Nathaniel J. Smith
Nathaniel J. Smith
It also would avoid having to answer the question "what if the metadata and the data disagree?"
Oh, fair enough. Maybe "the union of what we see in the wheel + what's listed in top-level.txt"
After reminding myself of how this is supposed to work: in theory everyone is supposed to use `PyDataMem_NEW`, so it might be fine. But there are enough weird C API...
Well, I guess there's an easy local fix... we already want to deprecate using `Process` objects as a context manager, so that will automatically take care of this :-) (#1104)...
To clarify: what we want to deprecate is the use of `async with `, which is the thing that's causing your test to fail. That might also be the quickest...
Because the existing fallback is to look for a `.sqrt` method on the given objects, but nobody except numpy has ever heard of this method. In particular standard python objects...
The speed issues seems irrelevant to me -- checking the type is not really going to be noticeable by the time we're already calling Python functions, checking for methods and...
Oh, and for errors, I think the logic we want is: import numbers def sqrt_fallback(obj): if hasattr(obj, "sqrt"): # Any exception here is allowed to escape: return obj.sqrt() transformed_obj =...
Emacs uses control-space to make a note of the beginning of the selection, but in classic Emacs (and I believe also readline) this does _not_ enter a special confusing mode....
Seems like a reasonable thing to fix, want to send a PR?