[mapserver-dev] testing CLUSTER object

[Stephen Woodbridge]

I was thinking about the issue of the cluster symbols getting clipped by 
tile edge effects and I think that we should treat symbols similar to 
labels and be able to say PARTIAL FALSE. The down side of this is that 
symbols near edges might never get drawn. Also, Am I correct in 
remembering the clusters are based on the features in the view, so a 
cluster on the right or left side of adjoining edges will be based on 
the meta-tile to the right or left which might make some strange effects.

Tamas, you probably have a better read on this since you thought through 
the implementation in much more detail than I ever have.

Jeff, did you setup the standard buffer error around the metatile that 
excludes labels so they are not chopped up? I assume you did so the 
cropped symbol is probably not aware of the exclusion zone.

-Steve W

On 4/14/2011 3:13 PM, Tamas Szekeres wrote:
>
> 2011/4/14 Jeff McKenna <jmckenna at gatewaygeomatics.com
> <mailto:jmckenna at gatewaygeomatics.com>>
>
>     - why isn't the LAYER TYPE ANNOTATION supported for the cluster
>     object?  A common trick for labelling symbols is to use an
>     annotation layer, I really think the cluster logic should be applied
>     to that type as well.
>
>
> Hi Jeff,
>
> We have no specific reason not to support the annotation layers since
> those are point layers in most cases. I will see how it could be enabled
> too.
>
>     - I think there is a problem with query results from a cluster.  If
>     I query a non-cluster-object group of features with mode=nquery (and
>     a template set at the LAYER level) I get the correct number of
>     features returned; If I query the same cluster-object group of
>     features with mode=nquery (and a template set at the LAYER level) I
>     only ever get one feature result (testing with MapServer trunk).
>       Thoughts?
>
>
> I'm not sure what you mean by querying a  non-cluster-object group or
> querying the same cluster-object group. With the current approach we can
> query only those features which are actually generated (displayed) for
> that layer (ie the clustered features) and there's no such mechanism to
> retrieve the features behind a cluster (if I understand the question
> correctly). This option would require the driver to behave differently
> regarding to the query and the normal drawing operations.
>
>     - If you look at my example layer
>     (http://www.mapserver.org/trunk/mapfile/cluster.html#example-clustering-railway-stations),
>     you can see that I had to do some tricks with an expression to not
>     show a cluster object for when the featurecount = 1; should this be
>     the case?  Would anyone using MapServer ever need a cluster for one
>     feature?
>
>
> Good question, I thought if that can be filtered by using the filter
> expression of the cluster object.
>
>     - I've had "challenges" trying to stop the cluster circle symbols
>     from being split at tile edges with mod-geocache
>     (http://labs.gatewaygeomatics.com/cluster/demo.html) Yes it could be
>     a missing setting either in my mapfile or in my geocache.xml, but I
>     still want to report this.  Maybe this is a good sign that we
>     (MapServer or mod-geocache) need to make it easier for new users to
>     prevent this from occurring?
>
>
> Is this dedicated to the cluster layer as a problem?
>
> Thanks for testing this, your comments and document additions were
> really helpful.
>
> Best regards,
>
> Tamas
>
>
>
> _______________________________________________
> mapserver-dev mailing list
> mapserver-dev at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev