will this crate ever be in the standard library?

[BurntSushi]

Originally asked by @sam0x17:

With the recent absorption of once_cell into std, do you think there is a path forward for regex eventually becoming part of the rust standard library? If not, why, especially considering the fact that most languages include some sort of regex facility built into the language?

[BurntSushi]

I don't think regex will ever be part of std. At the very least, I would oppose it. Given I'm on libs-api and the maintainer of regex, that probably means that, as a matter of practice, it has no path to std. I could have my mind changed (such things are known to happen), but I've held this position for several years now.

My little anecdote for this is that, once upon a time, the regex crate lived in rust-lang/rust. This was before Rust 1.0, and while the crate wasn't actually part of std, the regex crate (among other crates, such as rand) was shipped with Rust. This also pre-dated Cargo, so the Rust distribution tried to strike a balance between offering a small standard library but with some conveniences since re-using others' Rust code at that point in time was much harder and more annoying.

After I submitted the PR to add regex to rust-lang/rust, it stagnated for quite some time. I honestly found the process of contributing in that environment to have a lot of overhead. So I just kind of forgot about it and did other things. But in the run up to Rust 1.0, the regex crate was moved out to its own repository and published as a distinct crate. It was the best thing that ever happened to it, because once it did, I found it much easier to work on and iterate. I'm not sure if that would have ever happened if it moved into std.

Still, that was then and this is now. Perhaps now it's "mature" enough that it could be moved into std. So, concretely speaking, here are my main reasons against putting regex into std as of today:

• The API surface area is quite large. There are a lot of types and methods. It's rare for us to stabilize something in std this big. Critically, we generally treat std as something that cannot have breaking changes. So if it goes into std, then that's it. There is no 2.0. Since regex lives as a crate, I can publish a 2.0 version. Perhaps that will happen some day, perhaps years from now, but that's a much shorter time scale than "never."
• The implementation is huge. If you include all of its dependencies, it's just about 100,000 source lines of code. That's a huge maintenance burden to throw into std.
• I think there is still a lot of evolution on its internals that can happen, particularly among its dependencies. Speaking for me personally, this is much easier when it lives as a distinct repository from rust-lang/rust.

I feel like I'm forgetting something, but I think that probably covers it.

One possible alternative is to put something like regex-lite into std. Its API surface area is still quite large, but its implementation is an order of magnitude smaller. And since it specifically eschews performance and (some) functionality, its future evolution is much much much more limited. With that said, it is, as a result, much slower and missing crucial functionality one might expect from Rust's standard library (such as Unicode support). So while it might be nice to have basic regex support in std, I can already see the Stack Overflow answer saying, "don't use regex in std, it's slow, instead use the regex crate."