Question

I have two collections, as follows:

db.ships
document format: { mmsi: Long, …some other fields }
indexes: { {mmsi: 1}, {unique: true} }

db.navUpdates
document format: { mmsi: Long, time: ISODate, …some other fields }
indexes: { mmsi: 1 }, { time: 1 }

For each document within db.ships, I need to find the most recent document within db.navUpdates that matches by mmsi. I cannot use _id to find most recent as documents are not necessarily entered in chronological (as defined by timestamp time) order.

E.g.:

ship document:
{ mmsi: 12345 }

navUpdate documents:
{ mmsi: 12345, time: ISODate("2012-09-19T12:00:00.000Z") }
{ mmsi: 12345, time: ISODate("2012-09-18T12:00:00.000Z") }
{ mmsi: 54321, time: ISODate("2012-09-19T12:00:00.000Z") }

So for the ship with mmsi:12345, the most recent navUpdate is the first document in the list above, with time:ISODate("2012-09-19T12:00:00.000Z").

I tried the following mongo shell script, but it’s incredibly slow (multiple seconds for just 10 queries), and messages appearing on the server indicate I’m missing an index.

db.ships.find().limit(10).forEach(function(ship) {
    var n = db.navUpdates.find({mmsi:ship.mmsi}).count();
    if (n==0) { return; }
    var t = db.navUpdates.find({mmsi:ship.mmsi}).sort({time:-1}).limit(1)[0].time;
    print(t);
});

Why is this query so slow? I tried adding a {time: -1} index to navUpdate, thinking perhaps the sort({time: -1}) might be the culprit, but still no improvement.

Also, can this query be optimized? I have the count() call in there because there are some ship documents whose mmsis are not found within navUpdates.

Answer 1

A single query can only use one index, so you should add a compound index of { mmsi: 1, time: -1 } to navUpdates that can be used for both your find and sort needs.

Then use .explain() to determine if your indexes are being used in your queries.