[gdal-dev] Backward incompatible changes in the OGR style string format (r19724)

[Daniel Morissette]

(Cross-posted to mapserver-dev)

Okay, I finally found traces of the previous discussions on this topic
in one GDAL ticket, and a related MITAB bugzilla entry:

http://trac.osgeo.org/gdal/ticket/3571

and

http://bugzilla.maptools.org/show_bug.cgi?id=2206

<wrist_slap>
Tamas and I would have saved lots of time if winkey had included a
reference to the ticket number in his SVN commit change log... please
all developers take this as a reminder to include refs to ticket number
in SVN change logs, and references to the SVN revision number when
closing tickets... this makes it much easier to go back in history.
</wrist_slap>

Sounds like what needs to happen is:

1- Add version-specific code in MapServer's msOGRGetSymbolId() to look
for "," vs "." delimiter in symbol ids depending on GDAL/OGR version

2- Add prominent note about this backwards incompatibility in the MITAB
driver notes, the GDAL/OGR 1.8 release notes, and on the list so that
developers of other applications using style strings can update their
code before it breaks in the user's hands

3- Update MITAB upstream with the same change (so that it is not
overwritten in future updates of the OGR copy)

4- And perhaps ensure that no other driver in OGR currently generates
symbol ids with "." and need the same fix.

What do you think Tamas?

Daniel


Daniel Morissette wrote:
> Hi Tamas,
> 
> I remember that I discussed with other devs this a little while ago, I'm
> just not sure if that was on the list or on IRC... however I'm a bit
> surprised since I see that r19724 was done by winkey and I thought I had
> this discussion with aboudreault, so I'll need to dig a bit further to
> refresh my mind about this.
> 
> Anyway, I think in the end the conclusion was that we should bite the
> bullet and fix things to really use "," as a separator (instead of "."
> which is currently wrong and causes some side-effects which I can't
> remember just now). The only side-effect of this fix is on applications
> using Style strings, and there are probably very few of them, the only
> one I know about myself is MapServer, and the fix for them will be to
> add a version-specific #ifdef to look for either "," or "." depending on
> the GDAL/OGR version.
> 
> I'll try to dig out the discussions that we had about this and get back
> to you with more details.
> 
> Daniel
> 
> 
> Tamas Szekeres wrote:
>> Hi All,
>>
>> There have been a backward incompatible change in r19724
>> <http://trac.osgeo.org/gdal/changeset/19724> related to the OGR style
>> string format. According to this change the separator in the feature
>> style id has been changed from '.' to ','
>> Unfortunately such kind of changes may cause existing applications to
>> work incorrectly and would also need to be fixed. For instance this
>> change affects the mapserver function msOGRGetSymbolId in mapogr.cpp
>> <http://trac.osgeo.org/mapserver/browser/trunk/mapserver/mapogr.cpp> as
>> well. So as to fix this issue in mapserver, we would probably require to
>> keep the original code simultaneously, depending on the gdal version.
>>
>> I'd be curious to know whether we have any compelling reason to apply
>> these modifications, or we just wanted to follow the feature style
>> specification <http://www.gdal.org/ogr/ogr_feature_style.html> with
>> these changes? In the latter case I would personally in favour of
>> reverting these changes in gdal and handle this issue by fixing the
>> feature style documentation to follow the original syntax.
>>
>> Any ideas?
>>
>>
>> Best regards,
>>
>> Tamas
>>
>>
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> gdal-dev mailing list
>> gdal-dev at lists.osgeo.org
>> http://lists.osgeo.org/mailman/listinfo/gdal-dev
> 
> 


-- 
Daniel Morissette
http://www.mapgears.com/