Question

I am creating a mapping like this

"institution" : {
  "properties" : {        
    "InstitutionCode" : {
      "type" : "string",
      "store" : "yes"
    },
    "InstitutionID" : {
      "type" : "integer",
      "store" : "yes"
    },
    "Name" : {
      "type" : "string",
      "store" : "yes"
    }
  }
}

However, when I perform actual indexing operations for institutions, I am adding an Alias property (0 or more aliases per institution)

"institution" : {
  "properties" : {   
    "Aliases" : {
      "dynamic" : "true",
      "properties" : {
        "InstitutionAlias" : {
          "type" : "string"
        },
        "InstitutionAliasTypeID" : {
          "type" : "long"
        }
      }
    },     
    "InstitutionCode" : {
      "type" : "string",
      "store" : "yes"
    },
    "InstitutionID" : {
      "type" : "integer",
      "store" : "yes"
    },
    "Name" : {
      "type" : "string",
      "store" : "yes"
    }
  }
}

This is actually a simplified example, as I am actually adding more fields than just Aliases during the actual indexing of records.

How important is it to to fully define a mapping during mapping-creation?

Am I going to suffer any penalties by having the mapping automatically adjusted during indexing operations due to the indexing of institution records with additional properties? I expect institutions to gain additional properties over time and I wonder if I need to maintain the mapping-creation code in addition to the institution-indexing code.

Answer 1

I believe the overhead of dynamic mapping is fairly negligible…using them won’t hurt indexing speed. However, you can run into some unexpected situations where ElasticSearch auto-detects a field type incorrectly.

A common example is detecting an integer because the first example of a field is a number (“25”), when in reality the rest of the data for that field is a string. Or seeing an integer when the rest of the data is actually a float. Etc etc.

If your data is well standardized that isn’t much of a problem.

Alternatively, you can use dynamic templates to apply mappings to new fields based on a regex pattern.