[Visualize] Filter aggregation not working

[spalger]

Kibana version: master
Description of the problem including expected versus actual behavior:

Choosing the filter aggregation in visualize doesn't present any options to enter a filter, which allows the agg to be created without any params, and then sends a query to elasticsearch which triggers a error.

Steps to reproduce:

1. start a new vis
2. add a filter aggregation
3. click the play button

[ppisljar]

Filter aggregation should be disabled in all visualizations, as it is only used with tile map and is enabled with checkbox. @nreese do you want to look into this ?

[nreese]

I think @thomasneirynck had already started a PR but if not, I can work on it

[spalger]

I had agreed to take over @thomasneirynck's PR and implement the alternative/plugin-friendly approach but got side-tracked. If you want to take over @nreese I'd be happy to show you what I have.

[nreese]

@spalger Sure thing. What direction where you heading?

[thomasneirynck]

thx @spalger .

Fwiw, I think we can do a stop-gap here and exclude the Filters manually from the relevant visualizations for 6.0. Not sure about the effort involved in introducing the new API, but probably not something we should itnroduce before GA.

[spalger]

That's a good point @thomasneirynck, a short term solution is probably ideal for now.

@nreese this is what I came up with after talking with you and @thomasneirynck: master...spalger:implement/agg_groups. It is a lot more change than I anticipated, the gist of it is:

• ui/app_types/tags defines a set of tags that describe categories that aggs could fit into
• tags aim mostly to describe the type of result produced by an agg, which is generally enough to know if it's compatible with a vis/visSchema
• tags like METRIC_PARENT_PIPELINE describe metrics that are only useful when a visualization is bucketing
• other tags like METRIC_ONE_RESULT_BASIC describe a more ambiguous group that was discovered as the subset of metric aggs that could be used for ordering a terms agg... not sure if that's a feature or a bug
• the changes include some investigative work in the testbed plugin that tries to map out how the groups resolve, with an effort to ensure that they produce the same groupings as before
• https://gist.github.com/anonymous/aa9a3af41b8489792df29fa1b45d33aa has the changes I applied to a master fork to see the resolved agg types per vis schema in master, for comparison, but I realized kind of late that it doesn't include everything, like https://github.com/elastic/kibana/blob/master/src/ui/public/agg_types/buckets/terms.js#L23 for example

[thomasneirynck]

@spalger thx for the overview. In the meantime, I will put up a PR for this problem.