[gmx-developers] Shall we ditch gro and g96 files?

[Erik Lindahl]

Hi,

On Mar 31, 2008, at 6:09 PM, David van der Spoel wrote:
>>>
>> I agree that we need a better format, but then we want a flexible  
>> format.
>> An example of a good format is the Espresso package format, which  
>> Gromacs
>> can read and write already. Here is an example file:
>
> OK that looks interesting. The question is then the interpretation  
> of the strings, which are still somewhat cryptic in this example.  
> For instance is "id"  an integer or a string? How is the program  
> that reads this supposed to know? Are there pre-defined strings?  
> E.g. pos seems to imply that there are three coordinates.
>> # comment ...
>> {variable {box_l 0.000000 0.000000 0.000000}}
>> {particles {id pos type q}
>>       {0 0.230000 0.628000 0.113000 0 0}
>>       {1 0.137000 0.626000 0.150000 0 0}
>>       {2 0.231000 0.589000 0.021000 0 0}
>> }
>> All fields and variables are optional and are not fixed order or  
>> format.
>> For the box I don't see why using angles would necessarily be an  
>> improvement.
>> If you want angles or box vectors depends on the application.

Writing an extremely complex file format that can hold anything is  
trivial - then we can simply go XML. The challenge, imho, is to have a  
format that is simple enough that a vanilla fortran program can read  
it without excessive efforts in string parsing.


> What do you need to allow isotropic scaling of dodecahedrons and  
> other triclinic boxes?

That's an internal problem, and should thus be handled internally in  
the code, not by requiring a new box specification. To avoid rounding  
errors the easiest option is to calculate the angles at the start of  
the run, and make sure they don't change if we use isotropic scaling.

Cheers,

Erik

------------
Erik Lindahl   <lindahl at cbr.su.se>  Backup: <erik.lindahl at gmail.com>
Assistant Professor, Computational Structural Biology
Center for Biomembrane Research, Dept. Biochemistry & Biophysics
Stockholm University, SE-106 91 Stockholm, Sweden
Tel: +46(0)8164675  Mobile: +46(0)704218767  Fax: mail a PDF instead