Type-graphql Playground

• Kick off
• Graphql example
• Regenerate Data
• Statistics
• Next Phase

Kick off

# install dependency
yarn install

# compose docker
yarn project up

# start graphql server
yarn start

Graphql example

• query http://localhost:12305/graphql

query GetGroup($getGroupFilter: GetGroupDto!) {
  GetGroup(filter: $getGroupFilter) {
    _id
    name
    userCount
  }
}

# query with datalodare loadMany() without aggregation help
query GetGroups($getGroupsFilter: GetGroupsDto!) {
  GetGroups(filter: $getGroupsFilter) {
    _id
    name
    userCount
    users {
      _id
      email
      firstName
      lastName
    }
  }
}

# query with dataloader load() method with optmized mongo aggregation query
query GetGroupsViaSingleMongoQuery(
  $getGroupsFilterViaSingleMongoQuery: GetGroupsDto!
) {
  GetGroups(filter: $getGroupsFilterViaSingleMongoQuery) {
    _id
    name
    userCount
    users2 {
      _id
      email
      firstName
      lastName
    }
  }
}

• variables: ids is stored in generated groupIds.json in project root directory

{
  "getGroupFilter": {
    "_id": "61eabfc67d720e6372867487"
  },
  "getGroupsFilter": {
    "_ids": [
      "61eabfc67d720e6372867487"
    ]
  },
  "getGroupsFilterViaSingleMongoQuery": {
    "_ids": [
      "61eabfc67d720e6372867487"
    ]
  }
}

Regenerate Data

# generate the seed scripts
yarn project generateSeedScripts

# seed data to mongodb
yarn project seed

# create indices
yarn project index

Statistics

• for query 2000 groups with users as sub-document in graphql
• dataloader has two main functionalities here: batch requests (deduplicate) and cache request (memory)

Data Loader method	Aggregation Query	Plain Query	Time Consumed	use case
load(groupId)	✅		600-1000ms	huge amount of fanned out promises
loadMany(userIds)		✅	30000-60000ms	small amount of fanned out promises

Take away

• Graphql Flexibility: Favor field resolver to assemble exterior fields over assembling in service layer.
• Delegate cumbersome assembling logic from field resolver to mongodb aggregate pipeline.

Next Phase

• How to do performant sorting in graphql layer since field resolver happens after data get fetched from database?