Mapping

relevant docs

Mapping defines the structure of documents, ie their fields and their data types.

here

We can either:

1. Define custom mappings, OR
2. Let ES generate mappings for us

Data types

Available data types are here.

We have typical data types like text, float, integer, boolean etc. But there are specific data types in ES that power specific use cases that are of interest:

Object

Objects may be nested to form a hierarchy. Instead of defining the type, ES instead defines a properties for the object:

Arrays of objects

ES flattens arrays of objects, which can cause some issues.

E.g

{
  "group" : "fans",
  "user" : [
    {
      "first" : "John",
      "last" :  "Smith"
    },
    {
      "first" : "Alice",
      "last" :  "White"
    }
  ]
}

Is flattened to

{
  "group" :        "fans",
  "user.first" : [ "alice", "john" ],
  "user.last" :  [ "smith", "white" ]
}

And we lose relationships between fields.

To solve this we can define the object type as Nested. Read more here.

Nested objects are hidden in the background

Keyword

reference

keyword is used for exact value matching. This is used for filtering, sorting and aggregating docs.

We can't use keyword fields for full text search. For keyword fields, a different analyzer is used, called the "keyword" analyzer. This is a noop analyzer, meaning it just re-outputs the string as a single token. This is because it's intended for exact matching.

You can still configure the analyzer, however, for example to lowercase keyword fields.

If you want a field to be of both type keyword and text, you can achieve this using the fields mapping parameter.

Type coercion

With dynamic mapping, ES will do type coercion. For example the string "7.4" will be converted to a float.