Feature: Consider providing StrBytes interface

[Diggsey]

This would act like the Bytes object but would guarantee that the contents are valid utf-8. It would not implement Buf but would deref to &str and have conversions to the String type.

There would be TryFrom/TryInto conversions from Bytes to StrBytes which would scan the string for validity but would not copy it. There would be zero-cost From/Into conversions from StrBytes to Bytes. There would be an unsafe zero-cost conversion method from Bytes to StrBytes for when the bytes are already known to be valid.

It would be convenient (but not essential) if there was a way to general substring-returning operations on a StrBytes that themselves returned a StrBytes.

This would enable zero-copy string deserialization in the same way as for raw bytes.

[carllerche]

Does something like this work? https://docs.rs/string/0.2.1/string/

[Diggsey]

Yes. 👍

[stackinspector]

The String<T> in the crate string has no serde support implementation. When I tried to implement support for serde, I found that not knowing whether the internal bytes are borrowed or owned or something like Bytes could lead to additional overhead during deserialization. It would be much easier to implement serde support if a wrapper was created on a deterministic Bytes.

[Darksonn]

The bytestring crate provides serde support.