[Python] Define API for user type checking of array types

[asfimport]

We have some subclasses of pyarrow.lib.DataType, but we haven't been designing with the intent of writing isinstance(arr.type, pyarrow.TimestampType). We should think about the public API for such type-checking or other type of schema validation.

Reporter: Wes McKinney / @wesm
Assignee: Wes McKinney / @wesm

PRs and other links:

• GitHub Pull Request #1183

_{Note: This issue was originally created as ARROW-1250. Please see the migration documentation for further details.}

[asfimport]

Phillip Cloud / @cpcloud:
The thing here is that it's hard to enforce only this usage. I removed this from the public API because it's too easy to have an unpleasant experience with these types. They are really only for wrapping their C++ counterparts and about the only thing useful you can do with them within Python is type checking. I think these should remain outside of the top level public API and be made public in one module deep (maybe pyarrow.types?).

[asfimport]

Wes McKinney / @wesm:
I agree. Having a types module might be a good approach, possibly with some functions like pyarrow.types.is_integer, pyarrow.types.is_timestamp, etc.

[asfimport]

Wes McKinney / @wesm:
Issue resolved by pull request 1183
#1183