Improve error message for unsupported datetime dtypes in Pandas extra

[Zac-HD]

In this StackOverflow question, the asker notes that column('timestamp', dtype=pd.datetime) doesn't actually create a datetime column in the generated dataframe. This turns out to be a fairly general issue: passing dtype=datetime.datetime will actually create an object column too, and even passing dtype="datetime64" is problematic as Hypothesis will choose various time-resolution units. I think we should therefore:

1. Ensure that if you get an error when we can tell that you probably wanted datetimes, we explicitly suggest dtype="datetime64[ns]" as part of the error message
2. note_deprecation(...) if we can detect this and it wasn't previously an error
3. note_deprecation(...) if the time-resolution unit is not ns - others are valid in Numpy but not Pandas
4. Investigate doing the same for pd.Timedelta and related dtype arguments (see docs here)

[big-g-squared]

I'm currently working on fixing this issue and wanted to ask for some guidance. I ran the example in the stack overflow question and got the following error:

practice.py:8: FutureWarning: The pandas.datetime class is deprecated and will be removed from pandas in a future version. Import from datetime module instead.
  @given(data_frames([column('A', dtype=str), column('B', dtype=int), column('Date', dtype=pd.datetime)]))
/home/spoderman94/basis_project/venv/lib/python3.8/site-packages/hypothesis/extra/pandas/impl.py:84: HypothesisDeprecationWarning: Passed dtype=<class 'datetime.datetime'> is not a valid Pandas dtype.  We'll treat it as dtype=object for now, but this will be an error in a future version.
  note_deprecation(
Traceback (most recent call last):
  File "practice.py", line 13, in <module>
    s.test_example_test_method()
  File "practice.py", line 9, in test_example_test_method
    def test_example_test_method(self, dataframe):
  File "/home/spoderman94/basis_project/venv/lib/python3.8/site-packages/hypothesis/core.py", line 1397, in wrapped_test
    raise the_error_hypothesis_found
  File "/home/spoderman94/basis_project/venv/lib/python3.8/site-packages/hypothesis/extra/numpy.py", line 183, in from_dtype
    raise InvalidArgument(f"No strategy inference for {dtype}")
hypothesis.errors.InvalidArgument: No strategy inference for object

I noticed that there was a deprecation message being sent out from the "extra/pandas/impl.py" file, specifically in a function called "elements_and_dtype." I was wondering: would it be appropriate to place a "dtype=pd.datetime"-specific deprecation notice in this area? This is what the function looks like:

hypothesis/hypothesis-python/src/hypothesis/extra/pandas/impl.py

Lines 60 to 99 in 173de52

	@check_function
	def elements_and_dtype(elements, dtype, source=None):
	if source is None:
	prefix = ""
	else:
	prefix = f"{source}."
	if elements is not None:
	check_strategy(elements, f"{prefix}elements")
	else:
	with check("dtype is not None"):
	if dtype is None:
	raise InvalidArgument(
	f"At least one of {prefix}elements or {prefix}dtype must be provided."
	)
	with check("is_categorical_dtype"):
	if is_categorical_dtype(dtype):
	raise InvalidArgument(
	f"{prefix}dtype is categorical, which is currently unsupported"
	)
	if isinstance(dtype, type) and np.dtype(dtype).kind == "O" and dtype is not object:
	note_deprecation(
	f"Passed dtype={dtype!r} is not a valid Pandas dtype. We'll treat it as "
	"dtype=object for now, but this will be an error in a future version.",
	since="2021-12-31",
	has_codemod=False,
	)
	if isinstance(dtype, st.SearchStrategy):
	raise InvalidArgument(
	f"Passed dtype={dtype!r} is a strategy, but we require a concrete dtype "
	"here. See https://stackoverflow.com/q/74355937 for workaround patterns."
	)
	dtype = try_convert(np.dtype, dtype, "dtype")
	if elements is None:
	elements = npst.from_dtype(dtype)
	elif dtype is not None:

Additionally, the place where the error is actually raised is from "extras/numpy.py" in an area that looks like it's checking the different numpy "dtype.kind" properties of the given dtype. Would it be appropriate to place a "dtype=pd.datetime"-specific error message here? My thinking is that the deprecation notice that I mentioned above should be enough for the user to deduce that they may have used a deprecated datetime object as the dtype, but in your raised issue message you said that:

Ensure that if you get an error when we can tell that you probably wanted datetimes, we explicitly suggest dtype="datetime64[ns]" as part of the error message

Does this mean to do it in the raised error message? There is another piece that I want to explore which is the "try_convert" method from the "hypothesis.internal.validation" part of the program but I wanted to reach out for guidance just in case I'm overcomplicating things (as I tend to do). Thanks!