Elasticsearch - Summing filtered array based on terms aggregation - elasticsearch

Question

I have the following data structure in my documents in elasticsearch which represents a purchase:
{
...
"lineItems": [
{ "id": "1", "quantity": 2 },
{ "id": "2", "quantity": 1 },
]
...
}
I'm trying to work out the most popular product id in a provided date range. Using a terms aggregation I can work out the number of appearances of a product id in baskets, but I'm having trouble summing the quantities to work out how many of an item was purchased.
My current search looks like this:
{
"query": ...,
"size":0,
"aggs":{
"basketAppearances":{
"terms":{
"field":"lineItems.id.keyword"
},
"aggs":{
"timesPurchased":{
"sum":{
"field":"lineItems.quantity"
}
},
"order":{
"bucket_sort":{
"sort":[
{
"timesPurchased":"desc"
}
]
}
}
}
}
}
}
The problem with the above is that it obviously takes the full lineItems array and sums all the values, so basketAppearances is correct and timesPurchased is incorrect. i.e. I get the following result:
{
...
"aggregations":{
"basketAppearances":{
"doc_count_error_upper_bound":0,
"sum_other_doc_count":0,
"buckets":[
{
"key":"1",
"doc_count":1,
"timesPurchased":{
"value": 3
}
},
{
"key":"2",
"doc_count":1,
"timesPurchased":{
"value": 3
}
}
]
}
}
...
}
I need to sum only the rows in the array with the same ID as the terms bucket it resides. i.e. filter the array based upon the term of the terms aggregation
I appreciate the best answer here is probably to change my data format to have a different document type of "line item" and add a line item per array entry, but the data structure matches my data elsewhere (and it makes sense elsewhere) and I'd ideally like to keep it the same.