Improve accessibility of project/tutorial project templates

[vados-cosmonic]

Currently, new users must clone the component-docs repo, head to a specific sub-directory then attempt to build projects.

It would be ideal if there were easier ways to download and set up resources to walk through the tutorials.

While we probably want to avoid building custom tooling in every possible language (different languages have different scaffolding mechanisms), we need a solution that is a bit less nested.

There were several good suggestions made, one of which being a single repo that contains all the languages at the top level (thanks @mkatychev), which is a bit easier to clone and filter out.

Another suggestion is that this could actually be a part of #219 -- having a "Current" release that always contains tarballs of the current code for any given language might make it easy to download and extract (ex. rust-starter.zip or js-starter.zip).

We could also rely on language-specific templating functionality, but this would likely require ensuring that it exists in the first place for all related languages (i.e. cargo component new, jco new, componetize-py new, etc), or in a shared tool like wit-bindgen which may not be feasible.

[mkatychev]

I vote (with my bias) for a broad and shallow git repo.

Tree-sitter takes this approach to address consuming the same grammar (subsitute for WIT) across many languages. tree-sitter init initializes a new project using the everything-but-the-kitchen-sink approach of including all bindings at a shallow depth¹ and keeps all language manifests (ex: go.mod, Cargo.toml, package.json) together in the directory root ².
Though this approach adds congestion, the gain of eliminating navigrational complexity (where to cd) is worth it since language specific needs are always resolved from the project root.

Ex - consuming/using a tree-sitter JSON grammar for go and javascript:

Javascript:

git clone https://github.com/tree-sitter/tree-sitter-json.git && cd ./tree-sitter-json
npm install

Go:

git clone https://github.com/tree-sitter/tree-sitter-json.git && cd ./tree-sitter-json
go install

Footnotes

1. JSON grammar bindings
   https://github.com/tree-sitter/tree-sitter-json/tree/master/bindings ↩

2. https://tree-sitter.github.io/tree-sitter/cli/init.html#binding-files ↩

[kate-goldenring]

I lean towards whatever is easiest to maintain. This repo is one nested level deeper than needed since originally this repo was intended to be used for multiple books, so one initial step may be to flatten the component-model directory.

I believe -- i may be misremembering -- the only time you need to clone anything is to pull down the example host to run the adder component. Instead, we could publish that as a package somewhere. The other current implementations that exist in the repo are example implementations. To this, i think another thing this issue brings up is how do we want readers to interact with these examples? As an answer sheet (current day) or as a quick start.

[vados-cosmonic]

Just to note it here -- one more thing we should probably add to this repo is tests and CI for the individual languages

[vados-cosmonic]

An update here -- after chatting with @yoshuawuyts a bit it looks like there are a couple things to consider here and some proposals for how we should move forward:

• For Templates, per-language repos under the BCA should likely be the way forward, so we can preserve the ability for anyone to easily click-to-clone (this is the work that's already underway, and will likely be HTTP-focused per-language for the forseeable future)
• For Examples, no strong opinion but I think here we can combine examples into a single repo, that we can point to & keep under test. These are more to show people working examples of some paradigm (ex. shared-everything linking, composition, etc)

[EDIT] - The example repo should also contain tests, as was planned.