[Gdal-dev] Re: DISOWN in ogr.i

[Charlie Savage]

Ari Jolma wrote:
> 
> The DISOWN method, which swig generates for Perl, is for the Perl 
> programmer to use, and does not automate the handing of the object from 
> Perl to C. This is an unnecessary extra task for him/her so I'd like the 
> bindings to do it, as it assumably does for the other languages. Now a 
> (not "the") solution (with swig 1.2.25 at least) is to write %perlcode 
> into the bindings to explicitly call DISOWN in these few cases 
> (SetGeometryDirectly etc).

I see.  Yes, having the bindings manage this is preferable like in Ruby 
and Python.

> 
> Regarding the files in the CVS and swig version, I'd think that would be 
> a task for the PSC to decide ;-)

PSC?

> Just a few thoughts: Are people developing GDAL in Windows? Couldn't 
> there be a nightly make dist for those who need a bleeding edge version? 
> The Fedora 5 still has swig 1.3.24, installing 1.3.29 is an extra step, 
> which is not needed for Perl bindings I believe.

Yup, Fedora 5 is using SWIG 1.3.24.  Don't know about other distros? 
Seems easier to me to avoid the whole issue and have an "approved" swig 
version and include the bindings in CVS as is currently done.

The argument for using SWIG 1.3.29 is that it is the second release of 
the new unified type maps for Ruby/Python/Perl/TCL.  Which means instead 
of having separate typemaps for all languages, we could have one set 
with tweaks for individual languages as needed.  The counter arguments 
are a) its only the second release of the unified type maps so perhaps 
they'll change b) they don't cover the C++ containers (vectors, etc.) 
yet c) it would take some short-term work.


Thanks,

Charlie