[Gdal-dev] RE: too precise projection comparison

[Ben Discoe]

Matt,

I debugged into it, and found that GDAL (or OGR in this case) is actually
very picky about projection parameters matching exactly.

It's not the text in the "degree" field, and it's not the cell size.  It's
this:
  PARAMETER["standard_parallel_1",61.66666666666672]
  PARAMETER["standard_parallel_1",61.66666666666673]

Yes, that's enough to cause OGRSpatialReference::IsSame to return false.

So, Matt, in lieu of OGR adding some kind of epsilon test, i'd recommend
that you track the changes back to find where the projections diverged in
your tool chain, and/or simply fix the text in your .prj files.  For
example, it's certain that the correct value above is actually:
  PARAMETER["standard_parallel_1",61.66666666666666]

-Ben

> -----Original Message-----
> From: Matt, Kim, Cole & Danni [mailto:patawi at polarcom.com] 
> Sent: Sunday, October 17, 2004 2:57 AM
> To: Ben Discoe
> Subject: Re: too precise projection comparison
> 
> >>Bug (sort of): VTBuilder comparing projections is too 
> >>precise. It offers to convert projections when the
> >>mismatch is down on the 14th decimal place:
> 
> Ahh, I was led astray by the error message. I assumed that 
> because there was a difference in the numbers in the message,
> that was the entire extent of the difference. After manually
> comparing the .prj files I found another difference:
> 
> A.prj:
> UNIT["degree"
> 
> B.prj:
> UNIT["degree (supplier to define representation)"
> 
> And of course the cell sizes differ. There must be another 
> one in there too though because when I manually edit the
> "supplier..." text I didn't get the message. Sample data
> attached.
> 
> I can't remember how I generated B, it was either vtconvert or 
> gdal_translate back in early May.
> 
> -matt