Feature Request: Batch query support

[zopf]

I'm submitting a ...

• bug report
• feature request
• question

Some other GraphQL server implementations support a batch query protocol, whereby multiple GraphQL requests can be assembled into an array, the array is sent to the server, each of the requests is separately executed on the server, and their results are returned in one array to the client.

Example usage can be found here for Apollo: https://blog.graph.cool/improving-performance-with-apollo-query-batching-66455ea9d8bc

And an example implementation in Apollo server can be found here: https://github.com/apollographql/apollo-server/blob/3fecfcd3218c2f7f2c5ed59a351d71022944bc35/packages/apollo-server-core/src/runHttpQuery.ts#L65

This feature request proposes that PostGraphQL support a similar scheme, in order to allow clients to optionally make GraphQL requests to the PostGraphQL server in a batch mode, enabling multiple operations to be performed with a single HTTP request.

An example use case would be a client that needs to perform ten separate mutations (say, updating ten separate DB objects with new data). With batching, instead of sending ten separate HTTP requests to the PostGraphQL server, the client could send one larger HTTP request that contains an array of those ten mutations, and receive and array of ten results.

[benjie]

I definitely want to support this and it should be fairly easy to add. PRs welcome.

(In the mean time you can work around this by performing multiple mutations in the same GraphQL query:

mutation MultiCreate {
  post1: createPost(...) { ...Payload }
  post2: createPost(...) { ...Payload }
  post3: createPost(...) { ...Payload }
  post4: createPost(...) { ...Payload }
  post5: createPost(...) { ...Payload }
}
fragment Payload on CreatePostPayload {
  ...
}

)

[benjie]

Query batching

[w9]

@benjie Does this workaround guarantee atomicity (my guess is not)? And if not, is there a work around that allows me to have the same atomicity as

INSERT INTO table (c1, c2) VALUES (1,2),(3,4),(5,6);

? Thanks!

[w9]

To answer myself---I found the following solution works:

PostgresQL setup:

create type add_many_people_input as (
  name text,
  age numeric
);

create function add_many_people(
  inputs add_many_people_input[]
)
  returns void
language plpgsql security definer as $$
begin
  insert into person (name, age)
    select * from unnest(inputs);
end;

GraphQL query:

mutation {
  addManyPeople(
    input: {
      inputs: [
        {
          name: "John",
          age: 24
        },
        {
          name: "Jane",
          age: 23
        }
      ]
    }
  ) {
    clientMutationId
  }
}

[wenzowski]

Not sure if this is the right place to consider @export(as: "foo") support like Sangria has. Would be lovely to execute multiple dependent mutations in one go, without having to write volatile functions for every mutation chain.

[benjie]

Query batching support is available now via --enable-query-batching

[barbalex]

@benjie Sounds great. Where is that documented?