Is this happening reproducibly from the demo app? Please try and reproduce it
there; if you can't then the problem is somewhere in your setup, I figure. If
you can reproduce it in the demo then please send me a direct email with the
polygon shape, and your expectation of a particular point that should fit.
~ David
On Jul 21, 2011, at 12:40 PM, Jamie Johnson wrote:
> I think I've missed something. From what I'm seeing it appears that a
> bounding box is being built from my polygon and any points in that
> bounding box are returned. This makes sense from the debug which says
> the query is
> +(+point__x:[-75.267333984375 TO -74.569702148438]
> +point__y:[39.512329101563 TO 40.523071289063])
> +DistanceValueSource(org.apache.lucene.spatial.base.distance.EuclidianDistanceCalculator@c7d9406)
>
> Given that does point support doing what I am trying to do or should I
> be using another field type?
>
> I understand that you're busy so no rush on this.
>
> On Thu, Jul 21, 2011 at 11:48 AM, Jamie Johnson <jej2...@gmail.com> wrote:
>> Thanks for the reply.
>>
>> My setup has a point in the field and a shape as the query. Given
>> this it sounds as if I can get more precise results by changing the
>> distErrPct on a query parameter. I'll give this a whirl. Again thank
>> you.
>>
>>
>> On Thu, Jul 21, 2011 at 11:13 AM, Smiley, David W. <dsmi...@mitre.org> wrote:
>>> If you are talking about indexed shapes, then there is an attribute on the
>>> field type definition in your schema called "distErrPct". Reasonable
>>> values are between .01 and .20, in my opinion. The default is .025, but
>>> try setting it to .01. For points, use the "maxDetailKm" parameter, which
>>> is the kilometer detail level. By default, that parameter is .001 -- 1
>>> meter.
>>>
>>> If you are talking about your query shape, then this same parameter can be
>>> supplied as a request parameter. Again, the default is .025. The
>>> RecursiveGridFieldType can handle infinite query side precision, so you can
>>> supply 0 and still get reasonable performance. However if your indexing
>>> side is a certain precision, then there's little point in using more
>>> precision on the query side since in-effect it's as accurate as your index
>>> side.
>>>
>>> If you're wondering more about the meaning of distErrPct, see this snippet
>>> from SpatialArgs.java:
>>> /**
>>> * The fraction of the distance from the center of the query shape to its
>>> nearest edge that is considered acceptable
>>> * error. The algorithm for computing the distance to the nearest edge is
>>> actually a little different. It normalizes
>>> * the shape to a square given it's bounding box area:
>>> * <pre>sqrt(shape.bbox.area)/2</pre>
>>> * And the error distance is beyond the shape such that the shape is a
>>> minimum shape.
>>> */
>>> public Double getDistPrecision() {
>>>
>>> ~ David
>>>
>>> On Jul 20, 2011, at 5:44 PM, Jamie Johnson wrote:
>>>
>>>> Thanks David. When trying to execute queries on a complex irregular
>>>> polygon (say the shape of NJ) I'm getting results which are actually
>>>> outside of that polygon. Is there a setting which controls this
>>>> resolution?
>>>>
>>>> On Wed, Jul 20, 2011 at 2:53 PM, Smiley, David W. <dsmi...@mitre.org>
>>>> wrote:
>>>>> The notion of a "system property" is a java concept; google it and you'll
>>>>> learn more.
>>>>>
>>>>> BTW, despite my responsiveness in helping right now; I'm pretty busy this
>>>>> week so this won't necessarily last long.
>>>>> ~ David
>>>>>
>>>>> On Jul 20, 2011, at 2:43 PM, Jamie Johnson wrote:
>>>>>
>>>>>> Where do you set that?
