Risk Trajectory Split 1 : Snapshots

[spjuhel]

As the Risk trajectory PR is too substantial, this is a first split.

This PR implements Snapshots as the foundational object for risk trajectories.

These objects regroup Exposures, Hazard and ImpactFuncSet instances for a specific date.

The rationale is to provide an immutable snapshot of risk. As such, Exposures, Hazard, ImpactFuncSet and date inputs are deep-copied, and "protected" behind read-only properties.

Likewise, to avoid any ambiguity regarding adaptation, i.e., as a user, if I provide an adaptation measure, should I give the base (Exposures, Hazard, ImpactFuncSet) triplet to which the measure is applied ? Or the triplet where the measure as already been applied, a measure is always None and can only be set through apply_measure() which returns a new object (with the measure set).

PR Author Checklist

• Read the Contribution Guide
• Correct target branch selected (if unsure, select develop)
• Descriptive pull request title added
• Source branch up-to-date with target branch
• Documentation updated
• Tests updated
• Tests passing
• No new linter issues
• Changelog updated

PR Reviewer Checklist

• Read the Contribution Guide
• CLIMADA Reviewer Checklist passed
• Tests passing
• No new linter issues

[peanutfun]

Looks good, but I somehow still have 19 comments 🙈 The linter pipeline does not seem to work, and some comments are related to would-be linter issues.

As discussed, feel free to use pytest when writing new test modules.

We also briefly discussed writing this as a frozen dataclass, but I think the ref_only parameter would make the initialization for a dataclass too cumbersome. You can leave it as is, but there might be a way to remove the property definitions (see comment).

[peanutfun]

Is it really useful to support ints? This result could be a bit confusing

[spjuhel]

Well, the default thing people will use are years, so I would say yes.

[peanutfun]

But string years are compatible with pandas and numpy:

>>> pandas.Timestamp("2022")
pandas.Timestamp('2022-01-01 00:00:00')
>>> numpy.datetime64("2022")
numpy.datetime64('2022')
>>> numpy.datetime64("2022", "D")  # Force interval 'day'
numpy.datetime64('2022-01-01')