CLN, TYP use string types in generics

[MarcoGorelli]

• Ensure all linting tests pass, see here for how to run them

[simonjayhawkins]

is this because of aliases? If so, we may also define aliases local to a module, so we would need to allow if preceeded by =

[simonjayhawkins]

we also have the issue of casts and TypeVars, which still require the quoting.

[MarcoGorelli]

It's not because of aliases (they're fine), but because in _typing we would otherwise get

ImportError while loading conftest '/home/marco/pandas-marco/pandas/conftest.py'.
pandas/__init__.py:22: in <module>
    from pandas.compat.numpy import (
pandas/compat/__init__.py:14: in <module>
    from pandas._typing import F
pandas/_typing.py:79: in <module>
    FrameOrSeriesUnion = Union["DataFrame", Series]
E   NameError: name 'Series' is not defined

because Series is only defined if TYPE_CHECKING (same situation as in the other comment)

[simonjayhawkins]

yep, I wasn't clear. it's not just type aliases https://www.python.org/dev/peps/pep-0563/#forward-references

This PEP addresses the issue of forward references in type annotations. The use of string literals will no longer be required in this case. However, there are APIs in the typing module that use other syntactic constructs of the language, and those will still require working around forward references with string literals. The list includes:

[simonjayhawkins]

I guess this is also an option. would prefer to update regex though.

[MarcoGorelli]

Sure - this wasn't for the regex though, but rather because otherwise we get

ImportError while loading conftest '/home/marco/pandas-marco/pandas/conftest.py'.
pandas/__init__.py:51: in <module>
    from pandas.core.api import (
pandas/core/api.py:31: in <module>
    from pandas.core.groupby import Grouper, NamedAgg
pandas/core/groupby/__init__.py:1: in <module>
    from pandas.core.groupby.generic import DataFrameGroupBy, NamedAgg, SeriesGroupBy
pandas/core/groupby/generic.py:69: in <module>
    from pandas.core.frame import DataFrame
pandas/core/frame.py:156: in <module>
    from pandas.core.series import Series
pandas/core/series.py:108: in <module>
    from pandas.core.tools.datetimes import to_datetime
pandas/core/tools/datetimes.py:74: in <module>
    ArrayConvertible = Union[List, Tuple, AnyArrayLike, Series]
E   NameError: name 'Series' is not defined

because here Series is only defined if TYPE_CHECKING

[simonjayhawkins]

https://www.python.org/dev/peps/pep-0563/#forward-references goes on to say

Depending on the specific case, some of the cases listed above might be worked around by placing the usage in a if TYPE_CHECKING: block. This will not work for any code that needs to be available at runtime, notably for base classes and casting. For named tuples, using the new class definition syntax introduced in Python 3.6 solves the issue.

so this is fine

[simonjayhawkins]

Thanks @MarcoGorelli lgtm except for the chances of false positives. We probably need to use a AST parser to do this checking in the longer term.

[MarcoGorelli]

We probably need to use a AST parser to do this checking in the longer term.

Agreed - it's in my backlog to make sense of that and to make a better solution (which could be its own package), will get back to you on it when I make progress

[MarcoGorelli]

We probably need to use a AST parser to do this checking in the longer term.

Agreed - it's in my backlog to make sense of that and to make a better solution (which could be its own package), will get back to you on it when I make progress

I just gave it a go and I think I have something which does the job! Will make a PR tomorrow, we can close this one, thanks for pushing me to do it properly 🙏