BytesMut lacks an implementation of std::io::Write about bytes HOT 8 CLOSED

Sounds like a good idea to me! I'd grow the buffer internally

from bytes.

I don't think a writer that doesn't grow the buffer is of much use. Usually, when I need a writer, I delegate the encoding or writing to some other code. And often I don't know in advance how much such code will write.

An example of code that would make sense:

use serde_cbor;
use tokio_io::io::Encoder;

struct E;

impl Encoder for E {
  ...
  fn encode(&mut self, item: Self::Item, dst: &mut BytesMut) -> Result<(), Self::Error> {
    serde_cbor::to_writer(dst)
  }
}

Now I have to wrap dst into a newtype that implements writer because it doesn't know how to be a writer (which isn't much work). If it didn't extend the dst as needed, I'd still need to wrap it, because I don't know the size in advance.

from bytes.

This seems fine to me. My only question is if BytesMut should grow as needed or return an error when it is full.

Thoughts @alexcrichton

from bytes.

@alexcrichton So BufMut has a writer adapter: BufMut::writer(). That implementation does not grow the buffer. It's built on BufMut::bytes_mut and remaining_mut.

from bytes.

This is now done: https://docs.rs/bytes/0.5.6/bytes/struct.BytesMut.html#impl-Write

from bytes.

@carllerche BytesMut still lacks an implementation of std::io::Write.
What your URL refers to is an implementation of std::fmt::Write, which is something different.

I'd say this issue is still unresolved.

from bytes.

@ColinFinck please see #478

from bytes.

Thanks for the reference, @taiki-e
In the meantime, I'm using BufMut::writer on BytesMut.

from bytes.