Create a CLI Client plugin.

[mtaylor]

This should take the config file that the API scraper classes creates and generate a CLI client.

[dmitri-d]

is this the one where we crawl the api? if so, there's a common component for both doc and client generator...

[mtaylor]

Actually no this is the piece that generates a CLI from the the config file that the crawler spits out.

[mtaylor]

I envisage we have many of these that generate clients for many different languages. Ruby, Java, Python etc...

[dmitri-d]

Hmm. I think it makes sense to generate docs like that; I think generating client (when there's no custom code involved) from a template makes less sense: this would result in needing to regenerate the client code every time the server API changes. We can skip this step and crawl the api as needed (possibly with caching results to speed up the client).

There's a bit of disconnect here though - docs need to be re-generated, while the client does not. This is possibly an indication that api docs belong together with api (or at least the change/life cycle of api docs is the same as that of api itself).

[mtaylor]

I'm not 100% I understand what you are saying here. We should have a chat in IRC to hash this out in more detail. What I envisage though is 3 components.

1. An upfront scraper class that outputs a description of the API (Again, I'm not sure if this will work since we may need to do this on the fly)

2. A client generator for multiple languages. Given we have 1) we can use this description to generate clients against a particular version of the API. I would argue that people want a fixed API to develop against though this view maybe different to what you have in mind.

3. The documentation generator. This again would be generated from 1) and would essentially be the documentation for a particular version of the API.

For 1) We may just want to make the generation of these descriptions manual. In the case of conductor for example. We probably want to generate client libs for Python, Java and Ruby and also CLI. If we have a single description of our API we can automate the building of these. We could update the description each time we release a new version of the API.

I take it you are thinking of something more dynamic? Where the client lib creates models as they are discovers as the client walks through the API?