Re: [Avogadro-devel] Packmol, Symmol and other Fortran through Webapps

[Marcus D. H. <mar...@ki...>]

On Fri, May 24, 2013 at 4:49 AM, Jens <lin...@gm...> wrote:
>
> On 23 May 2013 18:15, Geoffrey Hutchison <ge...@pi...> wrote:
>>
>> I've been silent on the lists, since I'm hard at work getting tenure (the
>> package gets submitted in early August).
>>
>> But I had a brainstorm about some of the features we've discussed, like
>> Packmol -- which is a Fortran code. While we can supply binaries and detect
>> if the program is available at run-time, it might be nice to offer some of
>> these programs as web applications as well.
>>
>> Right now, we offer integration with PubChem, the NIH resolver (also for
>> compound naming) and a few other web services.
>>
>> But it'd be pretty easy to offer a web app that runs Packmol. A similar
>> Fortran program which might be useful is symmol, which detects point group
>> symmetry and can "symmetrize" molecules.
>>
>> To be clear, these programs are already available, and I'd suggest making
>> the webapp interface layer available as open source (probably BSD-license).
>> I'm just thinking that for many users and for development purposes, a webapp
>> would be useful (i.e., no need to get a Fortran compiler for Windows or
>> Mac).
>>
>> Thoughts?
>
> I think that sounds like a really good idea.
>
> I've always thought that it would be great to work towards a kind of
> eco-system of tools with Avogadro sitting in the middle and providing the
> interface to all of them and creating a really powerful modelling
> environment - something a bit like CCP4 have (http://www.ccp4.ac.uk/).
>
> There are so many helpful tools out there, but many aren't known about or
> difficult to integrate together. Avogadro could provide that glue and make
> them easy to use too.
>
> I don't know whether that would work best as a distributed software bundle,
> or as something that calls on webapps, but then there's no reason why a
> mixture of both wouldn't work.
>
One of the things we have been looking at is looser integration of
components using MoleQueue and separate processes, but more recently
we have been exploring easy ways to make and expose RESTful APIs that
allow you to query data sources or run external codes. At the core it
is just a simple RESTful API that returns results as JSON and one or
two other things. We have been experimenting with using Python as the
glue language on the server, and don't have anything worth exposing
yet but this might be one approach.

I totally agree on using a combination of the two, there are some new
developments in CMake for example that make it easier to use gfortran
from MinGW with Visual Studio for the C/C++ code.

Thanks,

Marcus