BUG: recognize M8[ns, tzstr]

[jbrockmendel]

• closes #xxxx
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

ATM dtype == "M8[ns, UTC]" fails while dtype == "datetime64[ns, UTC]" succeeds

[jorisvandenbossche]

Is this needed to support? (what's the use case?)
M8 is a numpy short-hand name, but tz-aware dtype is not a numpy dtype.

[jreback]

lgtm.

[jorisvandenbossche]

see comment

[jreback]

see comment

I find this ok, all we are asserting here is that datetime64==M8 which is true

[jbrockmendel]

Is this needed to support? (what's the use case?)

needed? No. convenient? yes.

Until recently there as a test that checked dtype == "M8[ns, US/Eastern]" that would have failed if it wasn't missing the "assert".

[jorisvandenbossche]

What is the convenience? Just less typing?

[jbrockmendel]

What is the convenience? Just less typing?

Sure. Or maybe "lease surprise" is more accurate. e.g. is_dtype_equal(dtpye, "M8[ns, US/Eastern]") is True ATM while dtype == "M8[ns, US/Eastern]" is False

[jorisvandenbossche]

OK, but then it is fixing a bug, not providing something for convenience.

Eg, it seems that pd.Series([1, 2, 3]).astype("M8[ns, UTC]") already works. Then of course .dtype == "M8[ns, UTC]" should also work.
Do you know if it is tested that astype / pandas_dtype(..) works with this notation? If not, can you add a test for that?

[jbrockmendel]

If not, can you add a test for that?

added