[
https://issues.apache.org/jira/browse/LUCENE-5476?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13922293#comment-13922293
]
Rob Audenaerde commented on LUCENE-5476:
----------------------------------------
Thanks Shai,
I have fixed the points you noted about the collector. I renamed the
sampleThreshold to sampleSize. It currently picks a samplingRatio that will
reduce the number of hits to the sampleSize, if the number of hits is greater.
I have a general question about your remarks about the test, besides fixing
the obvious (names, commit, sops). Is there a reason to add more randomness to
one test? I normally try to test one aspect in a unit test. And if I also want
to test some other aspect, like random document counts (to test the sampleratio
for example), I add more tests.
{quote}
Make the two collector instances take 100/10% of the numDocs when you fix it
{quote}
Sorry, I don't get what you mean by this.
{quote}
I don't understand how you know that numChildren=5 when you ask for the 10
top children. Isn't it possible that w/ some random seed the number of children
will be different?
In fact, I think that the random collectors should be initialized w/ a
random seed that depends on the test? Currently they aren't and so always use
0xdeadbeef?
{quote}
There will be 5 facet values (0, 2, 4, 6 and 8), as only the even documents (i
% 10) are hits. There is a REAL small chance that one of the five values will
be entirely missed when sampling. But is that {{0.8 (chance not to take a
value) ^ 2000 * 5 (any can be missing) ~ 10^-193}}, so that is probable not
going to happen :).
> Facet sampling
> --------------
>
> Key: LUCENE-5476
> URL: https://issues.apache.org/jira/browse/LUCENE-5476
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Rob Audenaerde
> Attachments: LUCENE-5476.patch, LUCENE-5476.patch, LUCENE-5476.patch,
> LUCENE-5476.patch, LUCENE-5476.patch, LUCENE-5476.patch,
> SamplingComparison_SamplingFacetsCollector.java, SamplingFacetsCollector.java
>
>
> With LUCENE-5339 facet sampling disappeared.
> When trying to display facet counts on large datasets (>10M documents)
> counting facets is rather expensive, as all the hits are collected and
> processed.
> Sampling greatly reduced this and thus provided a nice speedup. Could it be
> brought back?
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
