aiida-cp2k cli

[dev-zero]

$ aiida-cp2k data structure --help
Usage: aiida-cp2k data structure [OPTIONS] COMMAND [ARGS]...

  Commands to create and inspect `StructureData` nodes from CP2K input.

Options:
  -h, --help  Show this message and exit.

Commands:
  import  Import a `StructureData` from a CP2K input file.

$ aiida-cp2k data structure import -n ../../cp2k/cp2k/H2.inp 
Success: parsed structure with formula H2

[dev-zero]

@yakutovicha what do you think?

[yakutovicha]

Thanks, @dev-zero. Looks very good.

The only thing that I would like to change is the workflow submission part. Currently, you added there a cp2k input parsing, which (I believe) shouldn't be there.

What I want to propose instead is to add the data/input_dict command that would create an input Dict object. Then, when launching work chains I would rather prefer to deal with the existing AiiDA objects.

[yakutovicha]

Please add the number of CPUs per node parameter.

[dev-zero]

Thanks for the review.

The only thing that I would like to change is the workflow submission part. Currently, you added there a cp2k input parsing, which (I believe) shouldn't be there.

What I want to propose instead is to add the data/input_dict command that would create an input Dict object. Then, when launching work chains I would rather prefer to deal with the existing AiiDA objects.

Just to make sure I understand it, you mean that the user will do a 2-step process? Something like:

$ aiida-cp2k data input_dict "/path/to/my/cp2k.inp"
Successfully generated the CP2K Parameter<6789>
Successfully created StructureData<6790>
Successfully imported BasisSet<6791>
...
$ aiida-cp2k workflow base cp2k@myhost -p 6789 -s 6790

Or that the import part should be factored out into a separate function?

[yakutovicha]

Just to make sure I understand it, you mean that the user will do a 2-step process?

Yes. I also wanted to suggest parsing different objects separately:

$ aiida-cp2k data input_dict "/path/to/my/cp2k.inp"
Successfully generated the CP2K Parameter<6789>

$ aiida-cp2k data structure "/path/to/my/cp2k.inp"
Successfully created StructureData<6790>

$ aiida-cp2k data basisset  "/path/to/my/cp2k.inp"
Successfully imported BasisSet<6791>

This would allow importing objects from the other file types: CIF, XYZ for structure, JSON, YAML for input dict, etc.

The submission part looks ok to me.

[dev-zero]

Just to make sure I understand it, you mean that the user will do a 2-step process?

Yes. I also wanted to suggest parsing different objects separately:

$ aiida-cp2k data input_dict "/path/to/my/cp2k.inp"
Successfully generated the CP2K Parameter<6789>

$ aiida-cp2k data structure "/path/to/my/cp2k.inp"
Successfully created StructureData<6790>

$ aiida-cp2k data basisset  "/path/to/my/cp2k.inp"
Successfully imported BasisSet<6791>

This would allow importing objects from the other file types: CIF, XYZ for structure, JSON, YAML for input dict, etc.

Ok, will do. But it would be still nice to have a single command to simply run any given CP2K input via AiiDA with AiiDA taking care of all the rest.

About the import interface:

• should we do like aiida-cp2k data (structure|basisset|...) (import|list|dump) <file> or if we are only doing import aiida-cp2k data import (structure|basisset|...) <file> resp. aiida-cp2k import (structure|basisset|...) <file>?
• should the import be a calcfunction in the sense that a singlefiledata is created and a calcfunction generates the resulting structuredata/parameterdata nodes? There was a similar discussion for the extended structuredata plans if I remember correctly.

[yakutovicha]

But it would be still nice to have a single command to simply run any given CP2K input via AiiDA with AiiDA taking care of all the rest.

I agree

should the import be a calcfunction in the sense that a singlefiledata is created and a calcfunction generates the resulting structuredata/parameterdata nodes? There was a similar discussion for the extended structuredata plans if I remember correctly.

I would say yes.

Btw, is your parsing tool capable of reading files that are referenced in the CP2K input (like structure, basis sets)?

[yakutovicha]

• should we do like aiida-cp2k data (structure|basisset|...) (import|list|dump) <file> or if we are only doing import aiida-cp2k data import (structure|basisset|...) <file> resp. aiida-cp2k import (structure|basisset|...) <file>?

I would say that aiida-cp2k data (structure|basisset|...) (import|list|dump) <file> is too much to ask from the cp2k plugin. The data objects we use are quite general. It would be better to implement listing and dumping directly in AiiDA: verdi data dict list.

aiida-cp2k import (structure|basisset|...) <file> sounds good enough to me. Maybe adding a file format flag would be useful.

We can also ask ping users, to hear their opinion.

[dev-zero]

should the import be a calcfunction in the sense that a singlefiledata is created and a calcfunction generates the resulting structuredata/parameterdata nodes? There was a similar discussion for the extended structuredata plans if I remember correctly.

I would say yes.

Hmm, ok, that makes it a bit more involved.

Btw, is your parsing tool capable of reading files that are referenced in the CP2K input (like structure, basis sets)?

It can correctly parse &COORD sections, returning the coordinates including unit normalization (but no rescaling for SCALED):
https://github.com/cp2k/cp2k-input-tools/blob/561926485a51b5d62319a0e333e1929ef3de0db7/cp2k_input_tools/parser.py#L229-L234

&CELL, and all *_FILE references are simply returned as such (no further check whether they exist). Embedded &BASIS_SET or &PSEUDOPOTENTIAL are returned as list under the * key. All CP2K preprocessor instructions are recognized (e.g. @INCLUDE are resolved).

[yakutovicha]

@dev-zero I see you posted the same comment twice. Is my input still needed? Or will you go ahead implementing the changes that we discussed?

[yakutovicha]

@dev-zero, please let me know if you need any help here. I can try to take over some of the parts.