Issue with mapping in elasticsearch

[Hanni (Hanane)]

Hi, I'm trying to index a field in elasticsearch with my index template. Except that my field can either be of type text (ex: No) or it can be of type float ( ex: 8.1). When I set the type to float in my mapping, I get errors parsing text values. I tried to define the type as "keyword". but when I want to create a data table visualization in kibana and I want to choose the range metric, I can't use my field because the range metric only works on numeric values. How can I proceed?

[dadoonet (David Pilato)]

You need to split the text content and the numbers into 2 fields.

It's better if you can do that before sending the data to Elasticsearch.
Otherwise, you can create an ingest pipeline which converts the foo field to a float into a foo_float field. If you add ignore_failure: true, the field won't be created if not a number.

PUT _ingest/pipeline/to-float
{
  "processors" : [
    {
      "convert" : {
        "field" : "foo",
        "target_field" : "foo_float",
        "type": "float",
        "ignore_failure": true
      }
    }
  ]
}