Great work!

Some lines look too long, and some lines look unnecessary short. Could you please reformat the text using the recommended maximal line length (I think 79 from PEP 12 is applied here)?

Alphabetical order is perhaps necessary for such large number of changes, but in some cases I think that it would be better to group semantically close changes together. For example, PyObject_GenericHash and Py_HashPointer. Or PyMapping_GetOptionalItem, which is close to PyDict_* changes and to PyObject_GetOptionalAttr, and to PyMapping_HasKeyWithError. But I do not see a way how this can be improved.

Thanks for the review, I've pushed two commits.

Some lines look too long, and some lines look unnecessary short. Could you please reformat the text using the recommended maximal line length (I think 79 from PEP 12 is applied here)?

The devguide also says 79/80, but in a few places I've gone a couple of characters over to avoid a mostly empty orphan line.

I've rebalanced the particularly egregious case you pointed out, but breaking lines early is a somewhat intentional choice (some people call it semantic line breaks) where I break the line at punctuation or a clause in a sentence. This tends to lead to smaller diffs later down the line, as re-wrapping an entire paragraph for changing a single word can be avoided.

(This is of course only visible in the .rst -- the vast majority readers will never know which lines were broken where in the HTML or PDF output)

A