[Bioperl-l] URI, Graph, and Module::Build

[Chris Fields]

On Nov 29, 2006, at 10:04 AM, Sendu Bala wrote:

> Chris Fields wrote:
>> Sendu,
>> For the Build.PL dependencies, do we want to use the absolute  
>> module requirements (URI::Escape, Graph::Directed, XML::SAX::Base)  
>> or the distributions they are in (URI, Graph, XML::SAX)?  Or does  
>> it really make a difference?
>
> It makes a difference right now only where the absolute module  
> doesn't have a version (because in Build.PL we ask for at least  
> version 0, which fails if version in the modules is actually  
> undef). I'll fix that in the future, so it is best to use the  
> absolute module for clarity.

That sounds fine to me.  If I remember correctly, at least one  
example (XML::SAX::Base) used to be a separate distribution, but it  
should install XML::SAX now.

>> Also, you had mentioned that Bundle::Bioperl is no longer needed  
>> since you can have them installed via Build.  Does that also hold  
>> true for Windows (since CPAN is, at best, very testy installing on  
>> Windows)?
>
> In what way is it testy? I set up a new WinXP machine today and  
> installed ActiveState. My Build.PL script happily offered to  
> install Module::Build for me with CPAN and it almost worked up  
> until CPAN needed nmake. After installing nmake I upgraded CPAN  
> with no problems and then everything worked after that (I only  
> tried to install one optional dep with Build.PL via CPAN, and that  
> worked fine).

It all depends on the modules installed.  For instance, a few that  
include C code or required external compiled libraries had generally  
been much harder to install, or they were historically; I admittedly  
haven't tried them recently.  GD, (I think) DBD::mysql, and anything  
using precompiled dynamic libraries were the most problematic, but  
PPMs and Win-compatible dynamic libraries were always available  
somewhere which made installing them via nmake a non-issue.

One thing I noticed (and mentioned previously) is that using 'Build'  
also generated HTML docs on WinXP (I didn't see this on Mac OS X); I  
don't know if this still occurs, so I'll give it a run today.  This  
may be a problem since the pod2html-generated formatting differs  
significantly from the PPM-generated docs.  I'm not even sure they  
install in the same location as the PPM-generated pages or whether  
this will cause any issues.  Do you also see this on your local WinXP?

> So yes, I'd say it holds true for Windows. Though I haven't sorted  
> out the ppd stuff yet.

I think recommending PPM for BioPerl installation on Windows is still  
the best option for now, at least until we are sure about CPAN  
installation and the HTML issue above.  Frankly, there will just be  
too much demand for it just for the HTML docs alone (they are quite  
nice).  If the prereqs are listed correctly it should install those  
first, but who knows with PPM4!  It will also generate HTML on the  
fly and install it in the doc tree as expected, which I have tested.

As for a new PPD/PPM, I haven't tried generating it yet via  
Module::Build (I thought maybe Nathan was doing this).  I may give it  
a go today on WinXP with a fresh ActiveState install (most of my  
important work is now on Mac OS X, so that shouldn't interfere).  My  
guess is the PPD will be in old PPM3 XML format.

Christopher Fields
Postdoctoral Researcher
Lab of Dr. Robert Switzer
Dept of Biochemistry
University of Illinois Urbana-Champaign