Decouple parsing from the Buffer type.
#257
Conversation
This is a step towards #251. Now, the `TryFrom<Buffer>` and `TryFrom<&Buffer>` are no longer implemented. All the remaining code now uses `TryFrom<&[u8]>`.
| @@ -88,7 +88,7 @@ pub trait RequestConnection { | |||
| fds: Vec<RawFdContainer>, | |||
| ) -> Result<Cookie<'_, Self, R>, ConnectionError> | |||
| where | |||
| R: TryFrom<Buffer, Error = ParseError>; | |||
| R: for<'a> TryFrom<&'a [u8], Error = ParseError>; | |||
There was a problem hiding this comment.
Okay, this is something that I would never have come up with. Today I learned about what for<'a> is good for... (I saw this before, but this use here makes a lot of sense to me).
Thanks!
There was a problem hiding this comment.
Unfortunately, I could not get around the single_use_lifetimes lint, so I disabled it.
| @@ -68,7 +68,7 @@ | |||
| //missing_docs, | |||
| private_doc_tests, | |||
| rust_2018_idioms, | |||
| single_use_lifetimes, | |||
| //single_use_lifetimes, | |||
There was a problem hiding this comment.
Sigh. I had hoped this can be done without it.
Do you think deny(single_use_lifetimes) is useful in general? If not, we can just remove this. Otherwise, some #[!allow... is needed instead of just removing it globally.
There was a problem hiding this comment.
I'm afraid these allow would be needed on each implementation of the RequestConnection trait. I think we can go without single_use_lifetimes. I believe it won't be that bad if an avoidable single-use lifetime steps in.
|
Thanks, this definitely looks like a step in the right direction. The whole Do you want to continue working on #251 after this is merged? |
I'll be glad to 😄 |
|
Okay, thanks! |
This is a step towards #251.
Now, the
TryFrom<Buffer>andTryFrom<&Buffer>are no longer implemented. All the remaining code now usesTryFrom<&[u8]>.