enhancement: support introspection

[0xmovses]

Closes #790

See the issue for full context, it's all there 👌

Changelog

• Add introspect method on Schema struct.
• Updates query_graph handler function to check for introspection and return that if requested.
• Updates documentation.

Testing

1. cargo run --bin fuel-indexer -- run --manifest packages/fuel-indexer-tests/components/indices/fuel-indexer-test/fuel_indexer_test.yaml
2. In your browser, go to http://localhost:29987/api/playground/fuel_indexer_test/index1
3. Send the query below
   (For extra testing you can test normal queries and other indexers, I did this, but if you're in a QA type of mood... )

query IntrospectionQuery {
  __schema {
    types {
      name
      fields {
        name
        type {
          name
          kind
        }
      }
    }
  }
}

You should get the response

{
  "data": {
    "__schema": {
      "mutationType": null,
      "queryType": {
        "name": "Query"
      },
      "subscriptionType": null,
      "types": [
        {
          "fields": [
            {
              "name": "QueryRoot",
              "type": {
                "kind": "NON_NULL",
                "ofType": {
                  "kind": "OBJECT",
                  "name": "QueryRoot"
                }
              }
            }
          ],
          "kind": "OBJECT",
          "name": "Query"
        },
        {
          "fields": [
            {
              "name": "gas_used",
              "type": {
                "kind": "NON_NULL",
                "ofType": {
                  "kind": "SCALAR",
                  "name": "UInt8!"
                }
              }
            },
            {
              "name": "id",
              "type": {
                "kind": "NON_NULL",
                "ofType": {
                  "kind": "SCALAR",
                  "name": "ID!"
                }
              }
            },
            {
              "name": "blob",
              "type": {
                "kind": "NON_NULL",
                "ofType": {
                  "kind": "SCALAR",
                  "name": "Blob!"
                }
              }
            },
            {
              "name": "result",
              "type": {
                "kind": "NON_NULL",
                "ofType": {
                  "kind": "SCALAR",
                  "name": "UInt8!"
                }
              }
            }
              ...

Note

mutationType and subscriptionType fields have been set to null.

[luizstacio]

Testing the PR locally here, I see the error

{
  "error": "Introspection result missing interfaces: { fields: [[Object]], kind: \"OBJECT\", name: \"Query\" }."
}

[0xmovses]

@luizstacio I forgot to mention in the PR description, that error shows on initial page load, but if you run the introspection query, it should work. Could you try that? Meanwhile I'll see if I can suppress that error on page load.

[ra0x3]

@rvmelkonian

• Just an FYI doesn't seem to work for me either.
• When I open the playground I do see the error @luizstacio mentions
• When I click the "Schema" and the "Docs" tabs, I don't see anything -- I just see the loading spinner which never resolves

[0xmovses]

@ra0x3 I had a sync with @luizstacio late last night. The scope of this PR was just to do the introspection query and get the response back with the schema data (this is used for TS code gen). To get the schema and docs stuff to work in the playground, I would consider a separate PR as that involves editing JS and html from async-graphql crate.

Did you try out the query in the description and get back a successful response?

[ra0x3]

@rvmelkonian Sounds good. I'll defer to @luizstacio for approval in terms of functionality working properly

[luizstacio]

@rvmelkonian Sounds good. I'll defer to @luizstacio for approval in terms of functionality working properly

Docs and Schema tabs are loaded using the return of the introspect query, to be a valid return it should be loading this tabs also, this would give you the certainty that is working.

I was talk with @deekerno yesterday about investigate the possibility to use async-graphql dynamic schemas. In order to replace the current GraphQL Parser. This should enable this features without the need to custom implementations, and also make sure that the response of GraphQL returns comply with GraphQL standard.

I'm mentioning this here, because if @deekerno is looking into it, this work may not be needed?

[ra0x3]

@deekerno Will defer to you with regard to @luizstacio 's comment here.

[0xmovses]

@ra0x3 @deekerno In that case I'll put this PR on hold then #800 as I'm making heavy use of our current graphql parser. I'm all up for shifting to async-graphql, if we can get a compliant graphql endpoint out of the box that can support our needs, but I think this may require some rewrites elsewhere.

[deekerno]

It will definitely require some rewrites; I think it's just a question of how many and how much effort we're willing to put into it. It seems like it may be a net positive as far as type safety, but I'm still investigating prior art to see if we would lose any speed or start running into any sort of "N+1 problem" when it comes to queries. I'll update you all on Slack.

[0xmovses]

@luizstacio the logic for fetching the schema and docs is tucked away inside async-graphql crate. I'm only using that crate here to render out the playground, but everywhere else we are using our own fuel-indexer-graphql-parser package. This is why schema and docs do not load. But when you run the introspection query you should get all the data you need to be able to do TS gen. Let me know if this gives you what you need. 👌

[0xmovses]

Closing this as async-graphql provides the functionality and we will be switching to this shortly.