concurrent-resolvers example is not correct

[aneeskA]

As per https://github.com/graphql-go/graphql/issues/592 if the resolver is made as a thunk, they will be concurrent. To confirm I took the example given at https://github.com/graphql-go/graphql/blob/master/examples/concurrent-resolvers/main.go and changed QueryType to this

// QueryType fields: `concurrentFieldFoo` and `concurrentFieldBar` are resolved
// concurrently because they belong to the same field-level and their `Resolve`
// function returns a function (thunk).
var QueryType = graphql.NewObject(graphql.ObjectConfig{
	Name: "Query",
	Fields: graphql.Fields{
		"concurrentFieldFoo": &graphql.Field{
			Type: FieldFooType,
			Resolve: func(p graphql.ResolveParams) (interface{}, error) {
				var foo = Foo{Name: "Foo's name"}
				fmt.Println("concurrentFieldFoo: sleeping for 5 secs...")
				time.Sleep(5 * time.Second)
				fmt.Println("concurrentFieldFoo: sleeping for 5 secs... done")
				return func() (interface{}, error) {
					return &foo, nil
				}, nil
			},
		},
		"concurrentFieldBar": &graphql.Field{
			Type: FieldBarType,
			Resolve: func(p graphql.ResolveParams) (interface{}, error) {
				var foo = Bar{Name: "Bar's name"}
				fmt.Println("concurrentFieldBar: sleeping for 5 secs...")
				time.Sleep(5 * time.Second)
				fmt.Println("concurrentFieldFoo: sleeping for 5 secs... done")
				return func() (interface{}, error) {
					return &foo, nil
				}, nil
			},
		},
	},
})

concurrentFieldFoo and concurrentFieldBar have the same logic - deviating from the sample code where there was a go routine that was created to send the data in later - where both the resolvers sleep for 5 seconds.

In my output, I was expecting to see, the following:

concurrentFieldFoo: sleeping for 5 secs...
concurrentFieldBar: sleeping for 5 secs...
concurrentFieldFoo: sleeping for 5 secs... done
concurrentFieldFoo: sleeping for 5 secs... done

But I am seeing this:

concurrent-resolvers git:(master) ✗ go run main.go
concurrentFieldFoo: sleeping for 5 secs...
concurrentFieldFoo: sleeping for 5 secs... done
concurrentFieldBar: sleeping for 5 secs...
concurrentFieldFoo: sleeping for 5 secs... done
{"data":{"concurrentFieldBar":{"name":"Bar's name"},"concurrentFieldFoo":{"name":"Foo's name"}}}%      

This is not what I understood from https://github.com/graphql-go/graphql/pull/388 .

What am I missing here?

Full source attached here

package main

import (
	"encoding/json"
	"fmt"
	"log"
	"time"

	"github.com/graphql-go/graphql"
)

type Foo struct {
	Name string
}

var FieldFooType = graphql.NewObject(graphql.ObjectConfig{
	Name: "Foo",
	Fields: graphql.Fields{
		"name": &graphql.Field{Type: graphql.String},
	},
})

type Bar struct {
	Name string
}

var FieldBarType = graphql.NewObject(graphql.ObjectConfig{
	Name: "Bar",
	Fields: graphql.Fields{
		"name": &graphql.Field{Type: graphql.String},
	},
})

// QueryType fields: `concurrentFieldFoo` and `concurrentFieldBar` are resolved
// concurrently because they belong to the same field-level and their `Resolve`
// function returns a function (thunk).
var QueryType = graphql.NewObject(graphql.ObjectConfig{
	Name: "Query",
	Fields: graphql.Fields{
		"concurrentFieldFoo": &graphql.Field{
			Type: FieldFooType,
			Resolve: func(p graphql.ResolveParams) (interface{}, error) {
				var foo = Foo{Name: "Foo's name"}
				fmt.Println("concurrentFieldFoo: sleeping for 5 secs...")
				time.Sleep(5 * time.Second)
				fmt.Println("concurrentFieldFoo: sleeping for 5 secs... done")
				return func() (interface{}, error) {
					return &foo, nil
				}, nil
			},
		},
		"concurrentFieldBar": &graphql.Field{
			Type: FieldBarType,
			Resolve: func(p graphql.ResolveParams) (interface{}, error) {
				var foo = Bar{Name: "Bar's name"}
				fmt.Println("concurrentFieldBar: sleeping for 5 secs...")
				time.Sleep(5 * time.Second)
				fmt.Println("concurrentFieldFoo: sleeping for 5 secs... done")
				return func() (interface{}, error) {
					return &foo, nil
				}, nil
			},
		},
	},
})

func main() {
	schema, err := graphql.NewSchema(graphql.SchemaConfig{
		Query: QueryType,
	})
	if err != nil {
		log.Fatal(err)
	}
	query := `
		query {
			concurrentFieldFoo {
				name
			}
			concurrentFieldBar {
				name
			}
		}
	`
	result := graphql.Do(graphql.Params{
		RequestString: query,
		Schema:        schema,
	})
	b, err := json.Marshal(result)
	if err != nil {
		log.Fatal(err)
	}
	fmt.Printf("%s", b)
	/*
		{
		  "data": {
		    "concurrentFieldBar": {
		      "name": "Bar's name"
		    },
		    "concurrentFieldFoo": {
		      "name": "Foo's name"
		    }
		  }
		}
	*/
}

[aneeskA]

@chris-ramon Can you please take a look at this?

[LLLLancelot]

i guess 'concurrent-resolvers' means resolve logic runs concurrently with the graphql schedule, and the schedule will wait and handle all promising resolve functions. Which Not means resolve func is run concurrently by the schedule, you should setup concurrently tasks by yourself

[LLLLancelot]

here is my example:

"concurrentFieldFoo": &graphql.Field{
    Type: FieldFooType,
    Resolve: func(p graphql.ResolveParams) (interface{}, error) {
        var foo = Foo{Name: "Foo's name"}
        fmt.Println("concurrentFieldFoo: sleeping for 5 secs...")
        tick := time.NewTimer(time.Second * 5)
        return func() (interface{}, error) {
            <-tick.C
            fmt.Println("concurrentFieldFoo: sleeping for 5 secs... done")
            return &foo, nil
        }, nil
    },
},
"concurrentFieldBar": &graphql.Field{
    Type: FieldBarType,
    Resolve: func(p graphql.ResolveParams) (interface{}, error) {
        var bar = Bar{Name: "Bar's name"}
        fmt.Println("concurrentFieldBar: sleeping for 5 secs...")
        tick := time.NewTimer(time.Second * 5)
        return func() (interface{}, error) {
            <-tick.C
            fmt.Println("concurrentFieldBar: sleeping for 5 secs... done")
            return &bar, nil
        }, nil
    },
},

[aneeskA]

This is the conclusion I came to based on my experiments. :)

i guess 'concurrent-resolvers' means resolve logic runs concurrently with the graphql schedule, and the schedule will wait and handle all promising resolve functions. Which Not means resolve func is run concurrently by the schedule, you should setup concurrently tasks by yourself