fix: return proxy of attrs, behavior; dak.from_awkward correctly sets attrs

[agoose77]

Closes #447 by making .attrs and .behavior read-only. This addresses a confusing part of the API: currently these are writeable, but the underlying compute does not share the results of mutating attrs or behavior.

If we need to, we can add a dask_awkward.with_attrs to add a mechanism of mutating these objects. I'm not sure what the ramifications of this PR will be — we may internally be expecting to pass around the true attrs dicts, which would be broken by this change.

[lgray]

So I can only add attrs to an awkward array and then make a dask-awkward array from that?

This probably breaks nanoevents?

[lgray]

yep:

self = <coffea.nanoevents.factory.NanoEventsFactory object at 0x12a0f1330>

    def events(self):
        """Build events"""
        if self._is_dask:
            events = self._mapping(form_mapping=self._schema)
            report = None
            if isinstance(events, tuple):
                events, report = events
>           events.attrs["@original_array"] = events
E           TypeError: 'mappingproxy' object does not support item assignment

[agoose77]

You can still override attrs of a dask_awkward array, but it's harder to do. You just copy the meta and patch the .attrs. The intention of this PR is that the user-facing .attrs should not be writeable.

[lgray]

Ah, indeed I can see a few work arounds. I'll confirm one in an hour or so.

[lgray]

Yep - all good - simple work around found.

[douglasdavis]

based on conversation this morning it seemed that this is good to go?

[martindurant]

(test coverage would be nice)

[lgray]

@douglasdavis yes this is fine from my end

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (782b831) 93.10% compared to head (bd52045) 93.17%.
Report is 1 commits behind head on main.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #448      +/-   ##
==========================================
+ Coverage   93.10%   93.17%   +0.07%     
==========================================
  Files          23       23              
  Lines        3293     3298       +5     
==========================================
+ Hits         3066     3073       +7     
+ Misses        227      225       -2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[lgray]

@martindurant let me know if this is sufficient for tests, otherwise I think this one is good to go. Thanks in advance for the review.

[lgray]

@martindurant fixing a few things has led to scope creep of this PR... 🤷 Anyway it's ready for another look.