Closed keitherskine closed 1 year ago
And again, thanks for noticing stuff I didn't. Is there any way to create a unit test to compare the pyi file against the binary?
Long term I'm thinking we might want to split pyodbc into Python classes and a _pyodbc C library, as long as it doesn't degrade performance. A performance benchmark is necessary for determining if the stable ABI is a good idea too. Hmmm... Need to prioritize that.
The utility mypy
has a facility for doing a certain amount of stub checking, but I suspect that might depend on a pure-Python underlying module.
I think creating a pure-Python top/level module for pyodbc and C underlying extension classes would be a great idea. There's a lot of fairly straightforward non-ODBC stuff that could be done without invoking the power of C, e.g. parameter validation. It would be handy to be able to do that in Python.
Looks like a couple of entities are no longer in the pyodbc interface - "ansi" and "UNICODE_SIZE". This PR removes them.