MondoDB Aggregate - Absolute count for each day - mongodb

Question

I have the following (already aggregated)collection
{ "_id" : { "day" : "2015-02-01" }, "total" : 2 }
{ "_id" : { "day" : "2015-02-02" }, "total" : 3 }
{ "_id" : { "day" : "2015-02-03" }, "total" : 10 }
{ "_id" : { "day" : "2015-02-04" }, "total" : 10 }
{ "_id" : { "day" : "2015-02-05" }, "total" : 5 }
What i need is calculating an absolute value for each day, summing the previous days. So expected result would be in the case above
{ "_id" : { "day" : "2015-02-01" }, "absolutetotalforday" : 2 }
{ "_id" : { "day" : "2015-02-02" }, "absolutetotalforday" : 5 }
{ "_id" : { "day" : "2015-02-03" }, "absolutetotalforday" : 15 }
{ "_id" : { "day" : "2015-02-04" }, "absolutetotalforday" : 25 }
{ "_id" : { "day" : "2015-02-05" }, "absolutetotalforday" : 30 }
Currently no clue how to achieve this with 1 query. Of course i could do a sum for each day I'm interested in, but this might be a long time range
Any help appreciated

Answer 1

Because aggregation framework has no mechanism of knowing the value of a previous document, or the previous "grouped" value of a document, your best bet would be to use Map-Reduce in this case.
Map-Reduce will give you the "running total" for the current total at end of day you require although this won't be in the desired key, absolutetotalforday but in a key called value since the reduced values are always invalue key.
The following mapReduce() operation will give you the desired result, assuming the results from the previous aggregation operation were output to a separate collection named agg_results:
db.agg_results.mapReduce(
function() { emit( this._id, this.total ); },
function(key, values) { return Array.sum(values); },
{
"scope": { "total": 0 },
"finalize": function(key, value) {
total += value;
return total;
},
"out": { "inline": 1 }
}
);
Sample Output
{
"results" : [
{
"_id" : {
"day" : "2015-02-01"
},
"value" : 2
},
{
"_id" : {
"day" : "2015-02-02"
},
"value" : 5
},
{
"_id" : {
"day" : "2015-02-03"
},
"value" : 15
},
{
"_id" : {
"day" : "2015-02-04"
},
"value" : 25
},
{
"_id" : {
"day" : "2015-02-05"
},
"value" : 30
}
],
"timeMillis" : 0,
"counts" : {
"input" : 5,
"emit" : 5,
"reduce" : 0,
"output" : 5
},
"ok" : 1
}
Sorting the results will not work with inline results and with dates of String type. Instead, try converting the date strings to a JavaScript date object, write the results to a collection and then run a sort on that collection:
db.agg_results.mapReduce(
function() { emit( new Date(this._id.day), this.total ); },
function(key, values) { return Array.sum(values); },
{
"scope": { "total": 0 },
"finalize": function(key, value) {
total += value;
return total;
},
"out": "tmpResults"
}
);
Sample Output (with sort)
> db.tmpResults.find().sort({_id: 1})
{ "_id" : ISODate("2015-02-01T00:00:00Z"), "value" : 2 }
{ "_id" : ISODate("2015-02-02T00:00:00Z"), "value" : 5 }
{ "_id" : ISODate("2015-02-03T00:00:00Z"), "value" : 15 }
{ "_id" : ISODate("2015-02-04T00:00:00Z"), "value" : 25 }
{ "_id" : ISODate("2015-02-05T00:00:00Z"), "value" : 30 }
>