Aggregating Data in MongoDB based on properties values - mongodb

Question

I'm using MongoDB 3.4 and I need to create a query filter to fetch data from my mongo collection (ProductionEventsCollection).
Some values are calculated "on the fly".
As these records are event based and future values depend on past values, if a value is changed in the past it affects the future sums.
I have a collection with the properties as written bellow:
[{
_id: "5bfc2a16b4f11f3760ed4b64",
piece_id: "12345",
finish_date: "2018-11-26T17:15:09.795Z",
total_produced: 500
},
{
_id: "5bfc2a16b4f11f3760ed4b65",
piece_id: "12345",
finish_date: "2018-11-27T17:15:09.795Z",
total_produced: 750
},
{
_id: "5bfc2a16b4f11f3760ed4b66",
piece_id: "12345",
finish_date: "2018-11-28T17:15:09.795Z",
total_produced: 250
}]
The idea is to get a collection like this:
[{
_id: "5bfc2a16b4f11f3760ed4b64",
piece_id: "12345",
finish_date: "2018-11-26T17:15:09.795Z",
previous_value: 0,
total_produced: 500,
new_value: 500
},
{
_id: "5bfc2a16b4f11f3760ed4b65",
piece_id: "12345",
finish_date: "2018-11-27T17:15:09.795Z",
previous_value: 500,
total_produced: 750,
new_value: 1250
},
{
_id: "5bfc2a16b4f11f3760ed4b66",
piece_id: "12345",
finish_date: "2018-11-28T17:15:09.795Z",
previous_value: 1250,
total_produced: 250,
new_value: 1500
}]
Based on the finish_date property, I should be able to calculate the previous_value sums until that date, and the new_value will be the previous calculated sum plus the total_produced.
Previous value
previous_value = SUM(past total_produced) until finish_date
New Value
new_value = previous_value + total_produced
Based on these collection values I need to return a json array, because I'll allow users to download a spreadsheet.

Answer 1

You can try below aggregation to get the result, the logic is to use $reduce to calculate the running total
aggregation pipeline
db.t32.aggregate([
{$group : {_id : "$piece_id", data : {$push : "$$ROOT"}}},
{$addFields : {data :
{$reduce : {
input : "$data",
initialValue : [{previous_value : 0, total_produced : 0, new_value : 0}],
in : {$concatArrays :
[ "$$value",[{$mergeObjects : ["$$this", { previous_value : {$arrayElemAt : ["$$value.new_value", -1]} , total_produced : "$$this.total_produced", new_value : {$sum : ["$$this.total_produced",{$arrayElemAt : ["$$value.new_value", -1]}]}}]}]]
}
}}
}},
{$addFields : {data : {$slice : ["$data", 1, {$size : "$data"}]}}}
]).pretty()
sample collection
> db.t32.find()
{ "_id" : "5bfc2a16b4f11f3760ed4b64", "piece_id" : "12345", "finish_date" : "2018-11-26T17:15:09.795Z", "total_produced" : 500 }
{ "_id" : "5bfc2a16b4f11f3760ed4b65", "piece_id" : "12345", "finish_date" : "2018-11-27T17:15:09.795Z", "total_produced" : 750 }
{ "_id" : "5bfc2a16b4f11f3760ed4b66", "piece_id" : "12345", "finish_date" : "2018-11-28T17:15:09.795Z", "total_produced" : 250 }
aggregation result
> db.t32.aggregate([
... {$group : {_id : "$piece_id", data : {$push : "$$ROOT"}}},
... {$addFields : {data :
... {$reduce : {
... input : "$data",
... initialValue : [{previous_value : 0, total_produced : 0, new_value : 0}],
... in : {$concatArrays :
... [ "$$value",[{$mergeObjects : ["$$this", { previous_value : {$arrayElemAt : ["$$value.new_value", -1]} , total_produced : "$$this.total_produced", new_value : {$sum : ["$$this.total_produced",{$arrayElemAt : ["$$value.new_value", -1]}]}}]}]]
... }
... }}
... }},
... {$addFields : {data : {$slice : ["$data", 1,1000]}}}
... ]).pretty()
{
"_id" : "12345",
"data" : [
{
"_id" : "5bfc2a16b4f11f3760ed4b64",
"piece_id" : "12345",
"finish_date" : "2018-11-26T17:15:09.795Z",
"total_produced" : 500,
"previous_value" : 0,
"new_value" : 500
},
{
"_id" : "5bfc2a16b4f11f3760ed4b65",
"piece_id" : "12345",
"finish_date" : "2018-11-27T17:15:09.795Z",
"total_produced" : 750,
"previous_value" : 500,
"new_value" : 1250
},
{
"_id" : "5bfc2a16b4f11f3760ed4b66",
"piece_id" : "12345",
"finish_date" : "2018-11-28T17:15:09.795Z",
"total_produced" : 250,
"previous_value" : 1250,
"new_value" : 1500
}
]
}
>

Answer 2

I am not sure if such an aggregation is even possible in mongodb.
Sounds like a typical business logic for the sever side and therefore I would suggest fetching the relevant data to the server and perform all the computation there. That way things are much less complex and without any performance loss.