[gmx-developers] Thought on gmx distance

Teemu Murtola teemu.murtola at gmail.com
Thu Mar 13 05:41:54 CET 2014


thanks for the feedback, that is much valued. Since I no longer actively
use Gromacs for any research, but am more in a maintenance mode for the
code, there can be some rough edges related to changes from 4.5/4.6 to the
new way of working in 5.0. Most of the behavior should match g_select in
4.5/4.6, though.

On Wed, Mar 12, 2014 at 3:20 PM, Justin Lemkul <jalemkul at vt.edu> wrote:

> I thought I would share a bit of feedback on the new implementation of
> g_dist. First off, I love the flexibility in the selections now, so that's
> great.  The only problem is that gmx distance does not behave like the
> other tools. Specifically, no list of available index groups is printed
> (default or otherwise, when using -n index.ndx).  I found that simply by
> pressing Enter when prompted for a selection, I can get a list of available
> groups, but that's not documented anywhere that I could find.  It was sort
> of a random guess when I stared at the screen and said, "where are my
> groups?"

The rationale for not printing them by default was that with the
selections, it should be much rarer to actually need to use the index
groups. So in order to keep the output concise, they were printed only when
explicitly requested. As they are now, many of the default index groups
generate a lot of noise without providing much value: any group based on
splitting by residue names only can be easily expressed as "resname FOO".
For selecting a protein or such, there can be some value, at least until
the selections would provide convenient keywords for these. There's an open
issue in Redmine about how the default groups should behave with
selections: <http://redmine.gromacs.org/issues/921>

Nevertheless, I now added that information into the prompt in <
https://gerrit.gromacs.org/#/c/3262/>, since I agree that it may lower the
barrier to use the tool. If the user doesn't care about the selections,
they can now go as if they didn't exist. ;) The behavior was documented in
'help syntax', but that may not have been very easy to find. I thought of
only printing them if a -n option was explicitly provided to address my
concern from above, but didn't judge that worth the extra effort of passing
that information around.

So the suggestion I have is that gmx distance should act like the other
> tools and print a list of available groups.  That way, we don't force the
> user to have to quit the program to refer to the index file or open up a
> new terminal and then move back and forth.

There's no need to exit the program or open a new terminal, since the
information was still accessible. ;)

Best regards,
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20140313/287dcc8d/attachment.html>

More information about the gromacs.org_gmx-developers mailing list