About performance

METEOR@1.6.1.1 cultofcoders:grapher@1.3.6

i’ve tested my db and got performance issue.

GroupMembers.addLinks({
  group: {
    type: 'one',
    collection: Groups,
    field: 'groupId',
  },
  user: {
    type: 'one',
    collection: Users,
    field: 'userId',
  },
})

const groupmembers = GroupMembers.createQuery({
  user: {
    username: 1,
  },
  group: {
    name: 1,
  },
}).fetch()

I20180625-10:04:58.430(8)? … 6583 more items ] I20180625-10:04:58.430(8)? undefined: 1285.079ms

this is another test by fetch all content and assemble manually

const users = Users.find().fetch()
const groups = Groups.find().fetch()

const usersKeyById = _.keyBy(users, '_id')
const groupsKeyById = _.keyBy(groups, '_id')

const groupmembers = GroupMembers.find().map(gm => ({
  user: usersKeyById[gm.userId],
  groups: groupsKeyById[gm.groupId],
  ...gm,
}))

I20180625-10:05:58.910(8)? … 6583 more items ] I20180625-10:05:58.910(8)? undefined: 285.375ms

[theodorDiaconu]

@crapthings this has been fixed in 1.3.7. You will do it faster if you do it manually just because you don’t apply filters and the database returns your responses much faster, so it’s not necessarily a Grapher disadvantage. Anyway, I tested for the following:

1000 users 1000 groups Each user has 2 groups, so 2 group memberships per user ~2000

Results using your way: 58ms Results using Grapher: 70ms

This is acceptable if we take into account that it retrieves only the Users and Groups that you need. (In our case all bc the use-case is built that way, but you get it!)

[crapthings]

it looks like query was transform into mongo aggregate

https://github.com/cult-of-coders/grapher/issues/188#issuecomment-359643600