[jira] [Commented] (SOLR-3230) Performance improvement for geofilt by doing a bbox approximation and then Filter
[ https://issues.apache.org/jira/browse/SOLR-3230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13457906#comment-13457906 ] Robert Muir commented on SOLR-3230: --- What's the status on this issue: it hasn't been touched for months. Is it going to make 4.0? Performance improvement for geofilt by doing a bbox approximation and then Filter - Key: SOLR-3230 URL: https://issues.apache.org/jira/browse/SOLR-3230 Project: Solr Issue Type: Improvement Reporter: Bill Bell Assignee: Yonik Seeley Fix For: 4.0 Attachments: SOLR-3230-3.patch, SOLR-3230.patch This changes {!geofilt} to use a bounding box and then does a accurate filter. See attachment -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-3230) Performance improvement for geofilt by doing a bbox approximation and then Filter
[ https://issues.apache.org/jira/browse/SOLR-3230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13457967#comment-13457967 ] David Smiley commented on SOLR-3230: FYI, FWIW the new SpatialTwoDoubles does a bbox range query approximation and then does the distance calculation for a circle query shape after. I'm referring to the Solr adapter to a corresponding Lucene 4 spatial strategy class. Looking through the comments above, it seems another option I didn't consider was doing the bbox range on the fieldcache (so-called frange style). Performance improvement for geofilt by doing a bbox approximation and then Filter - Key: SOLR-3230 URL: https://issues.apache.org/jira/browse/SOLR-3230 Project: Solr Issue Type: Improvement Reporter: Bill Bell Assignee: Yonik Seeley Fix For: 4.0 Attachments: SOLR-3230-3.patch, SOLR-3230.patch This changes {!geofilt} to use a bounding box and then does a accurate filter. See attachment -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-3230) Performance improvement for geofilt by doing a bbox approximation and then Filter
[ https://issues.apache.org/jira/browse/SOLR-3230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13233162#comment-13233162 ] Bill Bell commented on SOLR-3230: - Yonik, Can you take a peak at my patch? Would love to get some feedback. We are seeing x2 performance improvements on filtering using this patch. Performance improvement for geofilt by doing a bbox approximation and then Filter - Key: SOLR-3230 URL: https://issues.apache.org/jira/browse/SOLR-3230 Project: Solr Issue Type: Improvement Reporter: Bill Bell Assignee: Grant Ingersoll Fix For: 4.0 Attachments: SOLR-3230-3.patch, SOLR-3230.patch This changes {!geofilt} to use a bounding box and then does a accurate filter. See attachment -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-3230) Performance improvement for geofilt by doing a bbox approximation and then Filter
[ https://issues.apache.org/jira/browse/SOLR-3230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13229046#comment-13229046 ] Bill Bell commented on SOLR-3230: - Yonik... I am not that familiar with this code. I do notice 2 methods in LatLonType.java. Is this the right place? public Query getFieldQuery(QParser parser, SchemaField field, String externalVal) { public Query getRangeQuery(QParser parser, SchemaField field, String part1, String part2, boolean minInclusive, boolean maxInclusive) { I did not see how these 2 functions are called. In class SpatialDistanceQuery I did not see where you said we are using range or fc... ? Maybe example code ? Performance improvement for geofilt by doing a bbox approximation and then Filter - Key: SOLR-3230 URL: https://issues.apache.org/jira/browse/SOLR-3230 Project: Solr Issue Type: Improvement Reporter: Bill Bell Assignee: Grant Ingersoll Fix For: 4.0 Attachments: SOLR-3230.patch This changes {!geofilt} to use a bounding box and then does a accurate filter. See attachment -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-3230) Performance improvement for geofilt by doing a bbox approximation and then Filter
[ https://issues.apache.org/jira/browse/SOLR-3230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13227891#comment-13227891 ] Yonik Seeley commented on SOLR-3230: bq. There is no way to cache and force the order on 1st request? You would currently need to use the lucene query parser to construct a single query with both: {code} fq=+_query_:{!bbox} +_query_:{!geofilt} {code} Not pretty, but lets you experiment at least. bq. Parameter? style=range or fieldcache? Default to fieldcache as it is now? perhaps method=fc? As far as the default, it's really tough to tell. For small distances, I'd guess that range queries would normally be faster, and that's likely to be the common case? Performance improvement for geofilt by doing a bbox approximation and then Filter - Key: SOLR-3230 URL: https://issues.apache.org/jira/browse/SOLR-3230 Project: Solr Issue Type: Improvement Reporter: Bill Bell Assignee: Grant Ingersoll Fix For: 4.0 Attachments: SOLR-3230.patch This changes {!geofilt} to use a bounding box and then does a accurate filter. See attachment -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-3230) Performance improvement for geofilt by doing a bbox approximation and then Filter
