RUST-1887 Convince append to be more accepting #541
Conversation
| /// Types that can be consumed to produce raw bson references valid for a limited lifetime. | ||
| /// Conceptually a union between `T: Into<RawBson>` and `T: Into<RawBsonRef>`; if your type | ||
| /// implements either of those you should consider adding an impl for this as well. | ||
| pub trait BindRawBsonRef { |
There was a problem hiding this comment.
Because this is consuming a value to produce a reference, it couldn't just be a straightforward Into-style converter; for owned types that would invalidate the backing storage for the reference we're trying to get. Instead I ended up with this mildly loopy way to consume a value to pass a reference to a closure, which bounds the lifetime.
I named it bind because in a past life I was a Haskell programmer and this is giving me flashbacks to the infamous monadic bind operator.
src/raw/document_buf.rs
Outdated
| }; | ||
| } | ||
|
|
||
| raw_bson_ref_from_impls! { |
There was a problem hiding this comment.
This pair of lists are writing out all of the From impls for RawBsonRef and RawBson, with preference given to RawBsonRef when there's overlap; merging these lists is what the new trait enables.
There was a problem hiding this comment.
Does the overlap prevent blanket impls for T where T: Into<RawBson> and T where T: Into<RawBsonRef>?
There was a problem hiding this comment.
We can't do both but I realized we could do one, so I've revised this to have a blanket impl for T: Into<RawBsonRef> and a manual list jsut for RawBson.
| @@ -188,7 +188,6 @@ impl RawDocumentBuf { | |||
| /// result in errors when communicating with MongoDB. | |||
| /// | |||
| /// If the provided key contains an interior null byte, this method will panic. | |||
There was a problem hiding this comment.
Can we add some language here about the acceptable input types and show a borrowed value being appended in the code example?
src/raw/document_buf.rs
Outdated
| }; | ||
| } | ||
|
|
||
| raw_bson_ref_from_impls! { |
There was a problem hiding this comment.
Does the overlap prevent blanket impls for T where T: Into<RawBson> and T where T: Into<RawBsonRef>?
RUST-1887
This turned out to be a bit more of an adventure than I was expecting. Because we (still 😞) don't have specialization, I couldn't directly express "accept anything that's
impl Into<RawBson>orimpl Into<RawBsonRef>", so this PR implements the next best thing. It's a bit ugly but I lean towards it being worth it for the better user experience (appendis the core of therawdocmacro so this will make that transparently accept a wider range of types).