python graphql-tools imitation

graphql_tools.py

import graphql

# build_executable schema
#
# accepts schema_definition (string) and resolvers (object) in style of graphql-tools
# returns a schema ready for execution

def build_executable_schema(schema_definition, resolvers):
    ast = graphql.parse(schema_definition)
    schema = graphql.build_ast_schema(ast)
    
    for typeName in resolvers:
        fieldType = schema.get_type(typeName)
        
        for fieldName in resolvers[typeName]:            
            if fieldType is graphql.GraphQLScalarType:
                fieldType.fields[fieldName].resolver = resolvers[typeName][fieldName]
                continue
            
            field = fieldType.fields[fieldName]
            field.resolver = resolvers[typeName][fieldName]
            
        if not fieldType.fields: continue
    
        for remaining in fieldType.fields:
            if not fieldType.fields[remaining].resolver:
                fieldType.fields[remaining].resolver = \
                    lambda value, info, _r=remaining, **args: value[_r]
                
                
    return schema

usage.py

import graphql
import json

from graphql_tools import build_executable_schema

source_schema = """
    schema {
      query: RootQuery
    }

    type RootQuery {
      officers: [Officer]
    }
    
    type Officer {
      title: String
      first: String
      last: String
      uniform: Uniform
    }

    type Uniform {
      pips: Float
    }
"""

resolvers = {
  'RootQuery': {
    # returning direct values, but would normally load fetch db with some info.context
    'officers': lambda value, info, **args: [
        dict(first='william', last='riker'), 
        dict(first='geordi', last='laforge'),
    ]  
  },
  'Officer': {
      # only declaring the field here which is computed
      'title': lambda value, info, **args: 'Officer: %(first)s %(last)s' % value,
      # and the connection
      'uniform': lambda value, info, **args: value['first']=='geordi' and {'i':2.5} or {'i':3},
  },
  'Uniform': {
      'pips': lambda value, info, **args: value['i'],
  }
}


my_schema = build_executable_schema(source_schema, resolvers)

executed = graphql.graphql(my_schema, """
    query Example {
       officers {
         first
         last
         title
         uniform { pips }
       }
    }
""")

print(json.dumps(executed.data, indent=4))

usage_output.json

{
    "officers": [
        {
            "first": "william",
            "last": "riker",
            "title": "Officer: william riker",
            "uniform": {
                "pips": 3.0
            }
        },
        {
            "first": "geordi",
            "last": "laforge",
            "title": "Officer: geordi laforge",
            "uniform": {
                "pips": 2.5
            }
        }
    ]
}

And how to use with graphene?

How could I use it with graphene?

I've iterated on this idea a bit within a small project recently, and can hopefully find an opportunity from there to write it as a more complete package than the sketch above. I'll try to prioritize that in coming weeks and see what I can publish; will notify here when something comes to fruition.

(But as for usage with graphene, I'm not so sure -- this is more of an alternative to graphene's style of schema creation than an extension, as graphene's authors have likewise said, eg. here: graphql-python/graphene#704 )

Great Job!
I opened the issue graphql-python/graphene#704 .

I think it's a better idea to design schema by DSL format. I don't know much about why we have to get a extra graphene schema layer. Even though the graphene schema layer is necessary, we still must have a DSL => graphene schema parser.

I'm quite confused that why there are two types of definition. Graphene schema definition graphene.types.schema.Schema and graphql-core schema definition graphql.type.schema.GraphQLSchema.

Since graphene.types.schema.Schema is extended from graphql.type.schema.GraphQLSchema, how could I get graphene.types.schema.Schema from graphq.parse(DSL file)?