Generate chi based http handler

[dannymidnight]

This is a first pass at adding chi support following on from #2 to generate an alternative to the echo based routing. There's still a bunch of missing pieces but I wanted to push it up early for general feels.

Why?

The current http handler solution is built around echo and requires the calling application to bring along this dependency. Chi is "100% compatible with net/http" making it self-contained and simple to integrate with existing go applications whilst also leaning heavily into the core context API making it more "go-like".

Example

Here's an example of how you could go about registering your operations. You'll note that there's no peer dependency on chi as ChiHandler returns a http.Handler.

Implementation is definitely a little verbose having raw request, response and passing around Context but I think the http.Handler interface will be more flexible in the long run and should be familiar for most go developers.

type Operations struct {}

func (*Operations) FindPets(w http.ResponseWriter, r *http.Request) {
  params := petstore.ParamsForFindPets(r.Context())
  ...
}

func (*Operations) DeletePet(w http.ResponseWriter, r *http.Request) {
  params := petstore.ParamsForFindPets(r.Context())
  ...
}

func (*Operations) AddPet(w http.ResponseWriter, r *http.Request) {
  ...
}

func (*Operations) FindPetById(w http.ResponseWriter, r *http.Request) {
  ...
}

func Handler() {
  myApi := Operations{}
  petstore.ChiHandler(myApi)
}

Questions

• Is this something you're open to incorporating?! 😁
• Rename ChiHandler to Handler? Chi is really an implementation detail that doesn't need to be exposed externally.

TODO:

• Separate out templates from existing echo based ones.
• Add new flag for generating chi http handler
• Add example implementation
• Add some test coverage!

[deepmap-marcinr]

I've been thinking about this, and it's hard to do it cleanly, but your approach looks pretty nice.

Right now, we have a fair bit of code that's router agonstic (all the type generation), and what's echo specific is the server interface (the client interface is agnostic). So, I would refactor the functionality so that the entirety of Echo bindings are in, say, echo-serverinterface.tmpl and echo-wrappers.tmpl, and you could produce chi-specific templates. You would select the backend via a command line option, which would substitute in chi-whatever.tmpl in place of those two files.

I initially chose Echo because it's quick, and it simplifies a lot of stuff, and it works for us, but I'm not particularly attached it to one way or another.

[deepmap-marcinr]

One practical thing that I've found is that it's good to have an interface to implement in your generated code - for mocking, or whatever. So your operations{} struct might want to implement an interface.

I'm totally open to incorporating this, if we can constrain the chi-specific code to a few .tmpl files, just as we've done with Echo.

[dannymidnight]

Great. I'll try clean things up a bit and let you know when it's ready for review.

[dannymidnight]

One practical thing that I've found is that it's good to have an interface to implement in your generated code - for mocking, or whatever. So your operations{} struct might want to implement an interface.

Yep makes sense, in my petstore example the operations struct is implementing this interface.

[dannymidnight]

Hey @deepmap-marcinr I think this is now ready for review.

Major changes since last time:

• There's now a chi-server flag for generating the new implementation and I've kept the echo codepath using server for backwards compatibility.
• I've added an example implementation and test runner following a similar approach to the echo based tests.
• I've split the petstore examples into the following structure with the client split out from the server implementations:

- petstore
-- chi
-- echo
-- petstore-client.gen.go

Let me know what you think! Any and all feedback welcome.

[deepmap-marcinr]

Solid work! Thank you.

[deepmap-marcinr]

Hmm, that's an interesting way to structure tests. I've not seen this approach before.

[dannymidnight]

Cheers @deepmap-marcinr!

[ashwinrs]

Thank you for this @dannymidnight