proposal: bytes: add (*bytes.Buffer).SetBytes method about go

[randall77]

*b = *bytes.NewBuffer(buf) works. The NewBuffer call is inlined and the allocation inside it happens on the stack.
(Copy a struct of another package wholesale is generally frowned upon, but it works in this case.)

[dsnet]

I still wonder if SetBytes has value as an explicit API?

As prior precedence, we added the AvailableBuffer method to make it more obvious how to correctly append into a bytes.Buffer even though this functionally had always been possible by messing around with the capacity of the buffer returned by the Bytes method.

[seankhliao]

Why not just reset? https://pkg.go.dev/bytes#Buffer.Reset

[ajwerner]

Why not just reset? https://pkg.go.dev/bytes#Buffer.Reset

Reset truncates the existing buffer. The use case here is that I have a byte slice already and I have a bytes.Buffer object and I'd like to create an io.Reader to read those bytes without allocating.

A legitimate claim is that the bytes.Reader API should work, because it indeed allows for resetting with a new slice. Where that turns out to not work is that bytes.Reader doesn't expose the underlying byte slice (it doesn't have a Bytes method).

[ajwerner]

*b = *bytes.NewBuffer(buf) works. The NewBuffer call is inlined and the allocation inside it happens on the stack.

(Copy a struct of another package wholesale is generally frowned upon, but it works in this case.)

Cool! I'll close this and hope that some future wayward soul may discover this trick.

[dsnet]

*b = *bytes.NewBuffer(buf) just blew my mind... I've been programming Go for 10 years and that didn't occur to me.

[ianlancetaylor]

Anybody want to add a test to make sure it keeps working? It's not like it's documented.

[dsnet]

I can add a test. I'm about to use this trick.

[gopherbot]

Change https://go.dev/cl/581297 mentions this issue: bytes: add test to ensure shallow copy of NewBuffer does not allocate