弹性搜索：如何在单个索引中搜索两个文档？

Question

I've two documents configured in a single index named inventory.

1. profile (this have users and mobile information)
2. p2p-relation

Inventory index mapping properties are as follows:

{
    "mappings": {
        "properties": {
            "type": {"type": "keyword"},
            "id": {"type": "keyword"},
            "sourceId": {"type": "keyword"},
            "targetId": {"type": "keyword"},
            "firstName": {"type": "text"},
            "lastName": {"type": "text"},
            "createdBy": {"type": "text"},
            "p_type": {"type": "text"},
            "model": {"type": "text"},
            "OS": {"type": "keyword"}
            }
        }
}

profile has the following information

users:

{   "type":"profile",   "id" : "user1", "p_type" : "user",  "firstName" : "Ashok", "lastName" : "S" }
{   "type":"profile",   "id" : "user2", "p_type" : "user",  "firstName" : "Arun", "lastName" : "V" }

mobiles:

{   "type":"profile",   "id" : "mobile1", "p_type" : "mobile",  "model" : "samsung", "OS" : "Android" }
{   "type":"profile",   "id" : "mobile2", "p_type" : "mobile",  "model" : "iPhone", "OS" : "iOS" }

p2p-relation has the which user used which mobile information:

{   "type":"p2p-relation",  "id" : "user1-owns-mobile1", "sourceId" : "user1",  "targetId" : "mobile1", "createdBy" : "admin" }
{   "type":"p2p-relation",  "id" : "user1-owns-mobile2", "sourceId" : "user1",  "targetId" : "mobile2", "createdBy" : "admin" }

In our business case, we need to retrieve the list of android/iOS mobiles owned by an user which we get as input from the customer.

That is, if the user1 requests for /mymobiles?query=os==Android , it should translate this into ES and expecting

{ "type":"profile", "id": "mobile1", "p_type": "mobile", "model": "samsung", "OS": "Android" }

as the result and if the user2 requests for the same, it should return empty.

I've tried with query and bool. But it search only within a single document. How to achieve this in elastic search?

Answer 1

Above model falls under many to many relationship data model. An user can have many mobile phones assigned to him and a mobile device can be assigned to many users over the period of time. So changed the data model to storing it as nested objects.

{
  "mappings": {
    "properties": {
        "type": {"type": "keyword"},
        "id": {"type": "keyword"},
        "firstName": {"type": "text"},
        "lastName": {"type": "text"},
        "createdBy": {"type": "text"},
        "model": {"type": "text"},
        "OS": {"type": "keyword"},
        "mobiles" : {
            "type" : "nested",
            "properties" : {
                "id": {"type": "keyword"},
                "model": {"type": "text"},
                "OS": {"type": "keyword"},
                "createdBy": {"type": "text"}
            }
        },
        "users" : {
            "type" : "nested",
            "properties" : {
                "id": {"type": "keyword"},
                "firstName": {"type": "keyword"},
                "lastName": {"type": "keyword"},
                "createdBy": {"type": "text"}                               
            }
        }
}}}

But the pain is, whenever a mobile phone is assigned to an user, I have the extra effort of updating two rows.

Data stored as follows

{   "type":"profile",   "id" : "user1",  "firstName" : "Ashok", "lastName" : "S", "mobiles" : [ {"id" : "uuid1", "model": "samsung", "OS" : "Android"}] }

{   "type":"profile",   {"id" : "uuid1", "model": "samsung", "OS" : "Android"}, "users" : [ "id" : "user1",  "firstName" : "Ashok", "lastName" : "S"] }

And to answer the /mymobiles?query=os==Android, I've used the following query.

{ 
"query": {
    "bool": {
        "must": [ 
            {
              { "term": {"OS": "android"} },
              "nested": {
                "path": "users",
                "query": {
                  "bool": {
                    "must": [ {"term": {"users.id": "user1"} }]
                  }
                }
              }
            }
        ]   
    }
}}

I've referred the following link to do this: https://medium.com/@mena.meseha/briefly-describe-how-to-store-complex-relational-data-in-elasticsearch-f30277317b1