Delete wasmtime.loader and wasmtime.bindgen
#310
Conversation
This commit deletes two nice-to-have but difficult-to-maintain modules in this repository. Specifically: * The `wasmtime.loader` module is deleted as it was a proof-of-concept but doesn't map well to Python and Wasmtime. For example bytecodealliance#302 shows how destruction of Python modules isn't handled correctly. In general it's a "cute" implementation but isn't too too useful and doesn't map well to how Wasm is embedded in many contexts. I'm deleting this to reduce maintenance burden on this repository. It would be ok to reimplement in the future but it would likely require someone with more Python knowledge than I. * The `wasmtime.bindgen` module is deleted in favor of component support in bytecodealliance#308. This module still makes sense to have but it would need a ground-up rewrite to work with component types rather than component definitions. This is deferred to future work in bytecodealliance#309 and in the meantime it's removed to avoid any confusion about what should be used and what shouldn't. Closes bytecodealliance#105 Closes bytecodealliance#106 Closes bytecodealliance#107 Closes bytecodealliance#108 Closes bytecodealliance#109 Closes bytecodealliance#119 Closes bytecodealliance#143 Closes bytecodealliance#178 Closes bytecodealliance#181 Closes bytecodealliance#197 Closes bytecodealliance#202 Closes bytecodealliance#218 Closes bytecodealliance#245 Closes bytecodealliance#282 Closes bytecodealliance#302
|
Another point of clarification on this: I'm sort of the only maintainer in this repo and there's not really a lot of process in place right now for larger changes like this. Given that I'm "doing this by fiat" because I think it's a good idea to do. That being said I'm happy to continue to discuss further with others if they feel this is a bad move and/or similarly. One thing I'd ask for though is that if you'd like to keep seeing a feature like this I'd request that you also be willing to take ownership of the implementation and share in maintenance. I am not a Python expert and often get quite frustrated with maintaining this repository as a result. I'm doing the best I can, but I can only do so much as a one-man-show who primarily writes Rust. |
1 participant
This commit deletes two nice-to-have but difficult-to-maintain modules in this repository. Specifically:
The
wasmtime.loadermodule is deleted as it was a proof-of-concept but doesn't map well to Python and Wasmtime. For example "TypeError: 'NoneType' object is not callable" during calls to wasmtime.Managed.__del__ #302 shows how destruction of Python modules isn't handled correctly. In general it's a "cute" implementation but isn't too too useful and doesn't map well to how Wasm is embedded in many contexts. I'm deleting this to reduce maintenance burden on this repository. It would be ok to reimplement in the future but it would likely require someone with more Python knowledge than I.The
wasmtime.bindgenmodule is deleted in favor of component support in Bind the C API for components #308. This module still makes sense to have but it would need a ground-up rewrite to work with component types rather than component definitions. This is deferred to future work in Reimplement bindgen for components #309 and in the meantime it's removed to avoid any confusion about what should be used and what shouldn't.Closes #105
Closes #106
Closes #107
Closes #108
Closes #109
Closes #119
Closes #143
Closes #178
Closes #181
Closes #197
Closes #202
Closes #218
Closes #245
Closes #282
Closes #302