lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Tim Smith (JIRA)" <>
Subject [jira] Commented: (LUCENE-1821) Weight.scorer() not passed doc offset for "sub reader"
Date Fri, 21 Aug 2009 12:37:14 GMT


Tim Smith commented on LUCENE-1821:

I'm OK with having to jump through some hoops in order to get back to the "full index" context

It would be nice if this was more facilitated by lucene's API (IMO, this would be best handled
by adding a Searcher as the first arg to Weight.scorer(), as then a Weight will not need to
hold on to this (breaking serializable))

There are definitely plenty of use cases that take advantage of the "whole" index (one created
by IndexWriter), so this ability should not be removed
I have at least 3 in my application alone (and they are all very important)

You get tradeoffs working "Per-Segment" vs "Per-MultiReader" when it comes to caching in general
going per-segment means caches load faster, and load less frequently, however this causes
algorithms working with the caches to be slower (depending on algorithm and cache type)

for static boosting from a field value (ValueSource), it makes no difference
for numeric sorting, it makes no difference 

for string sorting, it makes a big difference - you now have to do a bunch of String.equals()
calls, where you didn't have to in 2.4 (just used the ord index)
Given this reason, you should really be able to do string sorting 2 ways
* using per segment field cache (commit time/first query faster, sort time slower)
* using multi-reader field cache (commit time/first query slower, sort time faster)

This same argument also goes for features like faceting (not provided by lucene, but is provided
by applications like solr, and my application). Using a per-segment cache will cause some
significant performance loss when performing faceting, as it requires creating the facets
for each segment, and then merging them (this results in a good deal of extra object overhead/memory
overhead/more work where faceting on the multi-reader does not see this)

In the end, it should be up to the application developer to choose what strategy works best
for them, and their application (fast commits/fast cache loading may take a back seat to fast
query execution)

In general, i find there is a tradeoff between commit time and query time. The more you speed
up commit time, the slower query time gets, and vice versa
I just want/need the ability to choose

> Weight.scorer() not passed doc offset for "sub reader"
> ------------------------------------------------------
>                 Key: LUCENE-1821
>                 URL:
>             Project: Lucene - Java
>          Issue Type: Bug
>          Components: Search
>    Affects Versions: 2.9
>            Reporter: Tim Smith
>             Fix For: 2.9
>         Attachments: LUCENE-1821.patch
> Now that searching is done on a per segment basis, there is no way for a Scorer to know
the "actual" doc id for the document's it matches (only the relative doc offset into the segment)
> If using caches in your scorer that are based on the "entire" index (all segments), there
is now no way to index into them properly from inside a Scorer because the scorer is not passed
the needed offset to calculate the "real" docid
> suggest having Weight.scorer() method also take a integer for the doc offset
> Abstract Weight class should have a constructor that takes this offset as well as a method
to get the offset
> All Weights that have "sub" weights must pass this offset down to created "sub" weights
> Details on workaround:
> In order to work around this, you must do the following:
> * Subclass IndexSearcher
> * Add "int getIndexReaderBase(IndexReader)" method to your subclass
> * during Weight creation, the Weight must hold onto a reference to the passed in Searcher
(casted to your sub class)
> * during Scorer creation, the Scorer must be passed the result of YourSearcher.getIndexReaderBase(reader)
> * Scorer can now rebase any collected docids using this offset
> Example implementation of getIndexReaderBase():
> {code}
> // NOTE: more efficient implementation can be done if you cache the result if gatherSubReaders
in your constructor
> public int getIndexReaderBase(IndexReader reader) {
>   if (reader == getReader()) {
>     return 0;
>   } else {
>     List readers = new ArrayList();
>     gatherSubReaders(readers);
>     Iterator iter = readers.iterator();
>     int maxDoc = 0;
>     while (iter.hasNext()) {
>       IndexReader r = (IndexReader);
>       if (r == reader) {
>         return maxDoc;
>       } 
>       maxDoc += r.maxDoc();
>     } 
>   }
>   return -1; // reader not in searcher
> }
> {code}
> Notes:
> * This workaround makes it so you cannot serialize your custom Weight implementation

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message