Delete all live docs when query matched a whole segment. #13395
Conversation
|
@mikemccand |
There was a problem hiding this comment.
I am nervous about this change -- altering the very widely used Scorer API to express whether it matches all documents feels invasive.
Also, the use case this is optimizing for, feels rare? Though, perhaps if an index sorts its segments by time range, and collates documents into segments by time range as well, delete by range query over a time range might more efficiently drop whole segments if we could do something like this cleanly? Even so, it's quite rare that an index would invoke this pruning, and the likely smallish time it takes today is fine?
| @@ -98,4 +100,14 @@ public int advanceShallow(int target) throws IOException { | |||
| * {@link #advanceShallow(int) shallow-advanced} to included and {@code upTo} included. | |||
| */ | |||
| public abstract float getMaxScore(int upTo) throws IOException; | |||
|
|
|||
| /** Set whether we can match all docs in this scored segment. */ | |||
| public void setMatchAll(boolean isMatchAll) { | |||
There was a problem hiding this comment.
Hmm adding this API to such a widely used class (Scorer) makes me nervous -- it means consumers of Scorer can suddenly setMatchAll without being true?
There was a problem hiding this comment.
it means consumers of Scorer can suddenly setMatchAll without being true?
Yes, I should found a better way to mark isMatchAll.
Yes, delete by range query over a time range, is the only case which came to my mind, but the improve maybe will not be significant. |
|
This PR has not had activity in the past 2 weeks, labeling it as stale. If the PR is waiting for review, notify the dev@lucene.apache.org list. Thank you for your contribution! |
Description
We can delete all live docs when a query matched a whole segment, in
FrozenBufferedUpdates.applyQueryDeletes.I have implemented it for
PointRangeQuery, and also trying to implement it for other queries.