pdb_mkensemble includes the PDB filename when using a redirect

[amjjbonvin]

pdb_mkensemble includes the PDB filename when using a redirect

[JoaoRodrigues]

I'm not sure that is a "bug" on our end, and I'm not even sure of how to catch it. We don't know that we are writing to a file, we're writing to stdout and that's being redirected to a file. That's outside of Python's control and therefore pdb-tools. See: https://unix.stackexchange.com/questions/257110/exclude-target-redirection-file-from-being-processed-in-for-loop

I'm afraid that short of adding an output flag (which none of the other tools has), the only option is for the user to write the ensemble to a different folder, or be more specific with the glob: pdb_mkensemble model*.pdb > ensemble.pdb. Ultimately, we could add a check to only write files that have content, but that can hide a bunch of errors.

[amjjbonvin]

Ok - but then why does pdb_merge does not show the same behaviour then?
Or may-be it does but since the file is non-existing (or empty) you don't see it in the resulting pdb file.

And I was not suggesting to add an output flag.

We leave it as is then, but this can give strange behaviours and lead to errors with other software using those ensembles.

[joaomcteixeira]

When I first read this thread, I had the same opinion as João. But looking to your explanations and the code in more detail, it seems to be a bug. Give me some minutes, and I will come back with more.

edited: see next comment

[joaomcteixeira]

Hi @amjjbonvin

I can't reproduce that behaviour. It works well in my case. For example:

pdb_fecth 2fej > 2fej.pdb
pdb_splitmodel 2fej.pdb
rm 2fej.pdb
pdb_mkensemble *.pdb > test.pdb

And the test.pdb is clean.

Check if you are running on the latest version:

$ pip show pdb-tools
Name: pdb-tools
Version: 2.5.0

In similar cases, I use a tmp file to bypass the output, for example:

pdb_mkensemble *.pdb > tmp && mv tmp test.pdb

Cheers,

[amjjbonvin]

Ok - must be a strange behaviour on my Mac then.

If running under tcsh (I know, it's dinosaur stuff...) I get the weird behaviour. When running under bash it works fine... In both cases the same Python3 ((3.7.2) from the system are called.

[joaomcteixeira]

Okay. So it's precisely as João said at first, it's a problem how the tcsh pass the arguments to the python sys.args. In bash the args are block to the > but likely in tcsh the whole command line is sent. If you see this can be a real issue because tcsh is used often in servers where pdb-tools may run, we can make a checkpoint in the python part. But that will bring a non-negligible burden to the code.