mongodb aggregate uniqueness and count at same time with higher level averages - mongodb

Question

Consider this dataset. For each name, we wish to find the average of x and the distinct set and count of game. For Steve, this is avg(x)=19, game A is 2, and game B is 1. For Bob, this is avg(x) = 58, game B is 4:
{"name":"Steve", "game": "A", x:7},
{"name":"Steve", "game": "A", x:21},
{"name":"Steve", "game": "B", x:31},
{"name":"Bob", "game": "B", x:41},
{"name":"Bob", "game": "B", x:51},
{"name":"Bob", "game": "B", x:71},
{"name":"Bob", "game": "B", x:79},
{"name":"Jill", "game": "A", x:61},
{"name":"Jill", "game": "B", x:71},
{"name":"Jill", "game": "C", x:81},
{"name":"Jill", "game": "D", x:91}
EDIT: Answer is below but leaving this incomplete solution as a stepping stone.
I am really close with this. Note we cannot use $addToSet because it is "lossy". So instead, we group by player and game to get the full list, then in a second group, capture list size:
db.foo2.aggregate([
{$group: {_id:{n:"$name",g:"$game"}, z:{$push: "$x"} }}
,{$group: {_id:"$_id.n",
avgx: {$avg: "$z"},
games: {$push: {name: "$_id.g", num: {$size:"$z"}}}
}}
]);
which yields:
{
"_id" : "Steve",
"avgx" : null,
"games" : [ {"name":"A", "num":2 },
{"name":"B", "num":1 }
]
}
{
"_id" : "Bob",
"avgx" : null,
"games" : [ {"name":"B", "num":4 } ]
}
but I just cannot seem to get the avgx working properly. If I needed the average within the game type that would be easy but I need it across the player. $avg in the $group context does not work with array inputs.

Answer 1

Try this:
db.collection.aggregate([
{
$group: {
_id: "$name",
avg: {
$avg: "$x"
},
gamesUnFiltered: {
$push: {
name: "$game",
num: "$x"
}
}
}
},
{
$addFields: {
games: {
$reduce: {
input: "$gamesUnFiltered",
initialValue: [],
in: {
$cond: [
{
$not: [
{
$in: [
"$$this.name",
"$$value.name"
]
}
]
},
{
$concatArrays: [
[
"$$this"
],
"$$value"
]
},
"$$value"
]
}
}
}
}
},
{
$project: {
gamesUnFiltered: 0
}
}
])
Output:
[
{
"_id": "Bob",
"avg": 60.5,
"games": [
{
"name": "B",
"num": 41
}
]
},
{
"_id": "Steve",
"avg": 19.666666666666668,
"games": [
{
"name": "B",
"num": 31
},
{
"name": "A",
"num": 7
}
]
},
{
"_id": "Jill",
"avg": 76,
"games": [
{
"name": "D",
"num": 91
},
{
"name": "C",
"num": 81
},
{
"name": "B",
"num": 71
},
{
"name": "A",
"num": 61
}
]
}
]

Answer 2

Got it! You need an extra $unwind and use $first to "carry" the a field from stage to stage. Threw in total_games for extra info. In general, the "group-unwind-first" pattern is a way to aggregate one or more things then "reset" to unaggregated state to perform additional operations with the aggregate values traveling along with each doc.
db.foo2.aggregate([
{$group: {_id:"$name", a:{$avg:"$x"}, g:{$push: "$game"} }}
,{$unwind: "$g"}
,{$group: {_id:{name:"$_id",game:"$g"}, a:{$first:"$a"}, n:{$sum:1}}}
,{$group: {_id:"$_id.name",
a:{$first:"$a"},
total_games: {$sum:"$n"},
games: {$push: {name:"$_id.game",n:"$n"}}
}}
]);