For your restart file problem, make sure you are writing
and reading the restart file (the latter via restart2data) with
the most current (fully patched) LAMMPS.
For your dump problem, I'm not clear. Are you saying
some VMD reader doesn't read the newer LAMMPS format
that puts column labels on the ITEM: ATOMS line?
If so, can some VMD guru comment?
Steve
2009/4/22 Fengchao Wang <[email protected]...>:
At least for the released version of VMD (1.8.6) this is probably the case.
The lammps plugin tries to locate the ATOMS record by evaluating
if (0 == strcmp(line,"ITEM: ATOMS"))
which will be false if the line contains additional characters.
1.8.6 was released in April 2007. I don't know if this behavior has changed
in the development version of VMD.
-- Lutz
If someone changed strcmp() to strstr() then it would
work by ignoring the trailing text. Anyone know how
to get that change into VMD proper or your own local
version?
Steve
If someone changed strcmp() to strstr() then it would
work by ignoring the trailing text. Anyone know how
to get that change into VMD proper or your own local
version?
steve, lutz,
please see attached the latest version of the lammps plugin
for VMD and the log of changes since the last release.
if you need some additional changes being made, please let
me know ASAP. vmd-1.8.7 is going to be released in the near
future and i can integrate changes quickly myself now.
as far as a i can read the parser, it should not be affected
by the problem mentioned. if there are still files that
don't get parsed properly, please send me and example and
i'll try to make it work.
cheers,
axel.
lammpsplugin.c (14.8 KB)
lammpsplugin.log (1.68 KB)
Axel - from my perspective the only 2 enhancements needed are:
1) match on the ITEM lines with just ITEM: keyword, ignoring any trailing stuff
2) for the 3 lines of box size, sometimes (xlo xhi) will be (xlo xhi
xy) and simliarly
for all 3 lines (ylo yhi xz), (zlo zhi, yz), if the user is
outputting a triclinic box
If VMD doesn't display triclinic boxes any different, then it could just read
the first two fields, as it does now.
Steve
Axel - from my perspective the only 2 enhancements needed are:
1) match on the ITEM lines with just ITEM: keyword, ignoring any trailing stuff
i put something along those lines already into the VMD cvs.
but with the additional labels, VMD can extract and store
additional information from a trajectory files. so i'm fixing
to implement that as well. also lutz reminded me of one problematic
assumption that the current plugin makes and we've discussed
a way to handle this. if all goes well, there will be a much
improved lammps plugin for VMD by tonight, and that one would
also be able to handle "custom" style dump files. a future
version of VMD will allow file readers to pass any information
to VMD (into per timestep "user" fields) and the lammps plugin
will then be one of the first to make use of it, at least this
is my current plan.
2) for the 3 lines of box size, sometimes (xlo xhi) will be (xlo xhi
xy) and simliarly
for all 3 lines (ylo yhi xz), (zlo zhi, yz), if the user is
outputting a triclinic box
If VMD doesn't display triclinic boxes any different, then it could just read
the first two fields, as it does now.
VMD handles non-orthogonal boxes in a not very satisfactory way,
due to its original design around .psf/.dcd files (which didn't
support non-orthogonal boxes for a long time).
in general this is not so much of a problem, unless you want
periodic display, and then you'd have to rotate the whole
coordinate set so that the three cell vectors follow the internal
conventions. before going through those hoops (i've done it before),
i'd rather wait until VMD has a better internal treatment of
non-orthogonal periodic data. there are a bunch of requests
of that kind in the TODO list and thus it is likely that
improvements will happen.
cheers,
axel.
all sounds good to me - thanks for working on this
please post an announce to the LAMMPS mail list
whens something new for LAMMPS is avialable in VMD,
I think people will want to hear about it
Steve