from_pymc3(save_warmup=True) puts tuning samples into posterior if draws=0

[michaelosthege]

Describe the bug
When a trace given to az.from_pymc3(trace, save_warmup=True) contains tuning samples, but no draws, the tuning samples end up in the posterior group and the warmup group remains empty.

To Reproduce

with pm.Model() as model:
    x = pm.Normal("x", mu=0, sigma=1)
    trace = pm.sample(
        tune=100,
        draws=0,
        discard_tuned_samples=False,
        chains=1,
    )
    idata = az.from_pymc3(trace, save_warmup=True)
# >> Sampling 1 chain for 100 tune and 0 draw iterations (100 + 0 draws total) took 0 seconds.

print(idata.warmup_posterior.dims)  # >> Frozen(SortedKeysDict({'chain': 1, 'draw': 0}))
print(idata.posterior.dims)         # >> Frozen(SortedKeysDict({'chain': 1, 'draw': 100}))

Expected behavior
Tuning samples should end up in the warmup group - even if there are no further draws.

Additional context
ArviZ master
PyMC3 master

[michaelosthege]

The existing test can be parametrized to detect it:

Line 403 in test_data_pymc.py

    @pytest.mark.skipif(
        not hasattr(pm.backends.base.SamplerReport, "n_draws"),
        reason="requires pymc3 3.9 or higher",
    )
    @pytest.mark.parametrize("save_warmup", [False, True])
    @pytest.mark.parametrize("chains", [1, 2])
    @pytest.mark.parametrize("tune,draws", [(0, 200), (100, 200), (100, 0)])
    def test_save_warmup(self, save_warmup, chains, tune, draws):
        with pm.Model():
            pm.Uniform("u1")
            pm.Normal("n1")
            trace = pm.sample(
                tune=tune,
                draws=draws,
                chains=chains,
                cores=1,
                step=pm.Metropolis(),
                discard_tuned_samples=False,
            )
            assert isinstance(trace, pm.backends.base.MultiTrace)
            idata = from_pymc3(trace, save_warmup=save_warmup)
        prefix = "" if save_warmup else "~"
        test_dict = {
            "posterior": ["u1", "n1"],
            "sample_stats": ["~tune", "accept"],
            f"{prefix}warmup_posterior": ["u1", "n1"],
            f"{prefix}warmup_sample_stats": ["~tune"],
            "~warmup_log_likelihood": [],
            "~log_likelihood": [],
        }
        fails = check_multiple_attrs(test_dict, idata)
        assert not fails
        assert idata.posterior.dims["chain"] == chains
        assert idata.posterior.dims["draw"] == draws
        if save_warmup:
            assert idata.warmup_posterior.dims["chain"] == chains
            assert idata.warmup_posterior.dims["draw"] == tune

[michaelosthege]

It's a slicing problem.. Doing a PR...