reloading VS ncc

the most complicated/interesting thing i figured out was probably the AST manipulation and decorator fuckery required for my rewrite of julvo/reloading

Write a similar thing [1] which can also reload functions from source before every execution but also load loop bodies from source before each iteration. Can be helpful for long running loops like when downloading or training deep learning models.

[1] https://github.com/julvo/reloading

During my search for deploying Lambdas via GitHub actions, I came across a tutorial that utilized ncc for converting TypeScript and bundling. While ncc is effective, I discovered esbuild, which proved to be significantly faster and perfectly suited to my requirements.

You could use ncc to compile a binary:

https://github.com/vercel/ncc

This appears to be caused by lazy loading of image layer data, particularly during container initialization, and latency introduced by the read operations. In general, you want to access as few files and as little data as possible during the initialization of your functions. For example, we've seen improvement when using ncc to bundle Node.js applications.

What we do however, is bundle all these dependencies, and our own code into a single JS file using the nifty ncc tool (thanks Vercel!)

Vercel's ncc compiler

The most unclear and confusing part is packing the code using ncc. The necessity of this step is caused by GitHub’s approach to running your Action.

I filed an issue for the ncc bundler regarding a similar behavior

This may be slightly tangential but I recently discovered ncc[1] from vercel which can take a single node project and compile it and all dependencies to a single file.

As an added benefit it also collapses all contained dependencies license files into a single licenses.txt file too!

- [1] https://github.com/vercel/ncc

The current process of building a package locally is by running through a gulp task, using ttypescript to compile the TS and @vercel/ncc to build the binary: