- User Since
- Apr 25 2019, 5:20 PM (42 w, 5 d)
Mon, Feb 17
i would have biased towards going in the other direction and strongly typing the whole pandas API rather than reducing the surface we cover. one benefit of strongly-typed config is catching errors early. even if we do go the other direction, i don't like the introduction of a new parameter name that doesn't align with the underlying API -- why serialization_options and not kwargs? (i admit that taking this position to its extreme would replace path with filepath_or_buffer, which, maybe, is the wrong thing to do)
I'm not sure what you mean by prior art.
Sat, Feb 15
if i were concerned about this, i would also probably be using a username and db name that i didn't want anywhere -- we consider perhaps reading those in in the same way
Fri, Feb 14
Can we add a test also!
Thu, Feb 13
Wed, Feb 12
I think I'm on the same page as Alex
Tue, Feb 11
Respond to comments
I think we should hard break this now so that the changes are bundled with the other custom type changes