMBTree
......................................................................


Patch Set 10:

>Thanks. It actually solves a different problem than I thought (cardinality 
>mismatching between primary and secondary indexes?). I was thinking one can 
>omit inserting the anti-matter tuple when the tuple was not found in the 
>previous disk components.
Actually correct cardinality is a direct consequence of the fact that extra 
anti-matter is removed. See, if we find an anti-matter record <sk, pk, -> in a 
memory component during the flush there are two ways it ended up being there: 
1) it deletes the entry <sk,pk> in an earlier disk component, in which case 
it's a legit anti-matter. The entry was by created by delete\upsert and never 
got replaced in-place in memory component. 2) there was an earlier record 
<sk,pk> in the same memory component that got replaced by delete\upsert to <sk, 
pk, ->. In this case a physical entry <sk, pk> was updated in-place in a BTree 
and we can trace that using proposed update bit.

-- 
To view, visit https://asterix-gerrit.ics.uci.edu/1538
To unsubscribe, visit https://asterix-gerrit.ics.uci.edu/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I12a67eff8431b52d1f9051b793a5a64b15c009e9
Gerrit-PatchSet: 10
Gerrit-Project: asterixdb
Gerrit-Branch: master
Gerrit-Owner: Ildar Absalyamov <[email protected]>
Gerrit-Reviewer: Ian Maxon <[email protected]>
Gerrit-Reviewer: Ildar Absalyamov <[email protected]>
Gerrit-Reviewer: Jenkins <[email protected]>
Gerrit-Reviewer: Luo Chen <[email protected]>
Gerrit-Reviewer: Till Westmann <[email protected]>
Gerrit-Reviewer: abdullah alamoudi <[email protected]>
Gerrit-HasComments: No

Reply via email to