CEA-COSMIC / ModOpt

Modular Optimisation tools for solving inverse problems
MIT License
25 stars 15 forks source link

Development Roadmap and Future of ModOpt. #269

Open paquiteau opened 1 year ago

paquiteau commented 1 year ago

I reckon that I have started multiple PRs, moving forward here is a proposition for the integration of this work.

I would recommend we use a squash+rebase strategie, to minimize the effect on the main trunk. To smooth things up here is what I propose:

In a later future (e.g. start of Q2 2023), taking the opportunity of a coding sprint:

  1. Write test for the new algorithms. This might also be the opportunity to revisite the significance of all optimisation algorithm, to solve a less dummy problem (for instance a LASSO regularised least square)
  2. Move the configuration to pyproject.toml , and maybe use a src layout ^2, -> this seems to be the new standard ^3
  3. Release a new ModOpt version (2.0 ? )
  4. Write new a paper about it, add it to zenodo^5 to get a DOI for the package.
sfarrens commented 1 year ago

Sounds like a good plan to me @paquiteau!

Some other points to consider:

paquiteau commented 1 year ago

I mention the move to pyproject.toml in point 8 ;) Setuping the pyproject.toml should not be a big deal (I am getting pretty good at it and relies already on it for several of my projects, including pysap-fmri). this change of config file could happen simultaneously with the black formatting IMO.

sfarrens commented 1 year ago

Sorry I missed that, but since this is tied to the configuration of Black I think point 8 is linked to point 5 and should probably be done in conjunction.