Classification analysis fails when dependent variable is of type text.

[przemekwitek]

The error message is:

{
        "type" : "illegal_argument_exception",
        "reason" : "Text fields are not optimised for operations that require per-document field data like aggregations and sorting, so these operations are disabled by default. Please use a keyword field instead. Alternatively, set fielddata=true on [glass_type] in order to load field data by uninverting the inverted index. Note that this can use significant memory."
      }

[przemekwitek]

This reproduces easily in integration test.

[przemekwitek]

Ok, I was able to track down the root cause of this issue. As part of starting the analysis, we first perform cardinality aggregation on dependent_variable field to use it later for limit verification (see line 125 in ExtractedFieldsDetectorFactory.java) and only then we verify supported dependent_variable types in ExtractedFieldsDetector

This order is problematic as we simply cannot perform cardinality aggregation on a text field (it yields the error I provided in the first post).
So we would need to reverse this order and first check for supported types and then check cardinality.

[javanna]

heya @przemekwitek may I ask you to add labels to this issue please? Thanks!

[elasticmachine]

Pinging @elastic/ml-core (:ml)

[przemekwitek]

heya @przemekwitek may I ask you to add labels to this issue please? Thanks!

Done.

[przemekwitek]

PR #53874 makes the error message more understandable for the user.