[ https://issues.apache.org/jira/browse/SOLR-3230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13227077#comment-13227077 ] Yonik Seeley commented on SOLR-3230: Looks like this patch would probably break post filtering for geofilt since the spatial query would no longer be top-level. geofilt already does do a bounding box, but it does it frange style where it checks the fieldcache against documents. It will tend to be be faster when many points match, and slower when few points match. It will also tend to be faster when cache=false and few points match the main query + other filters. Hence, it's very important to test these differences in the exact context they will be used in (i.e. with typical queries and filters for your app). One way to put the bbox before the geofilt when both are non-cached is {code} fq={!bbox cache=false cost=50} fq={!geofilt cache=false cost=60} {code} I'm not sure what the right default is, or if we can somehow dynamically detect which will be better for a particular request. We could add a flag on geofilt to allow switching between the different bounding box styles (fieldcache vs standard range). Performance improvement for geofilt by doing a bbox approximation and then Filter - Key: SOLR-3230 URL: https://issues.apache.org/jira/browse/SOLR-3230 Project: Solr Issue Type: Improvement Reporter: Bill Bell Assignee: Grant Ingersoll Fix For: 4.0 Attachments: SOLR-3230.patch This changes {!geofilt} to use a bounding box and then does a accurate filter. See attachment -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-3230) Performance improvement for geofilt by doing a bbox approximation and then Filter
[ https://issues.apache.org/jira/browse/SOLR-3230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13227261#comment-13227261 ] Bill Bell commented on SOLR-3230: - Yeah that actually works too. There is no way to cache and force the order on 1st request? Parameter? style=range or fieldcache? Default to fieldcache as it is now? I'll try to do it I guess. Frange is cached well but has too much overhead on initial request for us. We need it to be fast on the initial request. BBox does that for us. Performance improvement for geofilt by doing a bbox approximation and then Filter - Key: SOLR-3230 URL: https://issues.apache.org/jira/browse/SOLR-3230 Project: Solr Issue Type: Improvement Reporter: Bill Bell Assignee: Grant Ingersoll Fix For: 4.0 Attachments: SOLR-3230.patch This changes {!geofilt} to use a bounding box and then does a accurate filter. See attachment -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-3230) Performance improvement for geofilt by doing a bbox approximation and then Filter
[ https://issues.apache.org/jira/browse/SOLR-3230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13227008#comment-13227008 ] Bill Bell commented on SOLR-3230: - I am pretty new to Filters and all of this. But it appeared that the SpatialDistanceQuery could just be added to the bbox logic to sequence it. But I could use a code review. {code} if (options.bbox) { spatial.bboxQuery = result; return spatial; } else { result.add(spatial, BooleanClause.Occur.MUST); return result; } } {code} Performance improvement for geofilt by doing a bbox approximation and then Filter - Key: SOLR-3230 URL: https://issues.apache.org/jira/browse/SOLR-3230 Project: Solr Issue Type: Improvement Reporter: Bill Bell Assignee: Grant Ingersoll Fix For: 4.0 Attachments: SOLR-3230.patch This changes {!geofilt} to use a bounding box and then does a accurate filter. See attachment -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-3230) Performance improvement for geofilt by doing a bbox approximation and then Filter
[ https://issues.apache.org/jira/browse/SOLR-3230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13227009#comment-13227009 ] Bill Bell commented on SOLR-3230: - I did not add a new filter, since it seems a no brainer to always do the bbox approximation first. Performance improvement for geofilt by doing a bbox approximation and then Filter - Key: SOLR-3230 URL: https://issues.apache.org/jira/browse/SOLR-3230 Project: Solr Issue Type: Improvement Reporter: Bill Bell Assignee: Grant Ingersoll Fix For: 4.0 Attachments: SOLR-3230.patch This changes {!geofilt} to use a bounding box and then does a accurate filter. See attachment -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-3230) Performance improvement for geofilt by doing a bbox approximation and then Filter
[ https://issues.apache.org/jira/browse/SOLR-3230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13227011#comment-13227011 ] Bill Bell commented on SOLR-3230: - Ohh. Using exampledate. Score is the distance haversine. http://localhost:8983/solr/select?q={!func}geodist()fl=score,storesort=score ascsfield=stored=348fq={!geofilt}pt=45.19614,-93.90341 -- returns 7 http://localhost:8983/solr/select?q={!func}geodist()fl=score,storesort=score ascsfield=stored=347fq={!geofilt}pt=45.19614,-93.90341 -- returns 6 http://localhost:8983/solr/select?q={!func}geodist()fl=score,storesort=score ascsfield=stored=347fq={!bbox}pt=45.19614,-93.90341 -- returns 7 (since it is not accurate) Performance improvement for geofilt by doing a bbox approximation and then Filter - Key: SOLR-3230 URL: https://issues.apache.org/jira/browse/SOLR-3230 Project: Solr Issue Type: Improvement Reporter: Bill Bell Assignee: Grant Ingersoll Fix For: 4.0 Attachments: SOLR-3230.patch This changes {!geofilt} to use a bounding box and then does a accurate filter. See attachment -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org