Elasticsearch - aggregation of unique counts - elasticsearch

Question

I have an Elasticsearch database of books:
{
"id": 1,
"name": "Animal Farm"
},
{
"id": 2,
"name": "Brave New World"
},
{
"id": 3,
"name": "Nineteen Eighty-Four"
},
{
"id": 4,
"name": "Animal Farm"
},
{
"id": 5,
"name": "We"
}
As you can see, the books with id of 1 and 4 have the conflict book name "Animal Farm". However, they are different books. One is by George Orwell, and another one is literally about farm animals.
I want to know how often do the book names conflict. For the example above, the expected results are:
{
"conflicts": [
{
"num_of_books": 2,
"count": "1"
},
{
"num_of_books": 1,
"count": "3"
}
]
}
The entry with num_of_books of 2 is the conflict of "Animal Farm", and it happened once (therefore the count is 1). The other 3 books has all different names, so they appears in the entry with num_of_books of 1 and count of 3. I don't need the names of the books. Only the counts matter.
I know SQL has "subquery" to do this:
SELECT num_of_books, COUNT(*) AS _count
FROM (
SELECT COUNT(*) AS num_of_books
FROM books
GROUP BY name
)
GROUP BY num_of_books;
I read the articles of Nested Aggregation and Sub-Aggregations, but failed to see the possibility to achieve my goal.
Any comment will help, thanks!

Answer 1

Running aggregations on aggregations is not yet possible in ES, as far as I know. I know of a few outstanding issues on allowing to apply additional logic on the results of bucket aggregations, but they are still being discussed and debated.
In your case, you can get away with the inner SQL query by using a terms aggregation in order to get the name of all conflicting books by using min_doc_count: 2.
{
"size": 0,
"aggs": {
"books": {
"terms": {
"field": "name",
"min_doc_count": 2
}
}
}
}
Then you can parse the buckets on the client side and re-bucket them into new num_of_books buckets depending on their count. For instance, using the head plugin you can add the following code in the Transform section
var num_of_books = {};
root.aggregations.books.buckets.forEach(function(b) {
num_of_books[b.doc_count] = (num_of_books[b.doc_count] || 0) + 1;
});
return num_of_books;
num_of_books would then contain something like this:
{
"2": 1,
"1": 3
}