Re: [patch #6245] Perl Module

[Jörg Beyer]

Thanks, John. 
Guido sent me the same URL off-list last week.
I replied in private, but it may be a good idea to re-post it to the list,
my experiences with this Perl script may be of public interest.

<mail to Guido Gay>

... 

I already found this script back in May or so. Tests with various not too
complicated SPSS-files of different versions all broke early with a
particular error, AFAIR -- possibly due to the fact, that I'm on a PPC with
OS X, here, so there may be some endianness related problems (just a guess).
Plus, it is a 'flat' script that had to be refactored as a module to be
really useful, plus the return value is far from what I need, plus it
doesn't seem to read dictionary information of newer versions, as you
already mentioned. 

So for me there are a lot of problems with it, but I stored it nonetheless
to get some basic insights and additional hints. Haven't looked too close at
the code yet, but it may turn out to be useful, should it come to a pure
Perl solution. 

... 

</mail to Guido Gay>


As for Python, yes, I think compatibility with SPSS in this respect is
necessary. In general, if I had to plan in which direction PSPP should go, I
would start with the premise 'make it compatible, but make it smarter, and
try to offer more than the competition'. Offering scripting support is a
good example: SPSS offers Python -- a kind of stripped down Python, AFAIK.
Going one step further and being able to offer _full_ support for more than
one language would make PSPP more attractive to more users. But the costs
have to be taken into account -- that's why SWIG came to my mind, it could
possibly help to streamline this aspect.

Cheers, 

Jörg



Am 30.10.2007 12:05 Uhr schrieb John Darrington (<address@hidden>):

> On Tue, Oct 30, 2007 at 11:20:14AM +0100, Jörg Beyer wrote:
>>      
>> Another thought I had during the week:
>> The current approach is to use Perl's own XS mechanism to set up the
>> bindings. I don't know much about Python, but I assume it offers
>> comparable means.
>> But what do you think in general about exploring SWIG in this context?
>> Again, this is all unfamiliar territory for me, but my idea would be to
>> use the benefits of _unifying_ bindings to whichever dynamic language comes
>> to mind. It may save some time, and it may even help to get some kind of
>> unified interface, or it may lead to some recommendations, how to design
>> the interface for language <xyz>.
>> SWIG would add another dependency, of course...
> 
> I'm not intimately familiar with swig nor with python,  but I'm sure
> it would be possible to build such an interface.
> Spss has recently come up with its own Python interface, so it would
> make sense to make the two compatible.
> 
> 
> Guido Gay sent me this Url http://czep.net/data/spssread/ It  is
> a Perl program which can apparently read PSPP/SPSS files.  It may be
> of interest to you.
> 
> 
> Regards,
> 
> 
> John