Hi Joel-

As I'm sure you are aware, the availability of the
needed Alien modules for PDL would be a home run
as far as improving the installability.  I plan to work
on an Alien::OpenGL module to support the perl
OpenGL development.

One thing that comes up is how to handle the case
that the desired feature is not installed.  I think the
Alien module should still install a stub that will fail
in the usual 'use Alien::XXX' sense *but* if called with
an import argument 'use Alien::XXX qw()' then it
would instead allow one access to the results of
the configuration testing,...

The specific example I'm thinking of is Alien::OpenGL
where I believe the correct approach would *not* be
to try to install OpenGL if missing since it could
result in breaking the users system and may
require root/admin priviledges.  However, it would
be useful to report what is there, and what is
needed and maybe some info on where to get...

--Chris

On Fri, Jan 20, 2012 at 10:33 AM, Joel Berger <[email protected]> wrote:
> Yeah, I touched base with kmx as I worked on Alien::GSL. As I mention
> in the proposal almost all the code had nothing to do with GSL, so I
> am trying to fork that code out. Alien::Base is born.
>
> On Fri, Jan 20, 2012 at 9:30 AM, Chris Marshall <[email protected]> 
> wrote:
>> Joel-
>>
>> That is great to hear!  I had some discussions with kmx on
>> this topic along the same lines.  He has a number of Alien
>> modules that actually work cross-platform as contrasted
>> with others that are there but only work on a limited set of
>> platforms...  Here is the link to the ticket/discussion:
>>
>>  https://rt.cpan.org/Ticket/Display.html?id=67568
>>
>> I've cc'd kmx as well.
>>
>> --Chris
>>
>> On Fri, Jan 20, 2012 at 10:25 AM, Joel Berger <[email protected]> 
>> wrote:
>>> Chris et al.
>>>
>>> I am working on a generic Alien:: framework called Alien::Base, hosted
>>> at my gh:
>>>
>>> https://github.com/jberger/Alien-Base .
>>>
>>> I submitted a proposal to tpf a small grant, here is the text for the 
>>> concept:
>>>
>>> https://gist.github.com/1616923
>>>
>>> Once this works, it should make Alien:: modules much easier to write.
>>>
>>> Joel
>>>
>>> On Fri, Jan 20, 2012 at 9:12 AM, Chris Marshall <[email protected]> 
>>> wrote:
>>>> On Thu, Jan 19, 2012 at 4:52 PM, Clifford Sobchuk
>>>> <[email protected]> wrote:
>>>>>
>>>>> ...don't fully understand what they need to do? Even installing on Linux. 
>>>>> I
>>>>> have tried to build PDL from CPAN and I never get it right. The best way 
>>>>> for
>>>>> me is to do the apt-get install pdl and then go into cpan and do the 
>>>>> upgrade
>>>>> /PDL/. All of the dependencies are accounted when I do it this way. When I
>>>>> install from cpan I always end up missing dependencies.
>>>>
>>>> Hi Cliff-
>>>>
>>>> Yes, the large number of external dependencies
>>>> used by PDL to provide "full" functionality is an
>>>> ongoing problem.
>>>>
>>>> The simplest way to install PDL on linux-ish
>>>> systems having some sort of package manager
>>>> is to use that to install PDL which should pull
>>>> in packages for the needed external library
>>>> and program dependencies.  Then use CPAN
>>>> to upgrade to the latest PDL.
>>>>
>>>> There has already been discussion and decision
>>>> on using the Alien module approach to address
>>>> this problem---just no one volunteering to write
>>>> the needed Alien::NetPBM, Alien::PROJ4,...
>>>> modules.
>>>>
>>>>> So am I the minority of target users (people who use PDL and don't have a
>>>>> clue at how to create their own PP modules) in not being able to get it to
>>>>> install from cpan directly? When I use R/scilab/matlab - I use modules 
>>>>> when
>>>>> ever I can because it would take me weeks more time to develop a script. 
>>>>> On
>>>>> the other hand if your target user is a C++/Perl expert, then please
>>>>> disregard the above as it doesn't apply.
>>>>
>>>> I don't think PDL has a chance of surviving as an
>>>> active and growing project unless we can make
>>>> it possible for anyone to "just use" PDL.  Without
>>>> a 1-click install, easy interoperability, and good
>>>> documentation all the non-C/Perl programmer
>>>> scientists will be using Matlab/Octave, or NumPy,
>>>> or ... instead.
>>>>
>>>> Cheers,
>>>> Chris
>>>>
>>>> _______________________________________________
>>>> Perldl mailing list
>>>> [email protected]
>>>> http://mailman.jach.hawaii.edu/mailman/listinfo/perldl

_______________________________________________
Perldl mailing list
[email protected]
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl

Reply via email to