Hi

JK, Graphs make "some" things easier

Let's take a look at a few examples.

Btw this whole talk is totally not a rip off from MongoDB's official blog

MongoDB Graph Database

• Let's add some sample data to our db that looks like this

[
  {
    "name": "Bob",
    "friends": ["Alice", "Kate"]
  },
  {
    "name": "Alice",
    "friends": ["Bob", "John"]
  },
  {
    "name": "Kate",
    "friends": ["Bob"]
  },
  {
    "name": "John",
    "friends": ["Alice"]
  }
]

• It looks like an adjacency list, doesn't it? Totally a coincidence /s.

• One of the common features in a social network is to show you friend suggestions.
• It would also help the user to know why a friend is being suggested - maybe they have common friends or common interests.
• Let's try to find friend suggestions for Kate. She only has 1 friend - Bob. Who else has Bob in their friends list?

db.people.find({ friends: "Bob", name: { $ne: "Kate" }})

{
  _id: ObjectId("63cb73f1bb0776fa5f0fd497"),
  name: 'Alice',
  friends: [
    'Bob',
    'John'
  ]
}

• That was simple enough. Alice is also friends with John so maybe Kate knows John too
• How can we query the data in a way John also shows up? And moreover we might want to know if the person is a 2nd connection (single level of mutual friend) or 3rd (2 levels).

Fortunately for us, we don't need to install a graph DB for simple queries like the one above.

MongoDB has an aggreation operator called $graphLookup.

db.people.aggregate([
  { $match: { name: "Kate", } },
  {
    $graphLookup: {
      from: "people",
      startWith: "$friends",
      connectFromField: "friends",
      connectToField: "name",
      as: "network",
      depthField: "connectionLevel",
      maxDepth: 2,
      restrictSearchWithMatch: { name: { $ne: "Kate" } }
    },
  },
  {
    $project: { name: 1, network: 1 }
  }
])

• Who are the 2nd and 3rd connections of Kate?
• Pretty self-explanatory right? I think we're done here.

Alright, let's break it down

• We start with Kate. $match.name: "Kate"
• GraphLookup operator starts a recursive search from all the documents from the previous output. In our case it's only one document.
  • The loop starts with the value in startsWith and looks for that value in the colleciton in from field.
  • A graph path is constructed with value in fromField and toField.
  • Traversal happens until the specified depth is reached or we run out of edges.
• TL;DR Mongo performs a DFS.