Mongo group aggregation with priority of record - mongodb

Question

I am trying to perform a MongoDB 3.6 aggregation and I can't figure out the right way.
The problem is following. After performing several aggregation steps I end up with result set like this:
[
{ _id: { month: 1, type: 'estimate' }, value: 50 },
{ _id: { month: 2, type: 'estimate' }, value: 40 },
{ _id: { month: 3, type: 'estimate' }, value: 35 },
{ _id: { month: 3, type: 'exact' }, value: 33.532 },
{ _id: { month: 4, type: 'estimate' }, value: 10 },
{ _id: { month: 4, type: 'exact' }, value: 11.244 },
]
It contains values grouped by month. Value for every month can be 'estimated' or 'exact'. Now I would like to reduce this result to achieve this:
[
{ _id: { month: 1 }, value: 50 },
{ _id: { month: 2 }, value: 40 },
{ _id: { month: 3 }, value: 33.532 },
{ _id: { month: 4 }, value: 11.244 },
]
Basically I want to use the value of type 'exact' whenever it's possible and only fallback to 'estimate' value in months where the 'exact' is not known.
Any help or tip will be greatly appreciated. I would like to perform that aggregation in the DB not on server.

Answer 1

You can simply $sort by type and then take use $first in next $group stage which will give you exact if exists and estimate otherwise. Try:
db.col.aggregate([
{
$sort: { "_id.type": -1 }
},
{
$group:{
_id: "$_id.month",
value: { $first: "$value" }
}
},
{
$sort: { _id: 1 }
}
])
Prints:
{ "_id" : 1, "value" : 50 }
{ "_id" : 2, "value" : 40 }
{ "_id" : 3, "value" : 33.532 }
{ "_id" : 4, "value" : 11.244 }
So sorting by type is considered as prioritizing here since we know that lexically exact will be before estimate. You can also be more explicit and add extra field called weight (evaluated using $cond) operator and then sort by that weight:
db.col.aggregate([
{
$addFields: {
weight: { $cond: [ { $eq: [ "$_id.type", "exact" ] }, 2, 1 ] }
}
},
{
$sort: { "weight": -1 }
},
{
$group:{
_id: "$_id.month",
value: { $first: "$value" }
}
},
{
$sort: { _id: 1 }
}
])