About zipping files at the end of a Job

[gpetretto]

I have always had troubles with the fact that atomate (even v1) zips all the files in a folder. This typically ends up zipping the cluster queue files, preventing the writing from queue manager of additional informations (at least this is what happens in the clusters I have used). Other files can also end up zipped even if they are not supposed to.

I propose that

1. the zipping can be disabled though an option in Atomate2Settings
2. the preferred option would be not to zip the entire folder but just the files related to the executed code. Or maybe even just those that are expected to be large.

For VASP these lists can be used: https://www.vasp.at/wiki/index.php/Category:Output_files https://www.vasp.at/wiki/index.php/Category:Input_files. For Abinit I am still not sure if the zipping is easily manageable and we will need to see in future PR. I don't know if list of files are available for CP2K and Lobster.

If there is an interest in this I can open a PR.

[JaGeo]

@gpetretto I would be certainly interested to have this option for VASP as well. It makes debugging especially hard if all files are zipped (including error files).

Within the Lobster workflow, I think we also zip everything at the moment:

atomate2/src/atomate2/lobster/jobs.py

Line 90 in c2d098a

gzip_dir(".")

Within atomate, we actually did it differently and only zipped all files related to Lobster.

[JaGeo]

@gpetretto The typical output files for Lobster are available here:

atomate2/src/atomate2/lobster/files.py

Line 15 in c2d098a

LOBSTEROUTPUT_FILES = [

Input file is only lobsterin at the moment. I think we wanted to implement it like this but I somehow did not do it.

[utf]

This was fixed in #414