Allow explicitly specifying 'ibrav'.

[greschd]

Fixes #347.

If an ibrav value other than zero is specified in the SYSTEM inputs, the cell of the input structure is converted into the appropriate A, B, C, cosAB, cosAC, cosBC values. As a check, the cell is reconstructed using qe_tools. The cells are compared element-wise, with an absolute tolerance controlled by the IBRAV_CELL_TOLERANCE setting.

Still needs testing and documentation. No need to check the correctness yet, but I'm soliciting feedback on the interface.

Note also: This adds numpy and scipy as explicit dependencies. They're implicit dependencies already, so I don't think that should be problematic.

TODO:

• add tests for the conversion between cell and parameters these will go into qe-tools directly
• document the explicit ibrav feature
• add test for immigration, possibly add conversion from parameters to cell
• move cell -> parameter conversion to qe_tools?

[greschd]

Ah, Python2, loving it 😅

[greschd]

@giovannipizzi should the "cell to parameter" conversion go into qe_tools?

One thing to consider here is that there are no checks in that function itself that the cell conforms to the specified ibrav - that is done later. If we want to move it into qe_tools and make it a publicly-available function, it should probably do some checking itself.

[giovannipizzi]

@greschd yes, maybe as they are general they can be put into qe-tools?

In this case I agree testing should be done by qe-tools, and raise there.
If we want to do it, I suggest this could be a good moment to release a v2, with py3 only support, and maybe changing a bit the interface? (I don't like so much the way you initialise things in qe-tools now, where it guesses what you want depending if the string you pass is a valid absolute path, otherwise it falls back to parse the string itself). Suggestions welcome there.

[greschd]

I suggest this could be a good moment to release a v2, with py3 only support, and maybe changing a bit the interface?

AFAIK aiida-quantumespesso still maintains py2 compatibility for now. How about we proceed with it here, then polish it for inclusion in qe_tools and replace it here once that's released and compatible and aiida-quantumespresso drops py2?

For inclusion in qe_tools, probably it should also be able to change between the celldm and A, B, ... output formats.

[sphuber]

Indeed, we first want to release aiida-quantumespresso==3.0.0 with python 2 support, because that is the first official release that is both Python 2/3 and AiiDA 1.0 compatibile. However, soon after I was planning to drop Python 2 and require aiida-core>=1.2 as I need certain features from those releases for advanced functionality. We will then only provide minimum support for 3.0.0 series and just continue with what will become 3.1 or 4.0, to be decided

[greschd]

Ok, what's your opinion on the interface of this PR, specifically the way specifying ibrav works with the single absolute tolerance on the cell?

If you're fine with this I will add tests and documentation.

[sphuber]

I don't have the time for it today, unfortunately, sorry. I am live demo'ing AiiDA in two hours and then have to prepare a proposal. Will try to look at it soon

[greschd]

No worries, it's not an urgent thing. Good luck on the live demo!

[giovannipizzi]

@greschd yeah sorry, you're right.
Then it's just a matter of moving the relevant code into qe-tools that still has py2+py3 compatibility.
Feel free to do a PR there! If you want I can also give you admin rights there so you can make releases yourself.

[giovannipizzi]

I guess we should first release qe_tools before we can finalise this, right?

[greschd]

That's the plan 😄

[greschd]

@sphuber what's the best place to put the documentation for this feature? Maybe a subsection in https://aiida-quantumespresso.readthedocs.io/en/latest/user_guide/get_started/examples/pw_tutorial.html#structure could work, but I'm not sure if there might be a better place.

[sphuber]

To be honest, I haven't yet been able to look at the documentation and it badly needs to be improved I feel. So I don't have an opinion really. Just put it wherever you feel is best for now.

[greschd]

Ok, I've added a short paragraph in docs/source/user_guide/calculation_plugins/pw.rst and docs/source/user_guide/calculation_plugins/cp.rst.

[greschd]

Rebased on the latest develop and ready to go!

[sphuber]

Thanks @greschd . Looking good. Just some minor suggestions and request to add a test.