[gmx-developers] analysis library

[David van der Spoel]

On 2011-04-08 01.02, Francesco Oteri wrote:
> Dear gromacs developers,
> I usually hack some gromacs analysis code and I found this very
> difficult because of the lacking of modularity.
> In fact in the gromacs website
> (http://www.gromacs.org/Developer_Zone/Roadmap) is stated that should be
> a big improvement doing this modularization.
>
> For this reason I propose myself as modularizing-man for gromacs library.
> Obviously, I will do it in the free time and giving the priority to the
> tools I usually use.
>
> In particular, I am starting using the existent code (version 4.5.4)
> with the aim to:
>
> 1. implementing Multi I/O (reading in many files, trjconv -f *.xtc -o
> out.xtc)
> 2. Output from all analysis tools also in an human readable style
> (something like a matrix) and not only in xpm format
> 3. Obtaining the modularization of rmsd, rmdf, gyration radius,
> clustering, secondary structure, hbond and saltbr computation, routine
> implemented in g_analyze.
>
> I will use C++ because I found it suitable for such a project.
> Any suggestion prior to start?
>
> Francesco
>
Hi Francesco, we definitely could use some help! Right now we are 
discussing how to implement fundamental parts of gromacs, mdrun is of 
course the most important, but also the low level stuff, file I/O etc, 
which is in turn relevant for analysis tools. You may want to check out 
some of the documents in 
http://www.gromacs.org/index.php?title=Developer_Zone/C%2b%2b_Transition
which in a very abstract manner describe some of the issues. Look in 
particular under Analysis_modules.

The idea is to make a framework that reads trajectories and generates 
output of different types, and that calls an specific analysis module at 
each time step. As you probably understand this is simple for 
time-dependent properties, but for others data must be accumulated over 
a whole traj (e.g. RDF, ACF).

We also would like to avoid significant new development based on the 4.5 
tree. The head branch is C++ only, but in practice there is not much 
there yet. Please try to get an overview of the new plans and then 
report back.


-- 
David van der Spoel, Ph.D., Professor of Biology
Dept. of Cell & Molec. Biol., Uppsala University.
Box 596, 75124 Uppsala, Sweden. Phone:	+46184714205.
spoel at xray.bmc.uu.se    http://folding.bmc.uu.se