doteric/funpack ?

[doteric]

Hey 👋
I just noticed your package.
Would https://www.npmjs.com/package/funpack maybe solve your problem also by the way? It's using esbuild under the hood so it might speed up your build process a little bit. Maybe also the additional options would work a little bit better for you?
I wonder for what exactly are you building functions for? Maybe I could improve my package to suit your need.

Cheers and have a nice day.

[WillBeebe]

Hi 👋

Your package looks like it accomplishes the same thing mine does but has way more functionality! This implementation was originally for GCP's Cloud Functions since those have to target a specific version of node (much like AWS Lambdas).

For the most part I've Dockerized everything and don't use this much anymore BUT where I do I'll look to replace my package with yours. Thank you for letting me know about it. 🥂

[doteric]

Hey @WillBeebe

Nice to hear, please let me know if you have any problems and if it works all fine with GCP, as I have never used GCP ;D

Btw. Pure functions instead of dockerized functions seem to work a lot better on AWS Lambda, that's not the case for GCP?

[WillBeebe]

That's a good question!

Since Cloud Functions was created as a product, Cloud Run has been introduced. Cloud Run requires a Docker image and behaves in sort-of the same way as Cloud Functions. In more recent times Cloud Functions V2 has been introduced and that's just Cloud Run under the covers. GCP has consolidated on the Cloud Run abstraction for both of those services.

If you don't use Docker w/ Cloud Functions (v1 or v2), I believe what happens in the CLI or UI is that a Cloud Build pipeline gets kicked off. That pipeline uses buildpacks to create a Docker image for you based off of a public base image. So on the Google side... it's containerized either way.

As far as I know, Lambdas does let you bring your own Docker image but if you use "pure functions" like you're saying it's most analogous to GCP's Cloud Functions V1. I'm not sure if Lambdas uses buildpacks under the covers or their own abstraction for the pure functions deployment process.

[doteric]

That's very interesting, it works pretty different in AWS and docker lambdas are a bit slower especially on cold start.
I assume that npm package won't be as useful for GCP Functions as it is for AWS Lambdas in a case where docker images are preferred.

Anyway, thanks you for the detailed explanation, really appreciate it and wish you a great day :D

[WillBeebe]

Anytime :) 🥳