Fix the way the mapmaker handles signals

[ziotom78]

This PR modifies _Toast2FakeCache so that it performs deep copies even when Signal.copy() is called.

[paganol]

Hi @ziotom78 @giuspugl @sgiardie, currently baseline_lenght is interpreted as the number of consecutive samples in a baseline. Digging in toast, it seems that it is actually the length in seconds. @giuspugl, do you confirm this?

[giuspugl]

Hi @ziotom78 @giuspugl @sgiardie, currently baseline_lenght is interpreted as the number of consecutive samples in a baseline. Digging in toast, it seems that it is actually the length in seconds. @giuspugl, do you confirm this?

I confirm that it should be the length in seconds !

[ziotom78]

Hi, I have renamed baseline_length to baseline_length_s. It's a breaking change, but it makes the code far easier to read:

    params = lbs.DestriperParameters(
        nside=16,
        baseline_length_s=100,
        return_destriped_map=True,
    )

I have updated @sgiardie 's notebook accordingly, but I fear that there will be changes to be made in the production scripts. Do you think this is reasonable to do?

[ziotom78]

@paganol , @sgiardie , what do you think about the breaking change? I would be in favor of implementing it and releasing 0.6.0, as in this way the simulation production will have a self-explaining code.

[sgiardie]

@paganol , @sgiardie , what do you think about the breaking change? I would be in favor of implementing it and releasing 0.6.0, as in this way the simulation production will have a self-explaining code.

Hi @ziotom78 , sorry for the late reply. I think it is a reasonable change, and also not so impacting on the production scripts (it would require just to rename that variable, I guess)

[ziotom78]

(it would require just to rename that variable, I guess)

Yes, that's right, it's something you can do automatically if you use an IDE like PyCharm or Spyder.

I'll draft release v0.6.0 then, instead of v0.5.1, as this is a breaking change.

[ziotom78]

I'll draft release v0.6.0 then, instead of v0.5.1, as this is a breaking change.

Wait, I just realized that we can keep baseline_length as well but mark it as «deprecated». In this way we can release 0.5.1 as planned.

[ziotom78]

Wait, I just realized that we can keep baseline_length as well but mark it as «deprecated». In this way we can release 0.5.1 as planned.

Mmm, no, deprecating a field of a dataclass is not trivial and I fear to complicate the code too much for little gain.

Sorry for the noise.