add nrexcl to mapping file. about votca HOT 9 OPEN

By the way, does this (nrexcl "issue") imply that in the case of "dlp" 
(DL_POLY) analysis the exclusions cover only nrexcl=1 ?

If so, I will have to do deal with that too!

AB

... Although I have not noticed any effects of that in the tutorial hexane case.

As I said earlier, this is not an issue! nrexcl=1 after the mapping is by 
design to leave the full control over the exclusion to the user.

dlpoly topologies behave exactly like all other topologies. Exclusion in the 
topology are honored if and only if the topology is not mapped using a mapping 
file (even dummy 1:1 mapping).

In the hexane case, there are 2 bonds and 1 angle per molecule, which 
effectively excludes all intra-molecular contributions to the non-bonded 
distributions.

> nrexcl=1 after the mapping is by design to leave the full control over the 
exclusion to the user.

- meaning ??? What is the user supposed to do about it, when csg_stat is used 
within any VOTCA iteration? Do you imply the user has to substitute csg_stat 
with his own analysis utility, or what?

The only way (that I know) the user can control the exclusion list is via 
nrexcl in .top/.tpr files in the case of Gromacs, for which one has to omit the 
mapping. 

DL_POLY has no equivalent parameter, as it creates the exclusion list based on 
the collective information in the FIELD file. Up to now (before I encountered 
this confusing "feature") I assumed that in the case of ".dlpf" option VOTCA 
creates exclusions for all bonded pairs within bonds, angles and dihedrals, if 
any (DL_POLY does essentially that).

For the purely non-bonded iterations the original gromacs/dl_poly topology is 
used and hence all exclusions are honored. Only for bonded iterations a 1:1 
mapping has to be performed and exclusions are overwritten with the ones 
generated by the mapping file(s). These exclusions are similar to nrexcl=1, 
meaning VOTCA excludes all atoms interacting with a bond, angle or dihedral. 
So, from what you described above, dl_poly uses nrexcl=1, too.

If the user needs more complicated exclusions for the distribution calculation 
there are a couple of options.
In the case of purely non-bonded iterations one can specify a special topology 
file (cg.inverse.$sim_prog.rdf.topol) when calling csg_stat. For bonded 
iterations one can put extra ("phantom") bonded into the mapping file to 
exclude more atoms. There might be cases where phantom bonds are not 
sufficient, especially something like nrexcl=0, and hence nrexcl in the mapping 
would be a nice feature to have. However one can always overload the 
distribution calculation with a custom script, e.g. running g_rdf (see 
calc_rdf_gromacs.sh). 

Bonded iteration is fairly new feature and it still needs some work.

Alright, thanks for the hints!

> So, from what you described above, dl_poly uses nrexcl=1, too.

- Nope! One can say that DL_POLY uses nrexcl=1 for bonds (if any), 2 for angles 
(if any), and 3 for dihedrals (if any). So it is a more complex algorithm for 
exclusions, where one can easily have a combination of nrexcl values within the 
same molecule (topology). 

I did not touch this complication before, because if the analysis in VOTCA 
iteration is always done with csg_stat (and any other VOTCA tools), then 
everything is updated consistently, even though exclusion lists might differ 
between VOTCA and DL_POLY (assuming VOTCA's single nrexcl >= DL_POLY's largest 
nrexcl).

BTW, in the case of the analysis for non-mapped full-atom hexane system (using 
the original trajectory), VOTCA and DL_POLY produce coinciding RDFs and other 
distributions! 

This was another reason for me not bothering about any difference in the 
exclusion lists between the two.

Open a separate issue for that problem (or re-open issue votca/csg#161) as the 
implementation of this new feature has very little to do with your specific 
issue.