Filling in with documents with default values after find/aggregate - mongodb

Question

I have a collection:
{ "name" : "A", "value" : 1, "date" : ISODate("2014-01-01T00:00:00.000Z") }
{ "name" : "B", "value" : 7, "date" : ISODate("2014-01-01T00:00:00.000Z") }
{ "name" : "A", "value" : 3, "date" : ISODate("2014-01-02T00:00:00.000Z") }
{ "name" : "B", "value" : 8, "date" : ISODate("2014-01-02T00:00:00.000Z") }
{ "name" : "B", "value" : 8, "date" : ISODate("2014-01-03T00:00:00.000Z") }
{ "name" : "A", "value" : 5, "date" : ISODate("2014-01-04T00:00:00.000Z") }
{ "name" : "A", "value" : 4, "date" : ISODate("2014-01-05T00:00:00.000Z") }
The document for A on 3rd Jan 2014 is not available. When I do a find/aggregate on A, I would like the document to appear in my result set with a default value (or better, value to be same as previous date). For example:
{ "name" : "A", "value" : 1, "date" : ISODate("2014-01-01T00:00:00.000Z") }
{ "name" : "A", "value" : 3, "date" : ISODate("2014-01-02T00:00:00.000Z") }
{ "name" : "A", "value" : 3 (or default value -1), "date" : ISODate("2014-01-03T00:00:00.000Z") }
{ "name" : "A", "value" : 5, "date" : ISODate("2014-01-04T00:00:00.000Z") }
{ "name" : "A", "value" : 4, "date" : ISODate("2014-01-05T00:00:00.000Z") }
How can this be done?

Answer 1

One thing you need in order to be able to do this in aggregation framework is an array of dates that you want your report to cover. For example, for input that you show, you might have an array:
days = [ ISODate("2014-01-01T00:00:00Z"), ISODate("2014-01-02T00:00:00Z"),
ISODate("2014-01-03T00:00:00Z"), ISODate("2014-01-04T00:00:00Z"),
ISODate("2014-01-05T00:00:00Z"), ISODate("2014-01-06T00:00:00Z") ];
to indicate that you want every one of these six days represented.
Here is the aggregation that you would run:
db.coll.aggregate( [
{$group : {_id:{name:"$name",date:"$date"},value:{$sum:"$value"}}},
{$group : {_id:"$_id.name", days:{$addToSet:"$_id.date"},docs:{$push:"$$ROOT"}}},
{$project : {missingDays:{$setDifference:[days,"$days"]},docs:1}},
{$unwind : "$missingDays"},
{$unwind : "$docs"},
{$group : {
_id:"$_id",
days:{$addToSet:{date:"$docs._id.date",value:"$docs.value"}},
missingDays:{$addToSet:{date:"$missingDays",value:{$literal:0}}}
} },
{$project : {_id:0, name:"$_id", date:{$setUnion:["$days","$missingDays"]}}},
{$unwind : "$date"},
{$sort : {date:1,name:1}}
] )
On your sample input with days defined as above it outputs:
{ "name" : "A", "date" : { "date" : ISODate("2014-01-01T00:00:00Z"), "value" : 1 } }
{ "name" : "A", "date" : { "date" : ISODate("2014-01-02T00:00:00Z"), "value" : 3 } }
{ "name" : "A", "date" : { "date" : ISODate("2014-01-03T00:00:00Z"), "value" : 0 } }
{ "name" : "A", "date" : { "date" : ISODate("2014-01-04T00:00:00Z"), "value" : 5 } }
{ "name" : "A", "date" : { "date" : ISODate("2014-01-05T00:00:00Z"), "value" : 4 } }
{ "name" : "A", "date" : { "date" : ISODate("2014-01-06T00:00:00Z"), "value" : 0 } }
{ "name" : "B", "date" : { "date" : ISODate("2014-01-01T00:00:00Z"), "value" : 7 } }
{ "name" : "B", "date" : { "date" : ISODate("2014-01-02T00:00:00Z"), "value" : 8 } }
{ "name" : "B", "date" : { "date" : ISODate("2014-01-03T00:00:00Z"), "value" : 8 } }
{ "name" : "B", "date" : { "date" : ISODate("2014-01-04T00:00:00Z"), "value" : 0 } }
{ "name" : "B", "date" : { "date" : ISODate("2014-01-05T00:00:00Z"), "value" : 0 } }
{ "name" : "B", "date" : { "date" : ISODate("2014-01-06T00:00:00Z"), "value" : 0 } }
The first group stage may not be necessary in your case - it's there in case there are multiple documents for the same name and date, in that case you want to add the values for them. The second $group and $project stage figure out the difference between the days present for each name and the array of days you want covered, creating missingDays which will be getting the value 0 in the next $group stage. That group stage creates for each name an array of dates that have data and array of missing dates that don't. It structures them the say way so that the following $project stage can create a union of them using the $setUnion operator. After that all that's left is to $unwind the array of dates and sort it whichever way you want.