Go: proposal: fmt: %p should accept uintptr

[seebs]

What version of Go are you using (go version)?

1.15

Does this issue reproduce with the latest release?

y

What operating system and processor architecture are you using (go env)?

N/A

What did you do?

fmt.Printf("%p", stringHeader.Data)

What did you expect to see?

A pointer value.

What did you see instead?

A fmt warning that %p got a non-pointer.

Rationale: It's true that uintptr is an integer-like thing rather than a pointer-like thing, but it's explicitly the type of integer which is used to represent pointers. It's certainly possible to just add the extra conversion everywhere, but I don't think it makes anything clearer, safer, or better. Allowing %p for uintptr shouldn't break any code except possibly "tests that check for getting that diagnostic" which seems like an edge case.

In the absence of moving GC, etc, I'd want %p to print a uintptr and the corresponding unsafe.Pointer identically.

(As a secondary issue, I would also sort of like %p to work on strings, printing the address of their backing store, but I'm enough less convinced of that that I'm leaving it as a parenthetical.)

[ianlancetaylor]

CC @robpike

[mvdan]

https://github.com/golang/go/issues/28865 is somewhat related, I think.

[robpike]

I'd rather that this not happen because uintptr is not a pointer, and the use of %p provides a modicum of type safety. If you want the hex you can always ask for it with %#x.

[seebs]

Yeah. This ties in somewhat with the occasional proposals for a slice-header-like-thing which contains an unsafe.Pointer instead of a uintptr. That said, I don't think there's a format specifier for uintptr that produces the same results as %p -- because %p can print "nil".

[rsc]

It sounds like %#x is fine.
Printing uintptr(0) as nil would be pretty surprising.
It's an integer, not a pointer.

[rsc]

Adding to minutes this week, but this seems headed for a likely decline based on the conversation.

[rsc]

Based on the discussion above, this seems like a likely decline.

[rsc]

No change in consensus, so declined.