Toast / sotodlib mapmaking

[Wuhyun]

Mapmaking utilities built for toast and sotodlib data. Depends on toast and sotodlib for some functionalities.

The code contains all utilities needed to load toast and sotodlib data objects and run mapmaking. However, there are currently two independent sets of codes. These codes will be refactored to use a single combined interface in the near future.

[ASKabalan]

I added some very high level comments
I did not go into all the details but looks great to me
Thank you very much for all of this 😃

[ASKabalan]

I am not a big fan of perf measure in production code
this could be in an outside script maybe?

[Wuhyun]

Agreed, removed the performance measures but allowed a custom logger for timing.

[ASKabalan]

I am assuming this is some static data class that it used internally
however it is still very easy to make into an actual jittable pytree in case it is passed around

Suggested change

	@dataclass
	class ObservationData:
	@jax.tree_util.register_dataclass
	@dataclass
	class ObservationData:

check
https://jax.readthedocs.io/en/latest/_autosummary/jax.tree_util.register_dataclass.html

[Wuhyun]

Thank you! I simply added this line for now. In the future, I will look into specifying meta/data fields for performance.

[ASKabalan]

I think that copy should not be used and XLA would automatically make a copy if it is needed

[sbiquard]

I think the copy() is justified here to create a shallow copy of the original list of pytree leaves. Maybe @pchanial has an opinion on this (this function is mostly taken from his default as_matrix() implementation for abstract linear operators).

[ASKabalan]

You can add scripts in pyproject.toml for convinience
see https://packaging.python.org/en/latest/guides/writing-pyproject-toml/#creating-executable-scripts

[Wuhyun]

This is cool! I added the 'furax-mapmaking' command (for now) by editing pyproject.toml

[sbiquard]

Squashed all modifications (before @Wuhyun's latest commits to address mypy errors and review comments) into one giant commit to avoid merge commits that were messing up the "changes" compared to main branch.

[sbiquard]

Added @Wuhyun as co-author on the mega-commit :)

[pchanial]

Quels sont vraiment les motivations / bénéfices de l'immutabilité ici ? On pourrait lire une config par défaut, puis la mettre à jour ensuite partiellement.

[sbiquard]

Dans l'idée, ne pas pouvoir modifier la configuration spécifiée au départ par inadvertance.
Si l'utilisateur fournit un fichier de configuration partiel, les champs restants prendront la valeur par défaut.