How to define introspection endPoint for each build variant separately?

[amansaryal]

I use introspection to download schema. The endpoint for which changes for every build variant.
I want to know how can I set the schema endpoint dynamically in my gradle script for each build variant in my app?

[BoD]

Hi! 👋

When configuring your services you can add an introspection block specific to each service.

If you're already using createAllAndroidVariantServices this could be done like this:

apollo {
    createAllAndroidVariantServices(".", "") {
        packageName.set("com.example.apollokotlin.graphql")
        if (name.startsWith("debug")) {
            srcDir(file("src/debug/graphql/"))
            introspection {
                endpointUrl.set("https://example.com/debug")
                schemaFile.set(file("src/debug/graphql/schema.graphqls"))
            }
        } else if (name.startsWith("prod")) {
            srcDir(file("src/prod/graphql/"))
            introspection {
                endpointUrl.set("https://example.com/prod")
                schemaFile.set(file("src/prod/graphql/schema.graphqls"))
            }
        } else if (name.startsWith("release")) {
            srcDir(file("src/release/graphql/"))
            introspection {
                endpointUrl.set("https://example.com/release")
                schemaFile.set(file("src/release/graphql/schema.graphqls"))
            }
        }
    }
}

Note that this creates a service for each variants, including the test ones. If you don't need this granularity you could declare the services manually, like so:

    apollo {
        service("debug") {
            srcDir(file("src/debug/graphql/"))
            packageName.set("com.example")
            outputDirConnection {
                connectToAndroidSourceSet("debug")
            }
            introspection {
                endpointUrl.set("https://example.com/debug")
                schemaFile.set(file("src/debug/graphql/schema.graphqls"))
            }
        }
        service("release") {
            srcDir(file("src/release/graphql/"))
            packageName.set("com.example")
            outputDirConnection {
                connectToAndroidSourceSet("release")
            }
            introspection {
                endpointUrl.set("https://example.com/release")
                schemaFile.set(file("src/release/graphql/schema.graphqls"))
            }
        }
    }

This will add gradle tasks named downloadDebugApolloSchemaFromIntrospection, downloadReleaseApolloSchemaFromIntrospection, etc.

[amansaryal]

Thanks! that works.
However, this relies on an implementation detail which I think is bad. Could we not have the variant name passed to the action block?

[BoD]

If "" is passed as the suffix, then the service name will be the variant name.

It's true that using this relies on knowing the implementation though. Passing the variant name to the lambda is a good idea, but it may be difficult to make this change in a compatible way. It may be something we can add to the next major version though!

[amansaryal]

Sounds good.
Closing this issue now. Thanks!

[martinbonnin]

@amansaryal Out of curiosity, how different are your different schemas? Do you also need different queries per variant?

If your use case is debug using the pre-prod schema and release using the production one, then I'd recommend using the 2nd approach from @BoD above, specifying explicit "debug" and "release" services.
We found out that using createAllAndroidVariantServices is a waste in most cases because it will create too many codegen tasks most of the times (debug, release, debugUnitTest, debugAndroidTest in a minimal setup) when all you need is 2 codegen tasks.

[amansaryal]

@martinbonnin correct. The second approach is what I'm going with as well. createAllAndroidVariantServices not only creates too many tasks, it also prevents me from keeping a common srcDir for my schema (because of the code below)

variant.sourceSets.forEach { sourceProvider ->
          service.srcDir("src/${sourceProvider.name}/graphql/$sourceFolder")
        }

I'm assuming that with schemas in different srcDir for each variant, I'll also have to copy my query and mutation files to all of those directories, which is unnecessary since those queries/mutations are not going to change.

Here's the use case in a nutshell:

I want a common srcDir for each build variant.
Ex: src/main/graphql

I should be able to configure introspection endpoint for each build variant.
Introspection should overwrite the schema.[ext] in my srcDir whenever I switch variants.

[martinbonnin]

Gotcha. Many thanks for the additional context 🙏 .

Introspection should overwrite the schema.[ext] in my srcDir whenever I switch variants.

I think that's the key point. If you have shared queries with different schemas, ideally you should put the schemas outside your queries dir:

src/main/graphql/query1.graphql
src/main/graphql/query2.graphql
...

src/debug/graphql/schema.graphqls
src/release/graphql/schema.graphqls

In that case, I'd recommend doing something like this (factoring code using a nice Kotlin function):

fun ApolloExtension.createService(sourceSet: String) {
  service(sourceSet) {
    srcDir(file("src/main/graphql/")) // You can actually omit this since it's the default
    schemaFile.set(file("src/$sourceSet/graphql/schema.graphqls")) // This is required
    packageName.set("com.example")
    outputDirConnection {
      connectToAndroidSourceSet(sourceSet)
    }
    introspection {
      if (sourceSet == "debug") {
        endpointUrl.set("https://example.com/debug")
      } else {
        endpointUrl.set("https://example.com/release")
      }
      schemaFile.set(file("src/$sourceSet/graphql/schema.graphqls"))
    }
  }
}
apollo {
  createService("debug")
  createService("release")
}

All in all, I think we could certainly deprecate createAllVariantServices. This was added in an effort to stay backward compatible with v2 but I think this is misleading in most cases.

[amansaryal]

So my assumption was wrong, we can indeed keep the queries and the schema in different places. Gotcha!
Thanks for this @martinbonnin!

[martinbonnin]

Yes indeed! The plugin tries to find a schema in srcDir by default but you can override this with schemaFile