Petascope returns wrong bounds with CRS:1 subsets

[Piero Campalani]

When subsetting via CRS:1 (grid) coordinates in a WCS request, the grid coordinates are not back-transformed to domain coordinates so e.g. the domainSet is ok while the boundedBy gets broken.

Es.:

http://localhost:8080/petascope/wcs2?
   version=2.0.0&
   service=WCS&
   request=GetCoverage&
   coverageid=eobstest&
   subset=x(0,3111)&
   subset=y,CRS:1(10,15)&

Returns:

[...]
  <boundedBy>
    <Envelope (...)>
      <lowerCorner>0.0 25.0 **10.0**</lowerCorner> WRONG Y BOUND
      <upperCorner>5.0 75.0 **15.0**</upperCorner> WRONG Y BOUND
    </Envelope>
  </boundedBy>
   <domainSet>
    <RectifiedGrid (...)>
      <limits>
        <GridEnvelope>
          <low>0 0 10</low>
          <high>5 100 15</high>
        </GridEnvelope>
      </limits>
    [...]
    </domainSet>
[...]

[Dimitar Misev]

Piero is it still valid? Otherwise we should close it.

[Piero Campalani]

It is still not fixed:

http://localhost:8080/petascope/wcs2?
  version=2.0.0&
  service=WCS&
  request=GetCoverage&
  coverageid=eobstest&
  subset=Long(0,3111)&
  subset=Lat,CRS:1(10,15)

However in the WCS KVP protocol binding corrigendum 1.0.1 (OGC 09-147r3), the subset shall not contain the CRS anymore (see Requirement 8).

So that request is illegal actually and Petascope should return an Exception, since the WCS CRS extension has been shutdown in r'c.

The CRS:1 is however still valid in WCPS syntax.

What should be do about this ticket?

PS The WCS request should look something like:

http://localhost:8080/petascope/wcs2?
  version=2.0.0&
  service=WCS&
  request=GetCoverage&
  coverageid=eobstest&
  subset=Long(0,3111)&
  subset=j(10,15) # or subset=Lat(10,15) ?
  subsettingCrs=http://www.opengis.net/def/crs-compound?
              1=http://www.opengis.net/def/crs/4326@Long
              2=http://www.opengis.net/def/crs/Index1D

[Peter Baumann]

WCPS 1.0 is outdated in CRS handling and does not reflect current OGC understanding, WCS is recent and upto date in this respect. Hence, let's not get stuck by WCPS. Indeed, as CRS handling is a separate extension it should be tested elsewhere. Based on these considerations I suggest to close this issue.

[Piero Campalani]

This has been fixed anyway somewhere in the middle of the sea of patches (might be changeset:db28696).