Mapping an associative array in Elasticsearch

Question

I'm relatively new to elasticsearch and I would like to map one of my field to an associative array structure.

What I currently Have :

One product of type product in the index products:

{
    "name":"Nexus 10",
    "category":"Tablets",
    "interests": {
        "1":30,
        "3":70
    }
}

As you can see, the field interests is an associative array. The key is a User ID and the value is the number of times the user talked about this product.

Dynamic mapping provided me with :

{
   "products":{
      "mappings":{
         "product":{
            "properties":{
               "category":{
                  "type":"string"
               },
               "interests":{
                  "properties":{
                     "1":{
                        "type":"long"
                     },
                     "3":{
                        "type":"long"
                     },
                  }
               },
               "name":{
                  "type":"string"
               }
            }
         }
      }
   }
}

What I want :

Since I'm going to have lots of entries in the interests field, I don't want the dynamic mapping to map each ID.

The idea is to be able to look for a specific user ID in the interests field so that I can boost the score of the result depending on the number of times this particular user talked about this product.

I would like to be able to map this kind of structure.

Do you have any idea how to achieve that?

Thanks,

Alex Brasetvik · Accepted Answer · 2014-02-20 16:16:05Z

9

You will need to change your object definitions.

Having keys based on what's actually values is problematic, as it will lead to an ever-growing mapping and thus cluster state. It'll also lead to quirky searches.

This is a common enough question/issue that I dedicated a section to it in my article on Troubleshooting Elasticsearch searches, for Beginners.

What you probably want to do is to either use nested documents, or parent-child-relations. If you update the counts very often, you'd want to use parent-child, as nested documents will cause reindexing of the entire thing. Nested doc will be faster search time, though, and you'll very likely want to throttle the updates somehow anyway.

If you structure the documents like this, you can easily use nested mappings, and you won't end up with one field per user.

name: Nexus 10
category: Tablets
interests: 
    - user_id: 1
      mentions: 30
    - user_id: 3
      mentions: 70

answered Feb 20, 2014 at 16:16

Alex Brasetvik

11.8k2 gold badges40 silver badges38 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Brice Over a year ago

Thank you very much for your answer. Your article is very interesting, I'll keep that :D. Just a quick question though, it means that instead of having a script like this: _score * (interests contains userId ? numberOfMentions : 1) I'll have to foreach interests : look for the userId passed as a param?

Alex Brasetvik Over a year ago

If you use the nested-query, you can use e.g. total as score_mode, to sum the scores of all matching documents. Here's an example: found.no/play/gist/4a68e0e1e724e0f9e693

Collectives™ on Stack Overflow

Mapping an associative array in Elasticsearch

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related