search

theam / aws-lambda-haskell-runtime

⚡Haskell runtime for AWS Lambda
https://theam.github.io/aws-lambda-haskell-runtime/
Other
270 stars 48 forks source link

Open questions in Docs #46

Closed ChristophP closed 5 years ago

ChristophP commented 5 years ago

I have a couple of open questions that aren't answered in the docs (https://theam.github.io/aws-lambda-haskell-runtime/index.html) or anywhere else I could find. I think it would be good to mention that as well some in the docs.

  1. The docs say this

    Well, the reality is that Haskell didn’t work as well as the supported languages did. We had to apply ugly hacks to make it work, and we ended up switching to TypeScript or other better supported languages for production projects. But that’s all in the past! Why didn't it fit well before? What changed so that those hacks were no longer needed?

  2. How to do deployment/ do I need a layer? Some time ago there used to be a medium article with some deployment instruction but I am not sure of that is no longer valid with version 2 of this runtime. What changed in version 2?

  3. What's the reason for Haskell being slower than node.js? Seems like a compiled language should be faster. Maybe this is hard to answer without some serious benchmarking.

Thanks for making the runtime.

NickSeagull commented 5 years ago

Hey @ChristophP , thanks for pointing this out 😄

  1. Basically one had to compile an executable/dynamic library and then wrap it around in a Node.js module that performed a native call. Nowadays, one can create their own runtime, like this one, and provide a compiled executable. I don't know if this should be mentioned in the docs or not.

  2. Yes, this should be definitely answered. One has to run a make in their project root, and then upload the generated build/function.zip

  3. To be honest we have not much idea, our guess is that AWS makes some fine tuning with Node.js, by perhaps using V8 processes rather than containers 🤷‍♀️, but that's just a guess. We did some benchmarks and Haskell was slower. Perhaps there is room for optimization in the runtime.

I hope this clears your questions ^_^

ChristophP commented 5 years ago

Awesome, thanks for getting to it so quickly. I left some comments on the PR.