Mongodb relative frequency in grouping aggregation - mongodb

Question

I have data that looks like this
{"customer_id":1, "amount": 100, "item": "a"}
{"customer_id":1, "amount": 20, "item": "b"}
{"customer_id":2, "amount": 25, "item": "a"}
{"customer_id":3, "amount": 10, "item": "a"}
{"customer_id":4, "amount": 10, "item": "b"}
Using R I can get an overview of relative frequencies very easily by doing this
data %>%
group_by(customer_id,item) %>%
summarise(total=sum(amount)) %>%
mutate(per_customer_spend=total/sum(total))
Which returns;
customer_id item total per_customer_spend
<dbl> <chr> <dbl> <dbl>
1 1 a 100 0.833
2 1 b 20 0.167
3 2 a 25 1
4 3 a 10 1
5 4 b 10 1
I can't figure out how to do this in Mongo efficiently, the best solution I have so far involves multiple groups and pushing and unwinding.

Answer 1

If you don't want to change the data structure there's no way around grouping all the data as we need to determine the total amount spent of each user, though this would require just a single $group stage and a single $uwind stage, it would look somethine like this:
db.collection.aggregate([
{
$group: {
_id: "$customer_id",
total: {$sum: "$amount"},
rootHolder: {$push: "$$ROOT"}
}
},
{
$unwind: "$rootHolder"
},
{
$project: {
newRoot: {
$mergeObjects: [
"$rootHolder",
{total: "$total"}
]
}
}
},
{
$replaceRoot: {
newRoot: "$newRoot"
}
},
{
$project: {
customer_id: 1,
item: 1,
total: "$amount",
per_customer_spend: {$divide: ["$amount", "$total"]}
}
}
])
With that said, especially when scale increases this pipeline becomes very expensive, Now depending on how big the scale is and the amount of unique pairs of costumer_id x item i would advice the following:
considering Mongo doesn't like data duplication and assuming a user does not "buy" new items too often it might be worth to actually save it as a field in the current collection. (which requires updating all the users items on purchase), I know this sounds "weird" and costly but again depending on frequency of purchases it might actually be worth it.
Assuming you decide not to do the above I would instead create a new collection with customer_total and customer_id. Mind you this field will still require upkeeping although much cheaper.
With this collection you can either $lookup the total (which again can be expensive).