To avoid costly errors, validate features when they're added about elasticsearch-learning-to-rank HOT 3 CLOSED

I think two different levels of validation we could do would be:

• fully validation will require an index that has all the mappings required to run the queries. Would that work if the user has to provide the "test index" the query will be run into for testing?
• Simple json validation, we could validate that the query will pass various query builders but not validate everything that is related to mapping and field types/options. Problem is that I need to have he parameters replaced.

from elasticsearch-learning-to-rank.

Can we include an optional (perhaps mandatory) block for testing the feature that includes

• Index for testing
• Sample arguments for the template?

For example, when you create a feature, you might append a test block like so:

 {
      "name": "user_rating",
      "type": "feature",
      "feature": {
         "name": "user_rating",
         "params": ["keywords"],
         "test": {
             "keywords": "rambo",
             "index": "tmdb" 
         }
         "template_language": "mustache",
         "template": {
            "query": {
               "match": {
                  "title": "{{keywords}}"
               }
            }
         }
      }
   }

from elasticsearch-learning-to-rank.

Closed by #89

from elasticsearch-learning-to-rank.