ElasticSearch & Tire: Using Mapping and to_indexed_json
elasticsearch, ruby, ruby-on-rails, tire
Solution
While the `mapping` and `to_indexed_json` methods are related, they serve two different purposes, in fact.
The purpose of the `mapping` method is to define mapping for the document properties within an index. You may want to define certain property as "not_analyzed", so it is not broken into tokens, or set a specific analyzer for the property, or (as you mention) indexing time boost factor. You may also define multifield property, custom formats for date types, etc.
This mapping is then used eg. when Tire automatically creates an index for your model.
The purpose of the `to_indexed_json` method is to define a JSON serialization for your documents/models.
The default `to_indexed_json` method does use your mapping definition, to use only properties defined in the mapping — on a basis that if you care enough to define the mapping, by default Tire indexes only properties with defined mapping.
Now, when you want a tight grip on how your model is in fact serialized into JSON for elasticsearch, you just define your own `to_indexed_json` methods (as the README instructs).
This custom `MyModel#to_indexed_method` usually does not care about mapping definition, and builds the JSON serialization from scratch (by leveraging ActiveRecord's `to_json`, using a JSON builder such as jbuilder, or just building a plain old Hash and calling `Hash#to_json`).
So, to answer the last part of your question, using both `mapping` and `to_indexed_json` will absolutely not create any conflicts, and is in fact required to use advanced features in elasticsearch.
To sum up:
- You use the `mapping` method to define the mapping for your models for the search engine
- You use a custom `to_indexed_json` method to define how the search engine sees your documents/models.
Problem
While reading the Tire doc, I was under the impression that you should use either `mapping` or `to_indexed_json` methods, since (my understanding was..) the `mapping` is used to feed the `to_indexed_json`. The problem is, that I found some tutorials where both are used. WHY? Basically, my app works right now with the `to_indexed_json` but I can't figure out how to set the boost value of some of the attributes (hence the reason I started looking at mapping) and I was wondering if using both would create some conflicts.