Rewrite most of the core_foundation module based on Foundation #88
Conversation
This happens if the contained method is an ObjCPartialMethod.
The latter gives better failure messages.
On 64-bit systems, the NS structures are exact typedefs of the CG ones.
On 32-bit, both have the same layout, but are separate declarations
with different names. This is significant in the type encodings:
on 64-bit, both NSPoint and CGPoint encode as {CGPoint=dd}, but on
32-bit, NSPoint encodes as {_NSPoint=ff}, and CGPoint as {CGPoint=ff}.
There was a problem hiding this comment.
I don't know why I bother doing a careful line-by-line teardown of your patches... they're always awesome, and I almost never find anything to fault.
The only thing that gave me pause was the way NSData conversion was being handled (line 1729 of runtime.py) - the name string_at() looked off on first inspection. However, a quick check of the ctype docs shows it's referring to a C string - which isn't very helpful from a Python naming POV, but is otherwise entirely accurate. A quick comment to explain the odd naming might be helpful; but otherwise - 👏 Once again, an amazing contribution.
|
Thanks :) I appreciate your reviews, it's always good to have someone else look for logic errors and such, and to get feedback on the API. Yeah, the naming of |
Implements #74.
This is a rewrite of most of the
core_foundationmodule's functions based on Foundation classes, resulting in more readable code with no manual function declarations needed. The only things left in thecore_foundationmodule are the basic types/constants/functions used by theasynceventloopmodule.The conversion functions have also been restructured a bit.
from_valueandto_valuehave been renamed tons_from_pyandpy_from_ns, because the old names weren't very clear on what direction the conversion went. Theatfunction, which previously only produced strings, is now exactly equivalent tons_from_py. This allows things likeat(42)orat([1, 2, 3]), similar to the@prefix/operator in Objective-C. The type-specific conversion functions (to_str,to_number, etc.) have been removed, becausepy_from_nscan be used in place of all of them.I don't know if the old functions from
core_foundationare used much externally. I did a quick search through the Toga repo and couldn't find anything exceptatbeing used. (athas only been moved internally, so that shouldn't be an issue.) Are there any other users of Rubicon that I should check for this sort of thing? I know Rubicon isn't officially stable yet, but I still prefer not to break other people's code.As usual, this PR includes some general cleanup and fixes, so the individual commit diffs may be easier to read than the full diff.