More robust seedname.wout parser

[AntimoMarrazzo]

We should use regular expression instead of split&strip to parse x y z and the spread at each iteration.

[giovannipizzi]

I think it's a bit more complex than that... the problem is that sometimes there are no spaces between coordinates!! (for long numbers starting with a minus)

What about adding a new output to W90 (if it's not already in the tb_parameters) with the centers spreads etc instead of trying to parse the wout?

[greschd]

Is there a good library to produce structured output (csv, xml, json, yaml) from Fortran? If so, I would argue for using that when creating a new output file. Otherwise we'll have to write yet another output parser..

[giovannipizzi]

Maybe things like YAML are simple enough to be written even without a library? (I hope...)

[greschd]

Or a CSV with just index, x, y, z, spread. Of course since it's per iteration it would also make sense to have a nested list, which is more suited for YAML.

[giovannipizzi]

I would consider this issue more general, and improve the names and parsing of the output.
We can then move the issue of creating a more machine-readable file in the Wannier90 repository, and when the feature is released implement its parsing from here.

[giovannipizzi]

@normarivano could you report here the table that we started discussing, with the suggested changes to the parser key names, for discussion with the others?

[normarivano]

I attach hereafter the table for the parsing of the standard output file. We propose some changes for the keys' names to be implemented within this release. Can you give me some feedback as soon as possible? Thank´s a lot.
@greschd @giovannipizzi @AntimoMarrazzo @qiaojunfeng

Keys in raw_wout_parser (i.e. seedname.wout parsing)
Suggested parsed key	Current key	Description	Where in the seedname.wout
warnings	warnings	List of warnings	General
number_wfs	number_wannier_functions	Number of Wannier functions.	MAIN
length_units	length_units	Units used to express the lengths, if not Ang we will have an additional warning in the list warnings¨.	MAIN
output_verbosity	output_verbosity	Level of verbosity of the output from 0 (low) to 3 (high). Default value is 1 and parsing is supported only in this case.	MAIN
convergence_tolerance	wannierize_convergence_tolerance	The convergence tolerance to find the final spread	WANNIERIZE
r2_mn_writeout	r2_nm_writeout	Boolean variable. If true, a file seedname.r2mn is written.	WANNIERIZE
xyz_wf_centres_writeout	xyz_wf_center_writeout	Boolean variable. If true, a file seedname_centres.xyz is written in xyz format for further visualization (jmol, vmd etc.).	WANNIERIZE
Omega_I	Omega_I	Gauge invariant spread.	Final State
Omega_D	Omega_D	Diagonal part of gauge-dependent spread.	Final State
Omega_OD	Omega_OD	Off-diagonal port of gauge-dependent spread.	Final State
Omega_total	Omega_total	Sum of the total gauge-dependent spread and gauge-invariant spread.	Final State
wf_ids	wannier_function	List of ordinal numbers identifying each Wannier function.	Final State
im_re_ratio	im_re_ratio	List of the ratio between the imaginary and real part of the each Wannier function in the same order given by wf_ids above.	Final State
wf_centres	coordinates	List of the coordiantes (x,y,z) of the centres of each Wannier function in the same order given by wf_ids and im_re_ratio above.	Final State
wf_spreads	spread	List of the spreads associated to each Wannier function in the same order given by wf_ids, im_re_ratio and wf_centres.	Final State

[normarivano]

I´m also going to open an other issue (not in Milestones) in which we can discuss additional information to parse for the future. Me and @qiaojunfeng already started discussing that.
#70

[greschd]

Looks all good to me 👍

[giovannipizzi]

Thanks a lot! Looks good to me as well. Very good also to go back to UK spelling considering this is what the W90 code uses.

Just a couple of comments of things I would change:

• r2_mn_writeout -> r2mn_writeout (there is no underscore neither in the name of the flag in the w90 input (write_r2mn), neither in the file extension)
• xyz_wf_centres_writeout -> xyz_writeout (similar reason as above, for consistency (e.g. the W90 flag is called simply write_xyz).

[AntimoMarrazzo]

All good for me - I close the issue.

[giovannipizzi]

I reopen it - let's close it when we have the PR that changes the names