Extended stats Agg

[spalger]

Added the extended stats agg, which includes the ability to choose which of the extended stats you want (count, min, max, avg, sum, sum_of_squares, variance, std_deviation) Closes #541

[rashidkpc]

@spenceralger is this ready for review?

[rashidkpc]

When I click to remove a metric it turns black, which doesn't really seem to do anything?

Also, it wasn't totally obvious how to get a removed item back. I now see that its a select2 input. Maybe we could add a plus button or something?

[spalger]

@rashidkpc fixed by replacing with checkboxes

[spalger]

@rashidkpc refined UX a bit more.

• displays error count above the apply button
• extended FormController, adding form.describeErrors() which returns something like 3 Errors
• extended NgModelController to force inputs (and in turn forms) to become dirty when the model moves from valid to invalid state.

[w33ble]

I'm guessing that, given this cryptic Java error, that this should only work on numeric fields.

ClassCastException[org.elasticsearch.index.fielddata.plain.PagedBytesIndexFieldData cannot be cast to org.elasticsearch.index.fielddata.IndexNumericFieldData]

[jwalton]

#1536 was closed by this, and #199 was closed by #1536, but I don't understand how this addresses #199.

Let's say I have a node.js process which are all logging out their process.memoryUsage() once a minute. So I have a bunch of log entries that have rss, heapTotal, and heapUsed fields. Can I plot all three of these to a single graph? This is what #199 was asking for.

Or, let's suppose I have a bunch of node.js processes which are all logging {pid, heapUsed} objects - can I plot one series for each PID? This would be amazing.

Since #199 was a pretty popular issue, if this doesn't address these points, #199 should probably be re-opened.

[spalger]

@jwalton you can definitely do this the latest master.

Case 1:
to plot rss, heapTotal, and headUsed over time you would probably want to start with a line or bar chart. Then create three metric aggregations calculating the average of each field you are interested in. Then you could create a single bucket aggregation, using a date_histogram on your timestamp field.

Here is what that would look like in the visualize editor:

Case 2:
First, setup a chart as described in Case 1. Then, add pid to the mix by adding a sub-aggregation under the date_histogram using the terms aggregation on the pid field.

Hope that helps!