Add dumping of domain data

[ericholscher]

This will let us index more interesting data about Sphinx domains in our search.

This will output something like this in the JSON:

    "objects": {
        "/api/v2/version/": [
            "http",
            "get",
            "",
            "get--api-v2-version-"
        ],
        "/api/v2/build/(int:id)/": [
            "http",
            "get",
            "",
            "get--api-v2-build-(int-id)-"
        ],
        "/api/v2/project/(int:id)/active_versions/": [
            "http",
            "get",
            "",
            "get--api-v2-project-(int-id)-active_versions-"
        ],
        "/api/v2/version/(int:id)/": [
            "http",
            "get",
            "",
            "get--api-v2-version-(int-id)-"
        ],
        "/api/v2/project/": [
            "http",
            "get",
            "",
            "get--api-v2-project-"
        ],
        "/api/v2/build/": [
            "http",
            "get",
            "",
            "get--api-v2-build-"
        ],
        "/api/v2/project/(int:id)/": [
            "http",
            "get",
            "",
            "get--api-v2-project-(int-id)-"
        ]
    },

[stsewd]

Honestly, I don't understand this code to the %100, I read http://www.sphinx-doc.org/en/stable/extdev/domainapi.html to have a little more of context p:

Is the example the full output of https://github.com/rtfd/readthedocs.org/blob/master/docs/api/v2.rst?

[stsewd]

Little typo geneate -> generate

[stsewd]

what's the reason to sort the data before? At the end, we save all this in a not-sorted dict

[safwanrahman]

I have build documents of readthedocs using the built in JSONHTMLBuilder. It seems like the searchindex.fjson file may already have all the objects we need, but I am confused about the data its providing.

"objects":{
      "":{
         "/api/v1/user/{id}/":[
            7,
            0,
            1,
            "get--api-v1-user-id-"
         ]
      }""
}

https://gist.github.com/safwanrahman/b7151e8fe47ceab145c700995e075cf6

[safwanrahman]

So I have type of data by reading the source code . Its in following order
index of docnames, key of objtypes, priority, shortanchor (if any).

we already get this file in our build. I think we can parse the information from there and build our index. what do you think @ericholscher ?

[ericholscher]

@safwanrahman We can perhaps just use the existing searchindex.json that is produced in the HTML build. I imagine we might want to output more data in the future, but if we can use that, it's fine.

[ericholscher]

Closing this, as I think all the data we need is already in objects.inv for this PR. So it's already on disk 👍