>>>>>>
>>>>>> On Wed, Jul 20, 2011 at 2:37 PM, Smiley, David W. <dsmi...@mitre.org>
>>>>>> wrote:
>>>>>>> You can set the system property SpatialContextProvider to
>>>>>>> com.googlecode.lucene.spatial.base.context.JtsSpatialContext
>>>>>>>
>>>>>>> ~ David
>>>>>>>
>>>>>>> On Jul 20, 2011, at 2:02 PM, Jamie Johnson wrote:
>>>>>>>
>>>>>>>> So I've pulled the latest and can run the example, I've tried to move
>>>>>>>> my config over and am having a bit of an issue when executing queries,
>>>>>>>> specifically I get this:
>>>>>>>>
>>>>>>>> Unable to read: POLYGON((...
>>>>>>>>
>>>>>>>> looking at the code it's usign the simple spatial context, how do I
>>>>>>>> specify JtsSpatialContext?
>>>>>>>>
>>>>>>>> On Wed, Jul 20, 2011 at 12:13 PM, Jamie Johnson <jej2...@gmail.com>
>>>>>>>> wrote:
>>>>>>>>> Thanks for the update David, I'll give that a try now.
>>>>>>>>>
>>>>>>>>> On Wed, Jul 20, 2011 at 10:58 AM, Smiley, David W.
>>>>>>>>> <dsmi...@mitre.org> wrote:
>>>>>>>>>> Ryan just updated LSP for Lucene/Solr trunk compatibility so you
>>>>>>>>>> should do a "mvn clean install" and you'll be back in business.
>>>>>>>>>>
>>>>>>>>>> On Jul 20, 2011, at 10:37 AM, Jamie Johnson wrote:
>>>>>>>>>>
>>>>>>>>>>> Thanks for responding so quickly, I don't mind waiting a bit. I'll
>>>>>>>>>>> hang out until the updates have been made. Thanks again.
>>>>>>>>>>>
>>>>>>>>>>> On Tue, Jul 19, 2011 at 3:51 PM, Smiley, David W.
>>>>>>>>>>> <dsmi...@mitre.org> wrote:
>>>>>>>>>>>> Hi Jamie.
>>>>>>>>>>>> I work on LSP; it can index polygons and query for them. Although
>>>>>>>>>>>> the capability is there, we have more testing & benchmarking to
>>>>>>>>>>>> do, and then we need to put together a tutorial to explain how to
>>>>>>>>>>>> use it at the Solr layer. I recently cleaned up the READMEs a
>>>>>>>>>>>> bit. Try downloading the trunk codebase, and follow the README.
>>>>>>>>>>>> It points to another README which shows off a demo webapp. At the
>>>>>>>>>>>> conclusion of this, you'll need to examine the tests and webapp a
>>>>>>>>>>>> bit to figure out how to apply it in your app. We don't yet have
>>>>>>>>>>>> a tutorial as the framework has been in flux although it has
>>>>>>>>>>>> stabilized a good deal.
>>>>>>>>>>>>
>>>>>>>>>>>> Oh... by the way, this works off of Lucene/Solr trunk. Within the
>>>>>>>>>>>> past week there was a major change to trunk and LSP won't compile
>>>>>>>>>>>> until we make updates. Either Ryan McKinley or I will get to that
>>>>>>>>>>>> by the end of the week. So unless you have access to 2-week old
>>>>>>>>>>>> maven artifacts of Lucene/Solr, you're stuck right now.
>>>>>>>>>>>>
>>>>>>>>>>>> ~ David Smiley
>>>>>>>>>>>> Author: http://www.packtpub.com/solr-1-4-enterprise-search-server/
>>>>>>>>>>>>
>>>>>>>>>>>> On Jul 19, 2011, at 3:03 PM, Jamie Johnson wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>> I have looked at the code being shared on the
>>>>>>>>>>>>> lucene-spatial-playground and was wondering if anyone could
>>>>>>>>>>>>> provide
>>>>>>>>>>>>> some details as to its state. Specifically I'm looking to add
>>>>>>>>>>>>> geospatial support to my application based on a user provided
>>>>>>>>>>>>> polygon,
>>>>>>>>>>>>> is this currently possible using this extension?
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>
>>>>>
>>>
>>>
>>
