mongodb/meteor: how to I get the value of one field corresponding to the $max value of another field? - mongodb

Question

I have a collection of messsages with the following fields: _id, senderId, receiverId, dateSubmittedMs, message, and for a given user I want to return the latest message to him from all other users. So, for example, if there are users Alex, Barb, Chuck, Dora, I would like to return the most recent message between Alex and each of Barb, Chuck and Dora. What is the best way to do this? Can I do it in one step using aggregation?
The aggregation examples in the official online documentation (http://docs.mongodb.org/manual/reference/aggregation/min/) show how to find the lowest age over groups within a collection, but what I need is something analogous to finding the name of the youngest person over groups of people.
Here is my current approach:
Step 1: Find the highest value for dateSubmitted over all messages sent and received by Alex, grouping over the other users:
var M = Messages.aggregate(
{$match:
{$or: [{senderId: 'Alex'}, {receiverId: 'Alex'}]}
},
{$group: {_id: "$receiverId", lastestSubmitted: {$max: "$submitted"} }}).fetch();
Step 2: Create an array of these highest values of dateSubmitted:
var MIds = _.pluck(M,'lastestSubmitted');
Step 3: Find these messages, by senderId, receiverId, and latestSubmitted:
return Messages.find(
{submitted: {$in: MIds}, $or: [{senderId: 'Alex'}, {receiverId: 'Alex'}]},
{$sort: {submitted: 1}}
});
There are two problems with this:
Can it be done in one step instead of three? Perhaps through a mapReduce or Aggregate command?
Instead of grouping only over the receiverId: 'Alex', is there a way to group over something like: $or [{receiverId: 'Alex', senderId: 'Barb'}, {senderId: 'Alex', receiverId: 'Barb'}]? (but for EACH of the other users) This would allow me to get the latest message in a conversation between any two participants that Alex conversed with. So for example:
Any suggestions?

Answer 1

The only thing you have to change is the group._id in the grougping phase. While you can use a document for this propose not just a field, you can apply the sort in the same pipeline. However it is not change if you use only the receiverId for grouping.
var M = Messages.aggregate(
{$match:
{$or: [{senderId: 'Alex'}, {receiverId: 'Alex'}]}
},
{$group:
{_id: {
receiverId: "$receiverId",
senderId: "$senderId"},
lastestSubmitted: {$max: "$submitted"} }
},
{$sort: {submitted: -1}
},
{$limit: 1}
).fetch();
This example above only adds the possiblity to check for the second most recently pinged connection or the third one, if you only likely to get the most recent message you do not even need to group. Just run this:
var M = Messages.aggregate(
{$match:
{$or: [{senderId: 'Alex'}, {receiverId: 'Alex'}]}
},
{$sort: {submitted: -1}
},
{$limit: 1}
).fetch();
THE ANSWER ABOVE THIS IS FOR GETTING THE MOST RECENT MESSAGE BETWEEN A SINGLE USER AND OTHERS, TO GET THE MOST RECENT MESSAGE BETWEEN ALL PAIRS INVOLVING A USER READ UNDER>>>
Based on the comments i misinterpreted a bit the question but the part above is useful anyway. The correct resolution for your problem is under. The difficulty in is to get a key for the pair to group on and identify the switched pair ( bob -> tom == tom -> bob) in this case. U can use condition and ordering to identify the swaps. (It is certainly a much more difficult question) The code look like this:
var M = Messages.aggregate(
{$match:
{$or: [{senderId: 'Alex'}, {receiverId: 'Alex'}]}
},
{$project:
{'part1':{$cond:[{$gt:['$senderId','$receiverId']},'$senderId','$receiverId']},
'part2':{$cond:[{$gt:['$senderId','$receiverId']},'$receiverId','$senderId']},
'message':1,
'submitted':1
}
},
{$sort: {submitted: -1}},
{$group:
{_id: {
part1: "$part1",
part2: "$part2"},
lastestSubmitted: {$first: "$submitted"},
message: {$first: "$message"} }
}
).fetch();
If you are not familiar with some of the operators used above, like $cond or $first, check out this.