cpoerschke commented on a change in pull request #151:
URL: https://github.com/apache/solr/pull/151#discussion_r641707464
##########
File path:
solr/core/src/java/org/apache/solr/handler/component/QueryComponent.java
##########
@@ -999,20 +1013,34 @@ protected void mergeIds(ResponseBuilder rb, ShardRequest
sreq) {
shardDoc.sortFieldValues = unmarshalledSortFieldValues;
- queue.insertWithOverflow(shardDoc);
+ if(reRankQueue != null && docCounter++ <= reRankDocsSize) {
+ ShardDoc droppedShardDoc =
reRankQueue.insertWithOverflow(shardDoc);
+ // FIXME: Only works if the original request does not sort by
score
Review comment:
\<thinking out aloud\>
> ... The reRankDocs-param which specifies the amount of docs that should be
reRanked, is used per shard, but also has to be applied while combining the
results. ...
> Example: ... reRankDocs = 2 shard1: ... shard2: ...
Might it be helpful to include in the example the original score for doc_1
and doc_4 and to have a "pre-example" of what the logic and outcome is if the
collection was single sharded and a "post-example" for (say) 4-sharded?
Can distributed, sharded re-ranking arrive at exactly the same outcome?
Might the _"but also has to be applied while combining the results"_ be
up-for-debate e.g. a reRankDocs on shard level and a (possibly different)
foobarReRankDocs on the federation level?
> ... We have no access to the score before reRanking at this point and I
currently see no possibility to retrieve it again. ...
Conceptually if a feature store had an `OriginalScoreFeature` and `fl=[fv]`
feature values were requested then in terms of "transport" the score before
reRanking could be available in the federator. As a proof-of-concept that could
be explored by making sure the store has the feature and the fl asks for the
feature values. As a real implementation it could be a bit messy to capture,
send and use the score that way. And actually, hmm, the score might be
available via the `fl` via the "get fields" part of the distributed use of the
query component but if/since the id merging happens before "get fields" then
yes no original score available during merging.
\</thinking out aloud\>
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
