Add escape hatches to and from quickjs_wasm_sys

[surma]

While quickjs_wasm_rs provides a much more ergonomic API, there are some features of QuickJS that are not exposed. For example, exotic object behaviors or QuickJS’s execution timeout.

In the long-run, it would be great to have those in quickjs_wasm_rs, but in the meantime, it seems like a good idea regardless to me expose escape hatches to the low-level crate.

Specifically:

• Expose the wasm_quick_sys module from quickjs_wasm_rs to avoid users having to add that crate themselves and potentially ending up with different version.
• Add a function to convert raw QuickJS value (u64) into a JSValueRef
• Add functions to convert a *mut JSContext to a JSContextRef and vice versa (in fact, *mut JSContext was already exposed via new_callback, but hard to utilize without the other parts).

[saulecabrera]

For example, exotic object behaviors or QuickJS’s execution timeout.

Could you add a bit more detail about which APIs you'd like to see exposed? I'm not very familiar with "exotic object behaviours" for example.

Generally my stance is that quickjs_wasm_rs should be as safe as it can be; I think this change is not changing that fundamental principle given that the safety around the methods is correctly annotated and we're not enforcing the usage of unsafe code, it's pretty much opt-in.

I'd be good though to have a list of features that you think we're missing in quickjs_wasm_rs to be able to get get rid of these "escape hatches", so I'd appreciate if you could list those more concretely if you can.

[surma]

I'd be good though to have a list of features that you think we're missing in quickjs_wasm_rs to be able to get get rid of these "escape hatches", so I'd appreciate if you could list those more concretely if you can.

I mean the short answer is: All QuickJS features 😅

Specifically for my explorations right now, I want to make use of the following QuickJS features:

• JS_SetInterruptHandler which calls a callback in regular intervals that can choose to terminate the execution (by returning true).
• Exotic Methods, which are a part of a QuickJS class definition. It effectively allows you to implement a JS class whose shape and implementation is backed by Rust. I’m currently working around this by using the ES6 Proxy API, but that is a level of indirection I’d like to avoid in the future :)

In general I think these escape hatches are valuable and I don’t think they really pose a risk, but it’s totally your call.

[saulecabrera]

In general I think these escape hatches are valuable and I don’t think they really pose a risk, but it’s totally your call.

I think they are too IMO, and we should land this, given that we currently don't cover the entire surface area of the QuickJS API. But in the future, ideally we should support all the features in a safe way, hence my comment about not making any stability guarantees. And as mentioned above, these new methods are are not forcing anyone to use any unsafe operations.

Exotic Methods, which are a part of a QuickJS class definition. It effectively allows you to implement a JS class whose shape and implementation is backed by Rust

Oh ok, that makes sense. This is probably going to be useful for #605

[jeffcharles]

Would there be any value in putting the re-export of quickjs-wasm-sys and the escape hatches behind a cargo feature? That might help communicate the stability guarantees in code as well as in documentation.

[surma]

@jeffcharles I don’t think I have seen that be done before, but I like the suggestion.

[saulecabrera]

It seems like one of the bindings tests is failing, I think it could be related to the fix in as_str, because all the other changes seem unrelated.

[jeffcharles]

Please update the CHANGELOG file for quickjs-wasm-rs and update the version of the crate to 3.1.0-alpha.1 since this adds meaningful new features.

[surma]

@jeffcharles @saulecabrera I think this is good to go