Changes to the output of the outxml_parser

[janssenhenning]

Up till now the output of the outxml_parser tries to mirror the current status of the hardcoded parser in aiida-fleur as closely as possible. Here we collect things that should be added/modified in the output. Feel free to add to this list

• The normal distance output is named charge_density. Maybe not the most telling name (alternative density_convergence for example)
• The names for the relax output are also not optimal (some are formatted containing the atomtype, maybe a list is more useful); also see fleur output node in case of relaxation aiida-fleur#101
• Add some quality criterions of calculations (lostElectrons, ...)
• Some calculation modes are missing for newer fleur versions (Hubbard1, greensfunction, ...)

[janssenhenning]

further modifications

• Remove targetStructureClass (this has been empty for a long time)
• Adjust some outputs with more fleur modes (e.g. bandgap output only in with mode='hist')

[janssenhenning]

Maybe add git describe as a more detailed description of the used version of fleur

[janssenhenning]

• number of kpoints is missing
• force_largest is in the old parser even if no force calculation is done (Is this desired behaviour for convergence checking or can this be removed ?)

[janssenhenning]

• Number of MPI/OMP should be available (can we reliably calculate core-h from this and the walltime ?)

[Tseplyaev]

@janssenhenning
The number of MPI/OMP used in the calculation is available in the attributes of a CalcJob node:

In [1]: a = load_node(55639)   #CalcJob node                                                                                                                            

In [2]: a.attributes                                                                                                                                       
Out[2]: 
{'job_id': '17021989',
 'sealed': True,
 'version': {'core': '1.4.0', 'plugin': '1.1.0'},
 'withmpi': True,
 'resources': {'num_machines': 4,
  'num_cores_per_mpiproc': 8,
  'num_mpiprocs_per_machine': 6},
  ...

There is also information on execution time and other useful data, related to submission

[Tseplyaev]

I think there should be no reason for keeping force_largest if forces were not calculated. If I am not mistaken, force_largest is used for information only and removing it should not affect functionality of other parts of the plugin.

[janssenhenning]

The number of MPI/OMP used in the calculation is available in the attributes of a CalcJob node:

Okay then the best thing is to leave this to the aiida side

[janssenhenning]

@Tseplyaev Would it be okay to convert the relax output to one list? So for example instead of

{
    'force_x_type1': 1,
    'force_y_type1': 2,
    'force_z_type1': 3,
    'force_x_type2': 4,
    'force_y_type2': 5,
    'force_z_type2': 6,
}

We write

{
    'force': [[1,2,3],
              [4,5,6]]
}

And the same for abspos. We could also call it force_atoms.

I would also like to rename force_largest to force_largest_component (Since that is what's it is :))

[Tseplyaev]

@janssenhenning
Sure, I think making a more compact form of storing the result is better. I believe the force output of the FleurCalculation output_parameters is used nowhere, hence there is no compatibilities to worry about.
The name change sounds reasonable, this is indeed the largest component.

[janssenhenning]

These changes are now implemented in masci-tools version 0.4.0-dev7