A million documents

Let's insert a million documents into a collections:

db.todos.insertMany(
  [...Array(1000000)].map((_, index) => ({
    text: "todo" + i,
    done: Math.random() > 0.5,
  }))
);

We can make mongo explain how the engine is going to find a todo by its text with:

db.todos.find({ text: "todo786231" }).explain("executionStats");

{
  ...
  "executionStats": {
    ...
    "executionTimeMillis": 604,
    "executionStages": {
      ...
      "stage": "COLLSCAN",
    },
    "totalDocsExamined": 1625292,
  }
}

The COLLSCAN way of searching is basically a linear search, so that is very slow in comparison with what could be. To fix this we have to create an index.

Working with indexes

With indexes (or indices), MongoDB will find things much faster. MongoDB uses B-Trees to store indexes. Indexes can be unique or not, depending if they allow repeated keys or not.

MongoDB already uses a unique index for the _id field, which is always present.

To create an index in mongosh you have to call the createIndex method in the collection and indicate what fields are involved:

db.todos.createIndex({ text: 1 });

The number associated with a field indicates if the ordering is ascending or descending. To create a compound index, just mention the involved fields:

db.users.createIndex({ userid: 1, score: -1 });

The methods of a collection involved in indexes are:

Hidden indexes are invisible to the query planner and therefore are not used in queries. This helps evaluate the impact of adding an index to a particular query, without actually dropping the index.