[BUG] WCS Error?

[Frank Warmerdam]

Ben Tuttle wrote:
> Hi all-
> I have two questions related to mapserver and WCS. First, I am running
> mapserver 4.10.0 on linux. I am trying to set up WCS access to global
> 30 arcsec data sets.
> Question 1:
>  It seems that mapserver does not respect resx and resy. Example: if
> I make the following request
> http://www.ngdc.noaa.gov/eog/maps/test5.php?SERVICE=WCS&REQUEST=GetCoverage&Coverage=vis_mosaic&STYLES=&CRS=EPSG:4326&BBOX=-125,30,-115,40&resx=0.00833333&resy=0.00833333&Format=geotiff 
> 
> the data comes back as  0.00834028,-0.00834028 instead of
> 0.00833333,-0.00833333 (the native resolution of the data is
> 0.00833333,-0.00833333). Additionally the given bounding box actually
> covers an area 1201px by 1201px in the data. However, the returned
> image is 1200px by 1200px. If compared to the original data one can
> see that mapserver has resampled the returned image to fit the area
> which is 1201 by 1201 into an image that is 1200 by 1200, which
> explains the change in the resolution (I'd be happy to provide more
> examples or information to anyone who is interested. It seems to me
> that this may not be the the desired behaviour for WCS. I would expect
> to get my data back at the same resolution requested. I haven;t seen
> any bug reports on this (I did a search for WCS and only found 6
> reports that matched wcs and none seemed to be about this issue). So
> does someone know if I this is a bug or if I don't fully grasp WCS and
> this is the desired response? Or is there a way to override this
> behaviour or is there perhaps something wrong wiht my build?

Ben,

I'd appreciate it if you could file a bug on this.  You can assign it
to me if you wish.  I've run into some "quirky" issues that may be
related as I test my WCS client support in GDAL against mapserver.
I think there may be some poor rounding choices and/or some
"middle of pixel vs. top left corner of pixel" origin issues at play.

> Question 2:
>  We would like to return data to people at the actual resolution they
> requested. To this end I am working on using wxs mapscript and php to
> examine the request and return the correct resolution. I want to be
> sure I don't stomp all over the WCS standard in the process though. So
> the question is does anyone have any thoughts onteh following process.
> If a user specifies no resx/resy or width/height I would calculate the
> appropriate width and height for the bbox to return full res data. If
> they specify resx/resy, but not width and height I'll caculate the
> appropriate width and height to return the data at the requested resx
> and resy for the bbox. If a user specifies a width and height I will
> assume they know how the width and height and bbox relate and return
> the data as requested. Does anyone feel like this goes againt the WCS
> standard?

Providing a full resolution default if no resx/resy or width/height
are provided seems reasonable, though I personally prefer for the
mapserver core wcs functionality to continue to treat this as an
unspecified condition and issue an error.

You write "If they specify resx/resy, but not width and height I'll
calculate the appropriate width and height to return the data at the
requested resx and resy for the bbox."  Doesn't this already happen?
Are you just looking to fiddle with things a bit to work around the
problem in point (1)?  If so, I think fixing the underlying problem
would be the best approach.

Best regards,
-- 
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | President OSGeo, http://osgeo.org