Aggregation $group $sum fields among one document - mongodb

Question

Here is document example, year field contains year keys, that contains some metrics with included days as keys:
{
"_id" : NumberInt(1),
"year" : {
"2017" : {
"g1" : {
"1" : {
"total" : 2.0
},
"2" : {
"total" : 5.0
}
},
"g2" : {
"1" : {
"total" : 3.0
},
"2" : {
"total" : 6.0
}
}
}
}
I dont want getting document into memory to summarize total field for each key field g#.
How could i say to mongodb, summ total field for each key in year field.
Result that i want: g1 = 7.0, g2 = 9.0

Answer 1

You have to change your year part of structure to something like below.(Preferred)
"year" : [{ "k" : "2017", "v":[{ "k": "g1", "v":[{ "k" : "1","v" : {"total" : 2 }},{ "k" : "2","v" : {"total" : 5}}]}, { "k": "g2", "v":[{ "k" : "1","v" : {"total" : 3 }},{ "k" : "2","v" : {"total" : 6}}]}]}]
You can the below aggregation. This will work without knowing the keys ahead of time.
The query $unwinds couple of times to reach the g & total document followed by group on the g key and calculate total sum.
db.collection.aggregate([
{$match:{_id:1}},
{$unwind:"$year"},
{$unwind:"$year.v"},
{$unwind:"$year.v.v"},
{
$group:
{
_id:"$year.v.k",
sum: {$sum:"$year.v.v.v.total"}
}
}
])
This is the solution if you can't change your structure.
You can use 3.4.4 version and use $objectToArray to convert all the dynamic keys into labeled key and value pair.
Stage 1 & 2: Match on _id filter and convert the dynamic year keys into label value pair.
Stage 3 & 4: $unwind year array & $reduce the total value to calculate sum before changing the g1 and g2 dynamic keys to labeled key and value pair.
db.collection.aggregate([
{$match:{_id:1}},
{$addFields: {"year": {$objectToArray: "$year"}}},
{$unwind:"$year"},
{
$project:
{
g1:
{
$reduce: {
input: {$objectToArray: "$year.v.g1"},
initialValue: 0,
in: { $sum: [ "$$value", "$$this.v.total" ] }
}
},
g2:
{
$reduce: {
input: {$objectToArray: "$year.v.g2"},
initialValue: 0,
in: { $sum: [ "$$value", "$$this.v.total" ] }
}
}
}
}
])