MongoDB use $last with $cond - mongodb

Question

Is there any way to get the last non-zero value in an aggregation. Is that possible?
Scenario:
I have an events collection, in which I store all the events from users. I want to fetch the list of users with last purchased items cost is $1.99 and logged in at least once last week.
My events collection will have records like
{_id:ObjectId("58af54d5ab7df73d71822708"),uid:1,event:"login"}
{_id:ObjectId("58db7189296fdedde1c04bc1"),uid:2,event:"login"}
{_id:ObjectId("5888419bfa4b69dc4af7c76c"),uid:2,event:"purchase",amount:3}
{_id:ObjectId("5888419bfa4b69dc4af7d45c"),uid:1,event:"purchase",amount:1.9}
{_id:ObjectId("5888819bfa4b69dc4af7c76c"),uid:1,event:"custom",type:3,value:2}
What am trying to do:
db.events.aggregate([{
{
$group: {
_id: uid,
last_login: {
$max: {
$cond: [{
$eq: ['$event', 'login']
}, '$_id', 0]
}
},
last_amount: {
$last: {
$cond: [{
$eq: ['$event', 'login']
}, '$_id', 0]
}
}
}
}
}, {
$match: {
last_purchase: {
$gte: ObjectId("58af54d50000000000000000")
},
last_amount: 1.9
}
}])
which obviously will fail because last will have 0 as the last item.
The output am expecting is
{_id:1,last_login:_id:ObjectId("58af54d5ab7df73d71822708"),last_amount:1.9}
The query is system generated. Please help.

Answer 1

To answer your question, you cannot modify the behaviour of $last using $cond. i.e. 'not the last entry because of a criteria'.
There are few alternatives that you could try:
Alter your document schema for the access pattern, especially if this is a frequently used query in your application. You should utilise the schema as a leverage to optimise your application queries.
Use one of the MongoDB supported drivers to process the expected outcome.
Depending on the use case, you could execute 2 separate aggregation queries; first query all the users uid that logged in at least once last week, and second to query only those users that have the last purchased item value of $1.99.

Answer 2

I found a workaround.Instead of $last, I used $push in the $group and added $slice with $setDifference in $project to remove null values. The query will now look something like
db.events.aggregate([{
{
$group: {
_id: uid,
last_login: {
$max: {
$cond: [{
$eq: ['$event', 'login']
}, '$_id', 0]
}
},
last_amount: {
$push: {
$cond: [{
$eq: ['$event', 'login']
}, '$_id', null]
}
}
}
}
},
{$project:{last_login:{$slice:[{$setDifference,'$last_login',[null]},-1,1]}}}
, {
$match: {
last_purchase: {
$gte: ObjectId("58af54d50000000000000000")
},
last_amount: 1.9
}
}])