[
https://issues.apache.org/jira/browse/JCR-1213?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12542418
]
Ard Schrijvers commented on JCR-1213:
-------------------------------------
"The first check (1) is the reason why you created this issue"
Not entirely: currently, the CombinedIndexReader instance is used as a
WeakReference, and this one is recreated for every search. The MultiIndexReader
instance is kept AFAICS as long as all indexes are the same. So, in
SearchIndex, changing
public int getParent(int n) throws IOException {
.....
return id.getDocumentNumber(this);
}
into
public int getParent(int n) throws IOException {
.....
return id.getDocumentNumber(subReaders[i]);
}
would already implement (1). This one holds when *every* index reader instance
is the same.
If, one of the instances has changed, we would need step (2) IIUC. Then we
could check wether the instance the parent was found in is still valid, and, as
you indicate, should return the 'corrected' DocNumber, which might be different
due to applyOffSet. When (1) and (2) are both invalid, then the search for the
parent node in subReaders[i] should be done again.
I agree, that (1) is redundant because (2) captures (1) , but I added it,
because first of all, it is something we can add right away, and secondly,
because I think (but I should measure) that if the subReaders[i] instance
(MultiIndexReader) did not change, it is useless to do a lookup of the index
reader segment the parent was in and check wether the instance is still valid.
I do agree with you that if removing (1) does not imply any performance loss,
we should only go for (2). But it is not correct that (1) does not solve
anything to the original problem: instead of the CombinedIndexReader which is
recreated all the time, I pass in the MultiIndexReader whose instance is kept
as long as no indexes change. This is at least what I understand from the
mechanism, but I am not as familiar as you are with it ofcourse, so I might be
off.
> UUIDDocId cache does not work properly because of weakReferences in
> combination with new instance for combined indexreader
> ---------------------------------------------------------------------------------------------------------------------------
>
> Key: JCR-1213
> URL: https://issues.apache.org/jira/browse/JCR-1213
> Project: Jackrabbit
> Issue Type: Improvement
> Components: query
> Affects Versions: 1.3.3
> Reporter: Ard Schrijvers
> Fix For: 1.4
>
>
> Queries that use ChildAxisQuery or DescendantSelfAxisQuery make use of
> getParent() functions to know wether the parents are correct and if the
> result is allowed. The getParent() is called recursively for every hit, and
> can become very expensive. Hence, in DocId.UUIDDocId, the parents are cached.
> Currently, docId.UUIDDocId's are cached by having a WeakRefence to the
> CombinedIndexReader, but, this CombinedIndexReader is recreated all the time,
> implying that a gc() is allowed to remove the 'expensive' cache.
> A much better solution is to not have a weakReference to the
> CombinedIndexReader, but to a reference of each indexreader segment. This
> means, that in getParent(int n) in SearchIndex the return
> return id.getDocumentNumber(this) needs to be replaced by return
> id.getDocumentNumber(subReaders[i]); and something similar in
> CachingMultiReader.
> That is all. Obviously, when a node/property is added/removed/changed, some
> parts of the cached DocId.UUIDDocId will be invalid, but mainly small indexes
> are updated frequently, which obviously are less expensive to recompute.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
