`_pyio.TextIOBase.readline` is missing a `size` arg

[ShaharNaveh]

Bug report

Bug description:

IOBase:

cpython/Lib/_pyio.py

Line 513 in a064b32

def readline(self, size=-1):

TextIOBase:

cpython/Lib/_pyio.py

Line 1931 in a064b32

def readline(self):

---

This pattern seems to be missing only for readline AFAICT.
for truncate (which is also "unsupported" for TextIOBase) it's preserved:

IOBase:

cpython/Lib/_pyio.py

Line 370 in a064b32

def truncate(self, pos=None):

TextIOBase:

cpython/Lib/_pyio.py

Line 1927 in a064b32

def truncate(self, pos=None):

(the same for write and others).

My proposed fix should be adding a size argument to readline with a default value of -1

CPython versions tested on:

CPython main branch

Operating systems tested on:

Linux

Linked PRs

• gh-151596: Added missing argument 'size' with default value of -1 to TextIOBase.readline #151679
• gh-151596: Add size to _pyio.TextIOBase.readline #151680
• [3.15] gh-151596: Add missing argument 'size' to pure-Python implementation of TextIOBase.readline (GH-151679) #151870
• [3.14] gh-151596: Add missing argument 'size' to pure-Python implementation of TextIOBase.readline (GH-151679) #151871
• [3.13] gh-151596: Add missing argument 'size' to pure-Python implementation of TextIOBase.readline (GH-151679) #151872

[picnixz]

cc @cmaloney

[cmaloney]

truncate is a bit different as "The current stream position isn’t changed" (https://docs.python.org/3/library/io.html#io.IOBase.truncate).

In the C implementation _io.TextIOWrapper, _io.BufferedIO, and _io.BytesIO do take an optional size argument today so this works as specified today. _pyio takes one in _pyio.TextIOWrapper.readline. _pyio isn't used in production anywhere as far as I'm aware and it's only the _pyio.TextIOBase.readline which is an abstract method missing the size=. I'm open to a tiny patch without NEWS adding size to _pyio.TextIOBase.readline it but I am doubtful that is worth the time / effort to write, review, and ship as _pyio is a CPython internal module.

[ZeroIntensity]

I'm open to a tiny patch without NEWS adding size to _pyio.TextIOBase.readline it but I am doubtful that is worth the time / effort to write, review, and ship as _pyio is a CPython internal module.

I don't understand this. _pyio is a fallback when the C version of io isn't available, isn't it? Why is it considered internal?

[cmaloney]

As currently coded it's not used as a fallback and requires _io / the C version to build/run.

[ZeroIntensity]

I suspect that other implementations might be using it anyway. _io can be written in an implementation's native language to get the primitive aspects, and then use _pyio from there to get the rest of the module.

That said, this is a bit of a gray area. Do you know of any precedent to point to here?

[cmaloney]

I looked at some of the other Python implementations (ex. PyPy) that have been brought up and all I have investigated either use _io or have a fully independent implementation of the io API. Have had some in person conversations around what makes sense for _pyio long term. All that said a NEWS file is not going to cause harm here; I thought it would lead to more review round trips and wasn't really valuable hence trying to just leave out for the one line patch.

[ZeroIntensity]

Yeah, no worries, I'm just being pedantic :)