[gmx-developers] Re: trjconv: problem of use "-skip" with "-split"

[Igor Leontyev]

> If split is implemented exactly as described in trjconv -h, then if you 
> choose an unlucky
> value for skip, then= t mod split might not be tested at all for frames 
> for which it
> would equal the first time (which is set with -b?). It's a matter of how 
> split and skip
> interact. You may need two passes with trjconv, or to use -dt cunningly.

Before posting the question I tried to play with all these flags 
(-skip, -split, -dt, -b), but did not find correlation with manual. Thus, -b 
option essentially does not change a number of split parts but -skip does, 
although, last should affect only a number of configurations saved in each 
part but not the number of parts. See manual:"-skip n:  Only write every 
nr-th frame". What is the actual correlation between the flags of trjconv?

>Well, be glad you read your output... -ndec has a default of 3, not 
>"preserve input
> precision", for better or worse... A default of preserving the input xtc 
> precision if the
> input is in xtc format would be very reasonable, however. Perhaps -ndec 
> should have
> a default value of -1, which maintains the precision of xtc input, and 
> defaults to 0.001
> precision otherwise. That maintains backward compatibility with all 
> previous usage of
> -ndec except "erroneous" trjconv of higher-precision xtc to 
> lower-precision xtc, and
> fixes this issue.

Despite the blabbing "-ndec has a default of 3 ...Using output precision of 
0.001 (nm)" the output trajectory is actually saved with the original 
precision 0.0001 (nm).