gh-65002: Make note that null bytes are used to pad bytes

[slateny]

https://docs.python.org/dev/library/struct.html#format-characters

• Issue: doc: clarify that struct.pack_into writes 0x00 for pad bytes #65002

[ghost]

All commit authors signed the Contributor License Agreement.

[terryjreedy]

It would be best if this footnote were the first. Is (0) allowed or must footnote numbering start with 1?
If the latter, I'm NOT asking for renumbering in this situation.

A deeper question: by 'null byte', I presume the footnote means \0. Why put this in a footnote? Why not

Suggested change

	| ``x`` | pad byte | no value | | \(7) |
	| ``x`` | \0 pad byte | no value | | |

[terryjreedy]

Alternative

Suggested change

	| ``x`` | pad byte | no value | | \(7) |
	| ``x`` | pad byte (\0) | no value | | |

[CAM-Gerlach]

As mentioned on Discord, this isn't a footnote at all, its just arbitrary text—so it could be made anything at all, including 0. They should be made actual footnotes ([1]_/.. [1]), or much better auto-numbered footnotes ([#note-x]_/.. [#note-x]) that would avoid this problem and also getting out of sync and the reader doesn't have to manually scroll multiple pages to find the note (which I've had to do quite a bit on this particular table). However, I considered that out of scope of this particular PR, and better done in a followup, which is why the numbering didn't bother much (as it would be quickly fixed anyway).

I presume the clarification in the footnote is for when this is used to output rather than input a binary struct. My concern here is that since the behavior is not 1:1 on input and output (pad bytes on input are ignored, while on output they area always \0), I'm concerned this might mislead users into thinking that input bytes must be \0, rather than simply ignored. Also, it should be a literal, so \0

Incorporating @zware 's suggestion to resolve this, as well as making it a literal:

Suggested change

	| ``x`` | pad byte | no value | | \(7) |
	| ``x`` | pad byte | no value | | |
	| | (output as ``\0``) | | | |

[slateny]

What do you mean by input a binary struct? Like a = bytearray(b'\0Y')?

[CAM-Gerlach]

What I mean is if you call struct.unpack with x in the format on some struct-packed bytes as input, the bytes ignored by x don't have to be \0 (as the previous suggestion could imply), but rather could be anything. \0 is only what is output/packed, not nessesarily what is input/unpacked. I.e. on output/packing,

>>> struct.pack('!xxc', b'a')
b'\x00\x00a'

but those bytes can be anything on input/unpacking, not just \0, i.e.

>>> struct.unpack('!xxc', b'\0\0a')
(b'a',)
>>> struct.unpack('!xxc', b'42a')
(b'a',)
>>> struct.unpack('!xxc', b'\r\na')
(b'a',)

where as "\0 pad byte" may imply to the user that only the first might be accepted.

[terryjreedy]

If either change suggested above is accepted,

Suggested change

	(7)
	For padding, ``x`` inserts null bytes.

[ambv]

I agree with CAM that it would be good to move to auto-numbered footnotes. This, however, is not the point of the PR in question. So, let's just merge that first. We can fix the pseudo-footnotes in a dedicated PR.

[miss-islington]

Thanks @slateny for the PR, and @ambv for merging it 🌮🎉.. I'm working now to backport this PR to: 3.10.
🐍🍒⛏🤖

[miss-islington]

Thanks @slateny for the PR, and @ambv for merging it 🌮🎉.. I'm working now to backport this PR to: 3.11.
🐍🍒⛏🤖

[bedevere-bot]

GH-98803 is a backport of this pull request to the 3.10 branch.

[bedevere-bot]

GH-98804 is a backport of this pull request to the 3.11 branch.

[slateny]

There was some discussion on whether to move the footnote into the table itself above, but now that this is merged perhaps we just leave that for another time if wanted.