[gmx-developers] Coordinate scaling in pdbio.c

[Szilárd Páll]

On Wed, Apr 11, 2012 at 1:40 PM, Mark Abraham <mark.abraham at anu.edu.au>wrote:

>
>
> On 11/04/12, *Erik Marklund *<erikm at xray.bmc.uu.se> wrote:
>
>
>  11 apr 2012 kl. 01.38 skrev Mark Abraham:
>
>  On 11/04/2012 5:21 AM, David van der Spoel wrote:
>
> On 2012-04-10 21:10, Anton Feenstra wrote:
>
>  On 09/04/12 21:47, Justin A. Lemkul wrote:
>
>
>
>  Radhakrishna Bettadapura wrote:
>
>   All,
>
>
>   I'm trying to figure out why a pdb input to pdb2gmx results in
>
>   coordinates that are scaled by a factor of 0.1. That is, if a line in
>
>   the PDB file reads
>
>
>
>   ATOM 1 N ALA B 3 -1.221 20.481 12.450 1.00 34.64 N
>
>
>   then the corresponding line in the .gro file reads
>
>
>   3ALA N 1 -0.122 2.048 1.245
>
>
>   The culprit seems to be the following lines in pdbio.c:
>
>
>   static int read_atom(t_symtab *symtab,
>
>   char line[],int type,int natom,
>
>   t_atoms *atoms,rvec x[],int chainnum,gmx_bool bChange)
>
>   {
>
>
>   // do stuff
>
>
>   x[natom][XX]=strtod(xc,NULL)*0.1; /*all coordinates scaled by 0.1...
>
>   why?*/
>
>   x[natom][YY]=strtod(yc,NULL)*0.1;
>
>   x[natom][ZZ]=strtod(zc,NULL)*0.1;
>
>
>   //do more stuff
>
>
>   }
>
>
>
>   There's also a line in the output function that multiplies all
>
>   incoming coordinates by 10 before writing it to output:
>
>
>   void write_pdbfile_indexed(FILE *out,const char *title,
>
>   t_atoms *atoms,rvec x[],
>
>   int ePBC,matrix box,char chainid,
>
>   int model_nr, atom_id nindex, atom_id index[],
>
>   gmx_conect conect, gmx_bool bTerSepChains)
>
>   {
>
>
>
>   //do stuff...
>
>
>
>   fprintf(out,pdbform,pdbtp[type],(i+1)%100000,nm,resnm,ch,resnr,
>
>   (resic == '\0') ? ' ' : resic,
>
>   10*x[i][XX],10*x[i][YY],10*x[i][ZZ],occup,bfac,atoms->atom[i].elem);
>
>   // do more stuff...
>
>
>   }
>
>
>
>   Can someone tell me why this coordinate scaling occurs? And why the
>
>   scale factor is a single hard-coded number than, say, something that
>
>   depends on the input?
>
>
>
>  PDB coordinates are in Angstrom, while .gro coordinates are in nm. Hence
>
>  the conversion factor of 0.1.
>
>
>  What happened to A2NM and NM2A?
>
>  In my (4.0.5, yes that is very old) includes/physics.h I still see:
>
>
>  #define A2NM (ANGSTROM/NANO) /* NANO */
>
>  #define NM2A (NANO/ANGSTROM) /* 10.0 */
>
>
>  Why aren't these still used for that?
>
>
>
> At some stage macros were removed from the code, even though these seem
> rather harmless. I agree that it is confusing to hard code these numbers.
> If we cannot use macro's like this we should probably replace them by
>
> static const real ANGSTROM=1e-10;
>
> static const real NANO=1e-9;
>
>
> etc.
>
>
> Comments?
>
>
> include/physics.h still has these macros. IMO, hard-coded constants are a
> greater evil than macros to prevent that, though I expect we will
> transition to const values at some stage soon.
>
>
> But are these the kind of macros we want to avoid? Aren't function-like
> macros the ones to kill in the first place?
>
>  They're taking no parameters, so they're hardly function-like. The code
> fragments above are compiled into constants by the pre-processor. A macro
> that is used as
>
> dist_in_nm = dist_in_angstrom * A2NM;
>
> is much less evil than a macro
>
> dist_in_nm = A2NM(dist_in_angstrom);
>
> and physics.h has only the first type. Inlined functions of the second
> type would be best of all, of course.
>

Agreed, in fact literal constants should be avoided as much as possible, a
macro (which has a name is always more verbose).

--
Szilárd


> Mark
>
> --
> gmx-developers mailing list
> gmx-developers at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-developers
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-developers-request at gromacs.org.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20120411/ae3d6eb2/attachment.html>