beddalumia / qcmp-lab Goto Github PK

Given the proprietary nature of MATLAB we definitely want full or at least more-than-decent compatibility with the free and open source alternative provided by the GNU project: Octave.

I don't really know how hard the task would be, since the project relies quite heavily on IO / string-manipulation capabilities of MATLAB, that being somehow out of its core scope, might not be fully covered by GNU Octave (or, worse, have not really compatible implementations). We'll see.

For sure at the moment I cannot even run a dry_line() without encountering several issues in Octave under WSL (win11).

Let's say that this issue would be closed whenever we reach full compatibility of the currently implemented basic functionality, i.e. the core content of a 'run-time + post-processing' tool for QcmPlab DMFT libraries. Later on, we could relax the compatibility requirement for graphical functionality, e.g. accepting slightly different appearence in produced figures / movies. Again, we'll see.

In any case an octave label has been created, for future issues on the matter.

We are at a point where we need to greatly improve the flexibility of runDMFT and postDMFT packages. Spanning only Uloc values in all the computed and analyzed lines is becoming too limiting. I want to be able to span whatever variable I want (passing its fortran name) and keep Uloc just the default. Yet I don't want to drop other optional arguments and the positional nature of default Matlab arguments makes all this a nightmare. Finally I'd very very much prefer to not break the API, as I want my old scripts (scattered around in Ulysses and elsewhere) to keep running, if I ever need to (I have to admit this might be hard, we'll see...)

So inputParser objects might be the way to go. They are not particularly new in the language (since R2007a), but for some reason I never heard of them before, so there might be some pain points. I need to find time to investigate about them.

https://it.mathworks.com/help/matlab/matlab_prog/parse-function-inputs.html
https://it.mathworks.com/help/matlab/ref/inputparser.html
https://it.mathworks.com/help/matlab/matlab_prog/input-parser-validation-functions.html

Here we want a refresh routine that is indeed totally generic, namely:

• we do not know how many ctrl variables have been touched -> we cannot trust the input* file!
• we do not know how input* specifically resolves (it's model dependent).

So we need to find a way to move used.input* in input*, despite the wildcards not being supported by either MATLAB movefile function or BASH mv command.

Opened by commit 38fd6cd