Calculated/computed Fields resolver calling order

I’m working through providing a graphql source to an energy monitoring application. Total load for a premises is essentially the sum of all of the measured loads on each circuit.

The following is a trivial piece of the work where there are two measured load fields and one total_load field

class Heating(graphene.ObjectType):
    total_load = graphene.Float()
    water_heater = graphene.Field(lambda: SensorReading,
                                  resolver=get_measurement,
                                  args={'id': graphene.String('Water Heater')})
    space_heater = graphene.Field(lambda: SensorReading,
                                      resolver=get_measurement,
                                      args={'id': graphene.String('Space Heater')})

I have tried to create a simple resolver to compute the total_load from water_heater and space_heater values but this total_load resolver always gets called first so both water_heater and space_heater are None, meaning I can’t do the calculation.

Both water_heater and space_heater use a DataLoader, part of the get_measurement resolver. This is important as the time series database (InfluxDB) has a tendancy to return quite large amounts of data the I need to resolve from a cache and which need bulk loading.

I have thought about using a promise or middleware to get the total_load but wanted to get a steer as to the right way to resolve this type of timing related issue and more generally how to approach calculated/computed values using both GraphQL and Graphene as much seaching hasn’t given me much luck so far.

[jkimbo]

@j-robin-hunter have you tried moving the data fetching to the level above the Heating type so that you can do it at the same time? For example:

class Query(graphene.ObjectType):
	heating = graphene.Field(Heating)

	def resolve_heating(_, info):
		water_heater = get_water_heater_load()
		space_heater = get_space_heater_load()
		total = water_heater + space_heater
		return {
			'water_heater': water_heater,
			'space_heater': space_heater,
			'total_load': total,
		}

[jkimbo]

@rarenatoe in that case I would recommend abstracting the day calculation into a seperate function that you either cache or compute it at the level above the object type that needs the data.