3 ms·
> .. excluding all tests and examples, .. How would you verify that the ported code actually works if you don't port the tests and examples? I see tremendous
by JaggerJo 1y ago
> .. excluding all tests and examples, ..
How would you verify that the ported code actually works if you don't port the tests and examples?
I see tremendous use in a tool that could be used to port "any" library to any language. I'm very skeptical if this works if the library itself depends on other language specific libraries, ... but we'll see.
If you'd seriously want to use that port in production it only makes sense to port the tests and examples. How would you verify the ported code works otherwise? This also means that if the original library has bad test coverage.. you have too.
- JonChesterfield 1y agoIt's much more convincing evidence that the port worked if you run the same tests and examples as before. If you've let the llm change the tests too it's very hard to tell whether the new codebase does the same thing as the original.
- JaggerJo 1y agoYes, that would be even better - but also way harder for unit tests as you’d need to interop between the langauges.
- frumplestlatz 1y agoClaude Sonnet has a funny habit of fixing test failures by deleting the failing tests.
- Anonbrit 1y agoIt really is like an over enthusiastic college intern!
- hn8726 1y ago> How would you verify that the ported code actually works if you don't port the tests and examples? Port tests, run them against unported code. Port code, run it against unported tests? That way you also verify that the public api hasn't changed