Code size and dynamic dispatch about borsh-rs HOT 2 OPEN

This also matters quite a bit for the primary use-case of borsh -- smart contracts compiled for WASM. And yeah, what you are saying makes sense -- there's indeed a tradeoff here between code-bloat and runtime performance (see serde vs deser (https://github.com/mitsuhiko/deser)).

I don't think anyone fully explored these tradeoffs for borsh though.

An obvious short-term win would be to fork borsh and publish a borsh-dyn crate which implements the same data format as borsh, but using a different implementation strategy. Would gladly link that from this crate's readme/docs!

In an ideal world, this would be a feature of the main borsh crate, but I think exentining existing APIs to support that nicely wouldn't be easy, mostly because today's APIs don't seem to be in a particularly good shape: #51.

from borsh-rs.

An obvious short-term win would be to fork borsh and publish a borsh-dyn crate which implements the same data format as borsh, but using a different implementation strategy. Would gladly link that from this crate's readme/docs!

Sounds like a smart way to go about this. Ideally at least for the common case with derived impls, it could be made to be an almost drop in replacement. I'll give this a go and see how it feels.

In an ideal world, this would be a feature of the main borsh crate, but I think exentining existing APIs to support that nicely wouldn't be easy, mostly because today's APIs don't seem to be in a particularly good shape: #51.

Yeah I was actually aiming for that initially, but quickly realized it would get pretty messy.

from borsh-rs.