Revisiting the structure of the code

[MIraola]

This is a PR which implements several modifications to the code. The code has been tested for all DFT codes and the compatibility with banduppy has been checked. Of course, we can never discard completely the potential existence of bugs.

Summary of changes

• Simplify the structure of the code
• Add tutorials for VASP, QE and Abinit
• Add automatic tests (via pytest) for QE and for the separation of bands in terms of symmetry eigenvalues.

Simplification of the structure

• The code is now separated clearly into 3 steps: parsing, processing, outputting.

• Parsing: The routines for parsing are in the readfiles.py file and each DFT code has its own parsing class. All parsing class have a similar structure: each method parses a different input data file.

• Processing: The Kpoint class is made general, i.e. independent of the DFT code.

• Outputting: methods dedicated exclusively to printing data and to constructing JSON dictionaries with the data have been created. The formatting of some data has been improved.

To do:

• [ ] Revisit docstrings
• [ ] BandStructure should inherit SpaceGroup
• [ ] Add a flag to delete wave functions if isymsep, wcc and zak are False.
• [ ] Revisit zak and wcc. Currently, we do not consider them stable functions!

[stepan-tsirkin]

Hi @MIraola

I would suggest you fix all 'F' problems that flake8 detects. Mostly it is just removing unnecessary imports, but one should be careful, sometimes it can signal a bug, for example if a variable is assigned and never used, there might be a typo in that variable, or you forgot to do something that should have been done.

See here:

stepan@stepan-desktop:~/github/irrep/irrep$ flake8 --select=F
./bandstructure.py:25:1: F401 '.utility.str2bool' imported but unused
./bandstructure.py:25:1: F401 '.utility.BOHR' imported but unused
./bandstructure.py:25:1: F401 '.utility.split' imported but unused
./bandstructure.py:26:1: F401 '.readfiles.Hartree_eV' imported but unused
./bandstructure.py:27:1: F401 '.readfiles.WAVECARFILE' imported but unused
./bandstructure.py:191:21: F841 local variable 'IBstartE' is assigned to but never used
./bandstructure.py:700:9: F841 local variable 'nmax' is assigned to but never used
./cli.py:22:1: F401 'sys' imported but unused
./cli.py:24:1: F401 'datetime' imported but unused
./cli.py:25:1: F401 'math' imported but unused
./cli.py:30:1: F401 '.spacegroup.SpaceGroup' imported but unused
./cli.py:32:1: F401 '.utility.str2bool' imported but unused
./cli.py:336:9: F841 local variable 'preline' is assigned to but never used
./cli.py:408:17: F841 local variable 'plotfile' is assigned to but never used
./gvectors.py:232:5: F841 local variable 'npw1' is assigned to but never used
./gvectors.py:321:21: F523 '...'.format(...) has unused arguments at position(s): 0
./gvectors.py:321:21: F524 '...'.format(...) is missing argument(s) for placeholder(s): other
./gvectors.py:322:29: F821 undefined name 'other'
./kpoint.py:22:1: F401 '.gvectors.calc_gvectors' imported but unused
./kpoint.py:22:1: F401 '.gvectors.sortIG' imported but unused
./kpoint.py:23:1: F401 '.readfiles.Hartree_eV' imported but unused
./kpoint.py:24:1: F401 '.readfiles.record_abinit' imported but unused
./kpoint.py:26:1: F401 'scipy.io.FortranFile as FF' imported but unused
./kpoint.py:161:17: F841 local variable 'err' is assigned to but never used
./kpoint.py:764:75: F821 undefined name 'd'
./kpoint.py:783:29: F821 undefined name 'preline'
./readfiles.py:327:24: F841 local variable 'occ' is assigned to but never used
./readfiles.py:373:9: F841 local variable 'title' is assigned to but never used
./readfiles.py:468:9: F841 local variable 'err' is assigned to but never used
./readfiles.py:487:9: F841 local variable 'nat' is assigned to but never used
./readfiles.py:604:21: F841 local variable 'j' is assigned to but never used
./spacegroup.py:561:9: F841 local variable 'nsym' is assigned to but never used
./spacegroup.py:690:13: F841 local variable 'write_ref' is assigned to but never used
./spacegroup.py:876:9: F841 local variable 'nmax' is assigned to but never used
./spacegroup.py:936:21: F841 local variable 'err' is assigned to but never used
./spacegroup.py:1037:13: F841 local variable 'found' is assigned to but never used
./spacegroup.py:1159:9: F841 local variable 'errtxt' is assigned to but never used
./tests/test_abinit_scalar.py:4:1: F401 'monty.serialization.dumpfn' imported but unused
./tests/test_abinit_spinor.py:4:1: F401 'monty.serialization.dumpfn' imported but unused
./tests/test_espresso_spinor.py:4:1: F401 'monty.serialization.dumpfn' imported but unused
./tests/test_refUC.py:4:1: F401 'monty.serialization.dumpfn' imported but unused
./tests/test_symsep.py:4:1: F401 'monty.serialization.dumpfn' imported but unused
./tests/test_vasp_scalar.py:4:1: F401 'monty.serialization.dumpfn' imported but unused
./tests/test_vasp_spinor.py:4:1: F401 'monty.serialization.dumpfn' imported but unused
./tests/test_wannier_scalar.py:4:1: F401 'monty.serialization.dumpfn' imported but unused
./tests/test_wannier_spinor.py:4:1: F401 'monty.serialization.dumpfn' imported but unused

[MIraola]

Hi @stepan-tsirkin ,

thank you for the recommendation. I have fixed the F errors of flake8 as you suggested. There is an error left which is in the WCC routines; I am pretty sure that removing it will do nothing, but I leave it since I try not to do changes to this routines until we fully test them.

The flake8 is really helpful and it works very well! I will use it from now on.

[MIraola]

I have...

• Implemented the possibility of removing wave functions from memory via the argument save_wf of BandStructure and Kpoint. By default: (1) in the CLI, WFs will be saved only if isymsep, ZAK or WCC were specified, removed otherwise. (2) If BandStructure or Kpoint are called in a Jupyter-notebook, WFs will be kept in memory.
• Updated docstrings
• Implemented the calculation of number of band inversions only if traces were calculated first (calculate_traces is True)

[MIraola]

Done, thanks!

I agree, we can address the remaining issues in smaller and more dedicated PRs.