go.mod support

[rcarmo]

Needs to be done now that it's stabilised.

[psugand]

Any plans for implementing this?

[rcarmo]

Yes, but no schedule. We tend to implement (or accept PRs) when people are actively using this and can do troubleshooting, and right now I have more Rust projects than Go ones (and haven't yet tackled that), so... contributions are welcome. It really shouldn't be very hard, but I have very little bandwidth currently.

[psugand]

I don't mind giving it a go if you need a pair of extra hands. I will explore the code and let you know if I have any questions.

If you got any pointers or quick overview which would expedite things please share.

[rcarmo]

Sure, you can look at #344 for an example. The Go detection and deployment functions are just beside the ones I linked to in that, and the key thing is trying to make sure that GOPATH and equivalents are set to a place that is easy to clean up when the app is destroyed.

[psugand]

Thanks.

How would you propose the actual execution of applications to work?

I can see you have a Procfile defined, however, when deploying a Go app would you want the application to be compiled as a binary and then execute the binary ? or simply let the user specify the run path by pointing to their main.go, i.e. go run cmd/main.go

[rcarmo]

We don't need to be very prescriptive... For compiled languages, we typically use the release worker to run the build (if that fails, the app isn't (re)started). And then you can just invoke the resulting binary.

For simple stuff, go run is fine. One thing you can do is look at the sample repos (go back to the org tree, we try to have at least one sample per runtime) and build a tiny sample repo - I'd also welcome having that, since I hope to some day use those sample repos for end-to-end CI testing...

[rcarmo]

Here's an example: https://github.com/piku/sample-bare-metal-c/blob/master/Procfile

[chr15m]

That brings up an alternative way to deploy go apps that is already compatible with Piku. What you do is create a gh-pages type of branch called e.g. 'live' and you commit the compiled binary in there and use that to deploy to Piku. Alternatively you can commit the compiled binary into a sub-directory. Some people find the idea of committing binary artifacts a bit weird but I think there are actually advantages to this. If you go down this route check out the 'release' worker which you can use to do set up things etc.

[rcarmo]

I don't judge :)