Re: [AMBER-Developers] namelist

From: Scott Brozell <sbrozell.rci.rutgers.edu>
Date: Wed, 17 Mar 2010 23:36:55 -0400

Hi,

On Wed, Mar 17, 2010 at 11:08:36PM -0400, Robert Duke wrote:
> Thanks. Yeah, there are better ways to do these things than namelists.
> You would probably need to support some sort of backward capability to read
> old namelist-based mdin's though, or at least that would be considerate of
> your installed user base. I guess a tool to convert an old mdin to an new
> mdin might be the best approach.

We decided to have the ciml parser create old style input files:
this is an easy path to complete backward compat. and expert mode.
This was also an easy path software dev wise because Columbus
has dozens of programs.

In fact, i made a fledging start at Amber Input Markup Language (AIML)
a few years back under GridChem funding; I mentioned this at an amber
dev meeting. It was mostly intended to satisfy the then single input file
requirement of GridChem. As such an aiml input file is just a wrapper
of amber input files; my plan was to broaden this along the lines
of CIML, but as usual i was spread to thin (and no one else seemed
interested). Here's an example:

<aiml>

  <program>
pmemd
  </program>

  <mdin> <![CDATA[
MD in vacuo: nonperiodic, dielectric = 1, 12A cut off, 300K Langevin, 100ps
 &cntrl
  imin = 0, ntb = 0,
  igb = 0, ntpr = 100, ntwx = 100,
  ntt = 3, gamma_ln = 1.0,
  tempi = 300.0, temp0 = 300.0
  nstlim = 100000, dt = 0.001,
  cut = 12.0
 /
    ]]>
  </mdin>

  <inpcrd>
TP3
     3
   0.0000000 0.0000000 0.0000000 0.9572000 0.0000000 0.0000000
  -0.2399880 0.9266270 0.0000000
  </inpcrd>

  <prmtop>
%VERSION VERSION_STAMP = V0001.000 DATE = 08/25/06 10:38:44
%FLAG TITLE
%FORMAT(20a4)
...
  </prmtop>
</aiml>

scott

> Regards - Bob
> ----- Original Message -----
> From: "Scott Brozell" <sbrozell.rci.rutgers.edu>
> To: "AMBER Developers Mailing List" <amber-developers.ambermd.org>
> Sent: Wednesday, March 17, 2010 10:45 PM
> Subject: Re: [AMBER-Developers] namelist
>
>
> >Hi,
> >
> >On Wed, Mar 17, 2010 at 05:54:27PM -0400, Robert Duke wrote:
> >>have any specs past 1995. This IS a really stupid feature of fortran, at
> >>least through 1995, but it is not like it is the only stupid feature of
> >>fortran :-) I would be overjoyed if someone with experience with 2003
> >>could say wrong, wrong, wrong. Scott? Joe Krahn? We could write our
> >>own
> >>code, but it would involve parsing namelists I would think...
> >
> >Right, right, right, as far as i know:
> >The New Features of Fortran 2003
> >www.fortranplus.co.uk/resources/john_reid_new_2003.pdf
> >
> >On Wed, Mar 17, 2010 at 07:45:40PM -0400, Volodymyr Babin wrote:
> >>you may also try using ncsu-parser to parse the input instead of
> >>namelist; it parses trees of key-value pairs and provides a
> >>way to know whether some variable (either a key-value pair, or
> >>a section containing key-value pairs) is present or not; OTOH,
> >>the invocation is somewhat more verbose then for a namelist
> >>(see one of the ncsu-*-hooks.f for an example)
> >
> >this is the way to go in general imo.
> >your input is more readable than typical amber, eg, ./abmd_FLOODING/mdin
> >
> >However, IF a serious effort was to be undertaken then in addition to
> >supporting expert and novice modes, we could go whole hog and create a
> >whole new input structure which handles a complete project and is based on
> >XML.
> >
> >Please redirect your OOhhh my its computer science comments
> >directly to /dev/null.
> >
> >The Columbus quan chem package has been down this road
> >and created a readable, single input file, completely backward
> >compatible Columbus Input Markup Language.
> >I bet even u can read it:
> >
> ><ciml>
> >
> ><molecule>
> > <description>
> >CH2 (4s5p1d/5s1p)->(2s2p1d/3s1p) (1.09A, 80). Christiansen ecp. From
> >argos.txt.
> > </description>
> > <symmetry> C2v </symmetry>
> > <geometry>
> > <units> bohr </units>
> >C 0.0 0.0 0.0
> >H 0.0 1.32401540 1.57790010
> >H 0.0 -1.32401540 1.57790010
> > </geometry>
> ></molecule>
> >
> ><basis>
> >...

_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Mar 17 2010 - 21:00:02 PDT
Custom Search