[udig-devel] accuracy of GetFeatureInfo

Tue Sep 5 10:10:42 PDT 2006

On Tue, 2006-05-09 at 17:02 +0200, Jody Garnett wrote:
> In that case we will need to ask Richard to implement a different plugin 
> for the InfoTool for each renderer he creates. Not fun but possible. I 
> did think he ended up making a 2x2 reference image and selecting the 
> first pixel (after having a few WMS fail on a 1x1 image reference image.

Err.. I made a couple experiments with this, but I didn't ever think it
was really feasible. I think the goal was to be able to provide a buffer
around a point clicked, or provide support for bbox-enabled
GetFeatureInfo requests (WMS spec says you can only use point
requests). 

Since the info tool in uDig is point based, I presume that the desire is
to provide a buffer? Paul is right in that shrinking the image a bit
would be closer to the desired effect. If we shrink the image to 3x3 and
ask for one pixel, we are going to get 1/9 of the screen back (and
possibly the wrong section too). 

So I think we should default back to just using the standard method of
doing it (ie. let the WMS provide tolerance if it wishes). We could add
a preference option to enable a buffer around the point and then use the
voodoo code, but I don't think it should be a default. 

I'll see if I can fix this on Thursday morning, before the RC release.

Richard

> Jody
> >
> > I think the problem with this approach, as now demonstrated, is that 
> > it doesn't work :)  It might take something a little more "arbitrary", 
> > like a pretend image of "about 300 pixels in size" with the click 
> > point in the "middle of that image" to get a correct-seeming answer.  
> > Sending back a one-pixel map with a click "on the pixel" is begging to 
> > hit corner cases in the WMS server side code.
> >
> > On 5-Sep-06, at 12:29 AM, Jody Garnett wrote:
> >
> >> Ah, interesting question ... because of a difference between what we 
> >> display and what we requested from the WMS.
> >> - often we will have to make several requests and tile content onto 
> >> the screen in WMS happy chunks,
> >> - or we will need to make slightly larger requests and resample them 
> >> onto the screen (as mentioned previously).
> >>
> >> It was just easier to back project a pixel and use it to make a 
> >> request in the correct location once, rather then trying to reverse 
> >> several rendering process es to exactly match the original request 
> >> (aka the approach we used for uDig 0.8).
> >>
> >> Cheers,
> >> Jody
> >>> Hi Jody,
> >>>
> >>> why did you end up changing the BBOX and width/height client-side? 
> >>> Why not leave it up to the WMS server like Gaia does?
> >>>
> >>> Best regards,
> >>> Bart
> >>>
> >>> Jody Garnett schreef:
> >>>> Indeed, we eventually figured out that WMS getInfo is considered a 
> >>>> check in the center of a single pixel.
> >>>>
> >>>> Thinking,
> >>>>
> >>>> Richard for a while we forced the renderers to leave us a hook so 
> >>>> we could make our info request, is there any way we can ask the WMS 
> >>>> renderers to leave some hints on the layer blackboard about the 
> >>>> image rendered? The same plugin making the rendering contribution 
> >>>> can provide the hint ...
> >>>>
> >>>> Jody
> >>>>> Hi Jody,
> >>>>>
> >>>>> the projections I am trying are both supported by the WMS.
> >>>>>
> >>>>> Unfortunately I have no Java hosting, but it's a simple shapefile 
> >>>>> which can be set up with Geoserver. I am using 1.4.
> >>>>>
> >>>>> The shapefile can be found here (200 Kb, native projection 
> >>>>> EPSG:28992, uDIG set to RD / New projection):
> >>>>>
> >>>>> http://www.osgis.nl/lausanne/nl-prov.zip
> >>>>>
> >>>>> An example request which uDIG is fyring off is (this request is 
> >>>>> really strange since the BBOX is not the BBOX of the map image 
> >>>>> shown anymore! and X and Y are always 1):
> >>>>>
> >>>>> /geoserver/wms?Y=1&X=1&SERVICE=WMS&INFO_FORMAT=text/html&LAYERS=ogctestbed:nl-prov&FORMAT=image/png&HEIGHT=3&REQUEST=GetFeatureInfo&WIDTH=3&BBOX=188907.50415179788,391802.21024004533,192453.4125990071,395348.1186872545&SRS=EPSG:28992&QUERY_LAYERS=ogctestbed%3Anl-prov&VERSION=1.1.1 
> >>>>> HTTP/1.1" 200 447
> >>>>>
> >>>>> Thanks in advance.
> >>>>>
> >>>>> Best regards,
> >>>>> Bart
> >>>>>
> >>>>> Jody Garnett schreef:
> >>>>>> Good question, it makes a direct call when able to make a direct 
> >>>>>> rendering of the WMS image (ie udig is using a coordinate system 
> >>>>>> supported by the WMS). In rare cases it may have to resample the 
> >>>>>> image onto the screen (making up for lack of WMS projections), we 
> >>>>>> could check that we back project the query into the coordinate 
> >>>>>> space of the original image... but as I recall I wrote this one 
> >>>>>> and went to some trouble to get it right.
> >>>>>>
> >>>>>> Can you point us at a service illustrating the problem, perhaps 
> >>>>>> in a jira report.
> >>>>>> Jody
> >>>>>>> Hi list,
> >>>>>>>
> >>>>>>> how accurate is uDIG with respect to GetFeatureInfo?
> >>>>>>>
> >>>>>>> I have tried it against 2 WMS's, Geoserver and Deegree, and at 
> >>>>>>> the borders of provinces, a lot of times the wrong element is 
> >>>>>>> returned. Ofcourse this could also be due to the WMS's.
> >>>>>>>
> >>>>>>> So I checked Gaia, and using Gaia it is very accurate for 
> >>>>>>> Geoserver, so I am assuming it is a uDIG problem.
> >>>>>>>
> >>>>>>> Any input appreciated.
> >>>>>>>
> >>>>>>> Best regards,
> >>>>>>> Bart
> >>>>>>>
> >>>>>>
> >>>>>>
> >>>>>> _______________________________________________
> >>>>>> User-friendly Desktop Internet GIS (uDig)
> >>>>>> http://udig.refractions.net
> >>>>>> http://lists.refractions.net/mailman/listinfo/udig-devel
> >>>>>>
> >>>>>>
> >>>>>
> >>>>>
> >>>>
> >>>>
> >>>> _______________________________________________
> >>>> User-friendly Desktop Internet GIS (uDig)
> >>>> http://udig.refractions.net
> >>>> http://lists.refractions.net/mailman/listinfo/udig-devel
> >>>>
> >>>>
> >>>
> >>>
> >>
> >>
> >> _______________________________________________
> >> User-friendly Desktop Internet GIS (uDig)
> >> http://udig.refractions.net
> >> http://lists.refractions.net/mailman/listinfo/udig-devel
> >
> > _______________________________________________
> > User-friendly Desktop Internet GIS (uDig)
> > http://udig.refractions.net
> > http://lists.refractions.net/mailman/listinfo/udig-devel
> 
> 
> _______________________________________________
> User-friendly Desktop Internet GIS (uDig)
> http://udig.refractions.net
> http://lists.refractions.net/mailman/listinfo/udig-devel