[ML] DataFrame UI cannot create value count agg with defaults

[hendrikmuhs]

Kibana version: 7.2-SNAPSHOT

When I want to group by terms and I also like to know how many datapoints belong to 1 bucket I can use the value_count agg to create a count. However with the current UI and the chosen defaults the UI doesn't let me create it:

Workaround: If I change the group_by settings and then create the value_count agg (afterwards I can even change the value_count agg and revert the change for group_by):

[elasticmachine]

Pinging @elastic/ml-ui

[sophiec20]

This occurs because of the default naming suggestions for the data frame fields. In this instance, the field user is being used as both a group-by and an agg.

The default naming for the data frame fields is as follows:

• group-by -> user
• aggs -> user.value_count

The workaround is to change the default field names. Because a user-facing error message is also displayed, this workaround was considered acceptable for the beta.

Consideration should be given as to how we can make this user experience better. Also take into account the users intentions, when selecting value_count is usually just to achieve the equivalent of count -- for the UI, we should perhaps make selecting count easier and perhaps move the field selection for value_count to be an advanced (JSON editing) option.

[walterra]

Another suggestion how we could solve this:

• Keep the default naming as is with dots, e.g. responsetime.avg, airline.value_count.
• Only if we identify a conflict switch to a naming pattern with underscores, e.g. responsetime_avg, airline_value_count and display a toast message like Used airline_value_count instead of airline.value_count because of a naming conflict with airline.

This would allow us both to keep the dot based naming and not block the user and require them to rename groups or aggs.