Ask HN: Is there a programming language that has this? | Hacker News Reader

Ask HN: Is there a programming language that has this?

3 pointstechsin1014 comments

I was thinking about cloud architecture, serverless to be specific. That how it sucks to develop in them but but it allows for really good scaling. Then I wondered, is there a programming language where every function call could be a serverless function.

Imagine a compiler like chrome v8 engine which analyzes code on runtime and figures out which code path are more resource intensive and manage them. So what if same way compiler notices that certain code path executions take longer and start spreading them on other threads/cores/machines like kubernetes and orchestrating all of it behind the scene.

So you as a developer don't worry about that.

Let's take javascript

suppose every promised function had potential to be it's own serverless function, depending on resources usage, compiler can either run it on main thread or delegate to cluster.

let data = await getData();// compiler sees getData takes time and is put its own serverless function or something.

I feel this could be ground breaking. Automatic app scaling based on runtime analysis of code paths.

We could have do this right now, with external a library in js using something like this...

let data = await SF(getData(123)); //SF spawns new serverless function based on body of SF...

but it'd be better if it was implemented at language level. I'd like it if chrome team did it.

This makes perfect sense. The closest what comes to my mind is CDK and bash.

CDK (Cloud Development Kit) lets you orchestrate creation of lambdas with TypeScript code, so in theory you can have one codebase with a file describing what methods to export and it’d provide a method to call the lambda with await. See example here: https://github.com/aws-samples/aws-cdk-examples/blob/master/...

And bash, because it does exactly this — spawns a process with an input for almost any work. If you think of the most basic bash commands, say “git ... | grep ... | sort ... | tee” you end up with 4 concurrent processes piping data from left to right. Compose bash functions instead of long lines and you got yourself a maintainable codebase. I understand this scales to one instance size, but you could throw 64 cores at the problem, writing many processes to work together, utilising all CPUs. Thinking outside of bash — using “await spawn()” in node you could start a new node process with execution of another function, just matter how you export them, or if a decorator can stringify the function.

Alternatively, you can create single lambda that calls any function in the codebase. Your decorator can register the function, and when lambda runs, it checks all registered functions and calls the one matching the request.path. For something more advanced you can build custom runtime which defines exactly what lambda executes: https://docs.aws.amazon.com/lambda/latest/dg/runtimes-custom...

I've been doing serverless in anger for years. My vision is to use annotations or decorators to tag functions and have the implementation transparently handled.

@spawn //makes this a lambda

function fn() ... {}

@queue //makes this a queued function sqs or similar

function fn() ... {}

im thinking along the same lines, it'd be even better if there was built in retrying or event based message queuing, time out, and default error message if it really fails for all funcs .... await fn(); // -> {err: 'timedout'}

https://www.youtube.com/watch?v=1vEjGKhSx8o

Reply on news.ycombinator.com