Aggregating and comparing scores of teams using mongodb - mongodb

Question

I'm just starting out with mongodb and have been reading through the documentation on aggregation but still struggling to relate equivalent knowledge of sql statements to the methods used in mongo.
I have this data:
{
"_id" : ObjectId("53ac7bce4eaf6de4d5601c19"),
"uid" : ObjectId("53ac7bb84eaf6de4d5601c15"),
"mid" : ObjectId("53ab27504eaf6de4d5601be4"),
"score" : 1
},{
"_id" : ObjectId("53ac7bce4eaf6de4d5601c1a"),
"uid" : ObjectId("53ac7bb84eaf6de4d5601c16"),
"mid" : ObjectId("53ab27504eaf6de4d5601be4"),
"score" : 5
}
...
And I'm trying to get to this result:
{
"uid" : ObjectId("53ac7bb84eaf6de4d5601c15"),
"uid_2" : ObjectId("53ac7bb84eaf6de4d5601c16"),
"mid" : ObjectId("53ab27504eaf6de4d5601be4"),
"score" : 1,
"score_2" : 5,
"difference" : 4
}
...
Where I am comparing every uid against every other uid around a single mid and calculating the difference in their scores (can't be a negative difference, only positive).
Most of the examples I'm running into don't quite fit my requirements and hoping some mongo guru can help me out. Thanks!

Answer 1

As stated, I think your data modelling is a little off here as you need something to "pair" the "matches" as it were. I have a "simplified" case here:
{
"_id" : ObjectId("53ae9da2e24682cac4215e0c"),
"match" : ObjectId("53ae9d78e24682cac4215e0b"),
"score" : 1
}
{
"_id" : ObjectId("53ae9da5e24682cac4215e0d"),
"match" : ObjectId("53ae9d78e24682cac4215e0b"),
"score" : 5
}
{
"_id" : ObjectId("53aea6cde24682cac4215e15"),
"match" : ObjectId("53aea6c1e24682cac4215e14"),
"score" : 2
}
{
"_id" : ObjectId("53aea6e4e24682cac4215e16"),
"match" : ObjectId("53aea6c1e24682cac4215e14"),
"score" : 1
}
{
"_id" : ObjectId("53aea6eae24682cac4215e18"),
"match" : ObjectId("53aea6e6e24682cac4215e17"),
"score" : 2
}
{
"_id" : ObjectId("53aea6ece24682cac4215e19"),
"match" : ObjectId("53aea6e6e24682cac4215e17"),
"score" : 2
}
What that basically represents is the scores for "six" teams in "three" distinct matches.
Given that, my take on getting to results would be this:
db.matches.aggregate([
// Group on matches and find the "min" and "max" score
{ "$group": {
"_id": "$match",
"teams": {
"$push": {
"_id": "$_id",
"score": "$score"
}
},
"minScore": { "$min": "$score" },
"maxScore": { "$max": "$score" }
}},
// Unwind the "teams" array created
{ "$unwind": "$teams" },
// Compare scores for "win", "loss" or "draw"
{ "$group": {
"_id": "$_id",
"win": {
"$min": { "$cond": [
{ "$and": [
{ "$eq": [ "$teams.score", "$maxScore" ] },
{ "$gt": [ "$teams.score", "$minScore" ] }
]},
"$teams",
false
]}
},
"loss": {
"$min": { "$cond": [
{ "$and": [
{ "$eq": [ "$teams.score", "$minScore" ] },
{ "$lt": [ "$teams.score", "$maxScore" ] }
]},
"$teams",
false
]}
},
"draw": {
"$push": { "$cond": [
{ "$eq": [ "$minScore", "$maxScore" ] },
"$teams",
false
]}
},
"difference": {
"$max": { "$subtract": [ "$maxScore", "$minScore" ] }
}
}},
// Just fix up those "draw" results with a [false,false] array
{ "$project": {
"win": 1,
"loss": 1,
"draw": { "$cond": [
{ "$gt": [
{ "$size": { "$setDifference": [ "$draw", [false] ] } },
0
]},
"$draw",
false
]},
"difference": 1
}}
])
And this gives you a quite nice result:
{
"_id" : ObjectId("53ae9d78e24682cac4215e0b"),
"win" : {
"_id" : ObjectId("53ae9da5e24682cac4215e0d"),
"score" : 5
},
"loss" : {
"_id" : ObjectId("53ae9da2e24682cac4215e0c"),
"score" : 1
},
"draw" : false,
"difference" : 4
}
{
"_id" : ObjectId("53aea6c1e24682cac4215e14"),
"win" : {
"_id" : ObjectId("53aea6cde24682cac4215e15"),
"score" : 2
},
"loss" : {
"_id" : ObjectId("53aea6e4e24682cac4215e16"),
"score" : 1
},
"draw" : false,
"difference" : 1
}
{
"_id" : ObjectId("53aea6e6e24682cac4215e17"),
"win" : false,
"loss" : false,
"draw" : [
{
"_id" : ObjectId("53aea6eae24682cac4215e18"),
"score" : 2
},
{
"_id" : ObjectId("53aea6ece24682cac4215e19"),
"score" : 2
}
],
"difference" : 0
}
That is essentially the results per "match" and determines the "difference" between winner and looser while identifying which team "won" or "lost". The final stage there uses some operators only introduced in MongoDB 2.6, but that really is not necessary if you do not have that version available. Or you could actually still do the same thing if you wanted to by using $unwind and some other processing.