ripasso VS semver-trick

Let's say crate B depends on crate A with a pinned dependency, and uses one of its types in a public interface.

Crate C depends on them both. It now can't bring in updates to A until B does, and when B updates that's a breaking change, so it better bump its major version.

Take a look at this teick, for example, for foundational crates updating their major version: https://github.com/dtolnay/semver-trick

Now imagine that being an issue every single patxh update.

Rust modules require a tiny bit more definition up-front, but they neatly decouple the module hierarchy from file layout so you can reorganize code however you like in future, and they support very fine grained control of privacy (such as being able to say pub(super) and pub(crate)). In extreme cases, you can even re-export symbols from one module in another without it counting as a breaking change, so you have even more options for evolving your project without breaking existing consumers. Look at the the semver trick as an example of how powerful this can be and how much freedom it gives library implementors. (And even if you're only a library consumer, wouldn't you rather be consuming libraries by implementors that had more freedom and power?)

A library author concerned about this can use the semver trick. TL;DR: if your current version is 0.42, you can do a 1.0 release, then do a 0.43 release that depends upon your 1.0 release and re-exports all the symbols.

I mean for all the parts of the standard library that do not change, one could presumably use the semver-trick.

The semver trick can help with libraries at least when they go to unify the ecosystem. Release new versions that replicate previous APIs in a compatible way while moving to the standard library implementation.

Because you still run into the problem that's been seen when various important crates upgraded and either didn't use the semver trick or had downstream crates specifying Cargo.toml version requirements too narrowly for it to be effective.

It's called the "semver-trick" [1].

[1]: https://github.com/dtolnay/semver-trick

The JVM is 114MiB on my machine. A near-minimal ggez program in debug mode is about 100MiB,¹ and ggez is small for a Rust application library. When you start getting into the 300s of dependencies (i.e. every time I've ever got beyond a trivial desktop application), you're lucky if your release build is less than 100MiB.

Sure, I could probably halve that by forking every dependency so they aren't duplicating versions, but that's a lot of work. (It's a shame Rust doesn't let you do conditional compilation based on dependency versions, or this would be a lot easier. As it is, we have to resort to the Semver trick: https://github.com/dtolnay/semver-trick/ — not that many people do that, so it's functionally useless.)

¹: I can get it down to around 8MiB with release mode, lto etc., but that significantly increases the build time and only about halves the weight of the intermediate build files.