How to wrap Executable line into a loop

General Usage
, ,
1

Hey :wave:

I have a small tool that handles cube files. Among other parameters, the tool expects the input name and output name of the file:

cubehandler shrink [options] input_file output_file

I am preparing a plugin for it to post-process the cube files created in a previous QE/CP2K/… calculation.

To be able to analyse multiple cube files in one go, I would like to wrap the execution in a bash loop:

for file in <files>; do
    cubehandler shrink [options] ${file} out_${file}
done

My only option now is to put for file in <files>; do into mpirun_extra_params and done in the cmdline parameters of the code, but this is ugly. Are there any better options?

Thank you very much.

Best,
Sasha

2

Hi Sasha, with the CalcJob options available in AiiDA there is not much better you can do than the hack you sketched out. One alternative I see is to instead of calling cubehandler as the “code” directly, to write bash script that contains the loop itself and call that as the code:

#!/bin/bash

options="$@"
for file in input*.cube; do
    cubehandler shrink "$options" ${file} out_${file}
done

Now to easily run this, I would recommend using aiida-shell.
Simply store that bash script on the remote computer, put it in the PATH and make it executable. Then you can run it with something like:

from aiida_shell import launch_shell_job
results, node = launch_shell_job(
    'cube_shrinker.sh',
    arguments='-s some-flag --quick -O option',
    nodes={
        'cube_1': SinglefileData(..., filename='input_1.cube'),
        'cube_2': SinglefileData(..., filename='input_2.cube'),
    },
    output=['out_*.cube'],
    metadata={
        'options': {
            'computer': load_computer('target-computer'),
        }
    }
)

I haven’t tried this, but I think it should be close to working. No dedicated plugin even necessary.

3

In addition to @sphuber’s comment: since you are writing the cubehandler script, you can just have the option to specify more than one input filename in the cmdline (or maybe a folder, where you process all .cube files), and just a folder for all the outputs. So you can just do a single call of the code, and then retrieve all files with some wildcard, e.g. outputs/.cube.
Also in this case of course you could still use aiida-shell directly (possibly combined with a custom parser if needed to convert .cube into some AiiDA-specific format?0