Add TGraph, RooCurve, and RooHist behaviours.

[beojan]

This adds a behaviour for TGraph (based on that for TGraphAsymmErrors), and behaviours for RooHist and RooCurve based on these. These make it easier to plot RooFit objects on a plot that's otherwise made with Python.

Draft because I haven't added any tests yet. I thought it would be better to discuss first.

[jpivarski]

It passes tests, but let me know when it's out of "draft" status (when you think it's ready for merging).

A comment: I'd prefer to not introduce new specialized exception types; it looks like both of these could be ValueError. I know that having more exception types can be good for writing more fine-grained try-catch logic, but it's hard to be consistent in defining exception types for different things; the system tends to get lopsided with specialized exceptions for one thing and not others. Uproot has a few specialized exceptions (subclasses of the exceptions you'd expect, such as KeyInFileError for KeyError), but all of those cases involve getting more information to users. (KeyInFileError gives a list of other keys, sorted by string difference from the one you gave; DeserializationError is part of a try, try-again-with-TStreamerInfo pattern, etc.)

[jpivarski]

Let me know when this is done!

[beojan]

Sorry, I was a little busy with other things. Here's the test file I'd like to use:

test-rooplot.zip

[jpivarski]

If you add that file in the directory below with name "uproot-issue-350.root", we can make a real test around it:

https://github.com/scikit-hep/scikit-hep-testdata/tree/main/src/skhep_testdata/data

I'll approve that scikit-hep-testdata PR and make a new version quickly so that you can test it here.

[beojan]

PR made here. Sorry for the delay.

[jpivarski]

@beojan, uproot-issue-350.root is now available in scikit-hep-testdata 0.4.4. Tests using it should now work!

[beojan]

Tests now added. It looks like it failed because of a conda issue though.

[jpivarski]

I restarted the tests, and they're failing in a relevant part now:

______________________ ERROR at setup of test_interpolate ______________________
file /home/runner/work/uproot4/uproot4/tests/test_0350-read-RooCurve-RooHist.py, line 117
  def test_interpolate(roocurve, roocurve_err):
E       fixture 'roocurve' not found
>       available fixtures: cache, capfd, capfdbinary, caplog, capsys, capsysbinary, datafile, doctest_namespace, monkeypatch, pytestconfig, record_property, record_testsuite_property, record_xml_attribute, recwarn, reset_classes, roocurve_err, roocurve_hist, roohist, tmp_path, tmp_path_factory, tmpdir, tmpdir_factory
>       use 'pytest --fixtures [testpath]' for help on them.

Maybe you've set up the new fixture locally but haven't checked it in?

(Unless it simplifies things in a noticeable way, I don't think it's necessary to use a lot of pytest features, such as fixtures. For instance, is this fixture used in many places, such that replacing redundancy with indirection is a good trade-off? I generally see tests as requiring a lower level of DRY than the main codebase, since the majority of them are written once and never updated, whereas the main codebase is frequently operated upon.)

[jpivarski]

_____________________________ test_interpretation ______________________________

roohist = <RooHist (version 1) at 0x7f66e4954650>
roocurve = <RooCurve (version 1) at 0x7f66fc0707d0>
roocurve_err = <RooCurve (version 1) at 0x7f66d7f49c10>

    def test_interpretation(roohist, roocurve, roocurve_err):
        assert roohist.classname == "RooHist"
>       assert roohist.behaviors[0] == uproot.behaviours.RooHist.RooHist
E       AttributeError: module 'uproot' has no attribute 'behaviours'

tests/test_0350-read-RooCurve-RooHist.py:38: AttributeError

Needs an explicit import uproot.behaviors.RooHist in the testing file. Behaviors aren't loaded preemptively because they might grow to be a large amount of code someday.

________________________________ test_to_boost _________________________________

roohist = <RooHist (version 1) at 0x7f66d75c7890>
roocurve = <RooCurve (version 1) at 0x7f66d75d6910>
roocurve_err = <RooCurve (version 1) at 0x7f66d75cf910>

    def test_to_boost(roohist, roocurve, roocurve_err):
        rh_boost = roohist.to_boost()
>       assert rh_boost.axes[0].edges == numpy.arange(0.0, 51.0)
E       ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

tests/test_0350-read-RooCurve-RooHist.py:51: ValueError

Could use NumPy's array_equal, or convert both sides with tolist, or use all as they suggest.

[jpivarski]

(I'm going offline now. Might be unresponsive; sorry!)

[beojan]

Needs an explicit import uproot.behaviors.RooHist in the testing file.

That wasn't it (I copied the TGraphAsymmErrors test). I just spelled "behaviors" the British way.

I've also fixed the other test.

[beojan]

All tests passed. I think this is ready for review.

[jpivarski]

This looks absolutely great, thank you very much!

[beojan]

By the way, are you planning to cut a new release soon? It's not urgent (I can just build from git), I'm just curious.

[jpivarski]

I am. A lot of things have been coming in on Awkward and I'm looking for a good gap in which to cut a release there, and I would then do both, but there's no reason I shouldn't do it for Uproot now.