User RootResolver Factory instead of creating it once

[arthurvaverko]

from the docs

First we need somewhere to track our state, lets put it in graph/resolver.go. The graph/resolver.go file is where we declare our app's dependencies, like our database. It gets initialized once in server.go when we create the graph.

type Resolver struct{
	todos []*model.Todo
}

while it is great to have a place for dependencies on the root resolver the main drawback is
It gets initialized once in server.go when we create the graph.

using golang its easy to init something once using sync.Once
while initializing something per-request is much harder and in many cases is implemented using context.Context injection of values.

Generally the community convention is against using context for DI one ref can be found here

My suggestion is to add a ResolverRootFactory to the generated config sturct

type Config struct {
	Resolvers  ResolverRoot
	Directives DirectiveRoot
	Complexity ComplexityRoot
        
        // add this
        ResolverRootFactory func(context.Context)ResolverRoot
}

then the generated code that is creating the resolver from the executable schema will create it using the Request Context.
this way we can inject loggers per request with some contextual information or DataLoaders or any other per-request dependencies and pass the explicitly to all other resolvers including Query, Mutation or any Type resolver.

I was wondering if thats something that only i find as required ?
how do you deal with per-request dependencies ? do you use context? or is there any other approach?

[arthurvaverko]

bump ...