Use an enum for infer_dtype return values?

[TomAugspurger]

While working in #29799, it was quite hard to add a new return value from lib.infer_dtype. There are many places where we do something like

inferred_dtype = lib.infer_dtype(values)

if inferred_dtype == "string":
    ...

if inferred_dtype in {"datetime", "integer"}
    ...

It was hard to grep for these. Could we instead define an enum

class InferredType(enum.Enum):
    STRING: "string"
    INTEGER: "intger"
    MIXED_INTEGER: "mixed-integer"
    ...

And then return InferredType.INTEGER, and update all our checks for

if inferred_dtype == InferredType.INTEGER

I think this should be entirely backwards compatible, since we're still just returning the string.

[jreback]

yep this is generally a good idea

[alimcmaster1]

Sounds good this would be neat.

I guess the check would be

if inferred_dtype == InferredType.INTEGER.value

I can potentially take a look at this is we feel its worth doing.

[theoniko]

Hello @TomAugspurger, @mroeschke, @alimcmaster1, @jreback
Could you please have a look to #52517 and give me feeback pr is in the right direction? I am also not sure about

pandas/pandas/_libs/lib.pyx

Line 1522 in b1dbbe0

return value.inferred_type

?

[theoniko]

Could somebody have a look to my pr and give feedback?