Visualize: Recommend filter aggs as first aggs

[spalger]

In the aggregation builder we should start recommending that users increase the priority for filter aggregations. Since they are static (they don't change when evaluated in the context of a bucket) there really isn't any reason to run them lower in the tree.

One of the side effects of running them lower in the tree is the creation of buckets in places you wouldn't expect. For instance, pr #6574 aims to add zero-filtering to visualizations, when the actual issue is that the filter is being evaluated in the context of every bucket, and producing a bucket regardless if there are documents or not.

[androidkencai]

@spalger I am doing research to see if Kibana can generate filter aggregation in Visualize - aggregation builder.
currently, kibana has

 Schemas([{
  group: 'metrics',
....
  aggFilter: ['count', 'avg', 'sum', 'min', 'max', 'cardinality', 'std_dev']
}])

additionally to have filter in the dropdown

is it the same thing as this issue? if so, it's labelled as P2, when do we expect to have this support?
Thanks!
KC

[spalger]

Actually this is about putting up a warning when filter aggs are not the top bucket agg

the difference between this:

and this:

The reason for this more apparent when you zoom in closer and start finding x-axis buckets that don't have any documents. When the filters are the first aggregation the lines in the chart skip over the empty buckets:

But when the filters are last calculated second they produce zero-values inside buckets that don't have any documents.

[androidkencai]

@spalger thanks for clarifying this, I will open a new issue for supporting the filter aggregation in the metrics group

[timroes]

We addressed this issue recently in some discussions and we currently don't have plans to add this warning in the existing visual editor anymore. The new visual editor (#17076) will use a different approach of ordering visualizations, which hopefully won't cause that kind of UX issues with users.