Hey 
I created an index with nested objects, kind of like this:
user:
properties:
name:
type: text
institutions:
type: nested
properties:
name:
type: text
description:
type: text
members:
type: nested
properties:
name:
type: text
description:
type: text
orders:
type: nested
properties:
[...]
How can i use the More Like This Query, to find similar users based on name, institutions, orders and so on. This doesn't work:
GET /users/user/_search
{
"query": {
"more_like_this" : {
"like" : [
{
"_index" : "users",
"_type" : "user",
"_id" : "1"
}],
"max_query_terms": 25,
"min_term_freq": 1,
"min_doc_freq": 1,
"minimum_should_match": 1
}
}
}
I've tested it (only institutions + members, without orders) with Nested Query also:
POST /users/user/_search
{
"query": {
"nested": {
"path": "institutions",
"query": {
"bool": {
"should": [
{
"more_like_this": {
"like" : [{
"_index" : "users",
"_type" : "user",
"_id" : "1"
}],
"max_query_terms": 25,
"min_term_freq": 1,
"min_doc_freq": 1,
"minimum_should_match": 1
}
}, {
"nested": {
"path": "institutions.members",
"query": {
"bool": {
"should": {
"more_like_this": {
"like" : [{
"_index" : "users",
"_type" : "user",
"_id" : "1"
}],
"max_query_terms": 25,
"min_term_freq": 1,
"min_doc_freq": 1,
"minimum_should_match": 1
}
}
}
}
}
}
]}
}
}
}
}
}
If there are the same terms in name + description in institutions from different users (A + B) it matchs. But if i use only the same terms in name + description in institution from user A and in one member of some institution of user B, it not matchs. It seems like the MLT Query compare the users only in the same stage (institution XOR member). Another disadvantage is that this kind of query becomes very complexe with more than 60 nested types.
What's the best practice to deal with this issue?