name of "retrieve_hoppings"
Closed this issue · 8 comments
What about 'retrieve_tbmodel' or anything else? the idea you retrieve the tight binding model...
How about retrieve_tb, to be consistent with the Wannier90 option name write_tb?
Just to be clear: The option should retrieve not only the _hr.dat file, but also the other files generated with write_tb, correct? Otherwise I'd be in favor of keeping the name, or changing it to retrieve_hr. I think it makes sense to be consistent with the Wannier90 write_* options.
Actually seedname_tb.dat is a brand new output file. So maybe we change retrieve_hoppings to retrieve_hr or retrieve_hamiltonian or we leave it as it is, and later implement write_tb that retrieves only _tb.dat. In other words, I agree we need to be consistent with Wannier90.
Yes, that makes sense 👍
Looking at this again, I wonder why we need an option to disable retrieving these files. Wannier90 already has options to configure if the files are written, so if they are written I think they should also be retrieved.
That is, I think all output files which aren't parsed specifically should just be put in the retrieve_list unconditionally at https://github.com/aiidateam/aiida-wannier90/blob/develop/aiida_wannier90/calculations.py#L328.
Related issue: #69.
Tagging @giovannipizzi for comment.
I see your logic, indeed one could just disable the writing of the Hamiltonian in the W90 input.
Originally, we introduced that flag because the Hamiltonian is a relatively heavy file so there are a number of cases where you just don't want to store that in the repository (e.g. high-throughput).
Now I am thinking at possible use cases where the Hamiltonian is passed to another code (e.g. tbmodels, postw90) for postprocessing: there you may avoid to store the Hamiltonian in the repo and, at the same, you need to write it on the remote directory.
Does it make sense?
Anyhow, we can also set the default to be "parse all you find" and leave an optional flag to disable parsing of selected files...
Hmm yeah that could be a use case for not retrieving them. In the end this only works if the plugin for the using code also supports remote input.. which in the case of aiida-tbmodels I haven't implemented 😅
But I agree that we want to at least enable this from the side of aiida-wannier90, so maybe the question is just if it should default to retrieving or not retrieving.
Fixed in #79 - now everything known is retrieved (but it is possible to say in the settings what you don't want), since anyway if the file is being generated, it's because it's asked for in the input of Wannier90