io.IOBase docs often talk about byte offsets, but the same classes are used for string io types which use character offsets

[andmis]

The io.IOBase docs often talk about byte offsets, but the same classes often act as base classes for the string io types, which use character offsets.

PR fixing one such issue: https://github.com/python/cpython/pull/121421/files

Linked PRs

• gh-122299: fix io.IOBase.seek docs WRT offset type #121421

[vadmium]

I wouldn’t call it a character offset; the number is not necessarily equal or even proportional to a number of characters. Consider multi-byte encodings, the newline encoder and decoder, and _pyio.StringIO with non-ASCII.

For character IO seeking, the offset is generally meant to be an opaque value returned by tell, or zero.

[picnixz]

Considering that multiple core devs expressed their reservation, I will close this one as not planned. If you want to discuss it more in details and/or may suggest a better wording that would be accurate as well, feel free to @ me.