[Discover] [Unified Field List] Field stats load exceptionally slow on large data sets

[davismcphee]

The new field stats popover seems to load exceptionally slow on large data sets. I imagine this is in large part because we're working with bigger samples rather than the 500 or fewer documents we used to, but it's quite disruptive and we should discuss if we're happy with this or can do anything to improve the UX:

One thought I had it would it be useful to fallback to the old field stats as a placeholder while the new field stats are loading? We could potentially do something like this:

• Display the loading indicator initially, and if the stats have not loaded within X seconds, display the old field stats.
• Show a loading indicator next to the old stats and a message along the lines of Calculated from a sample of up to 500 documents. More accurate stats are loading..
• Replace the old stats with the new stats once they have loaded.

[elasticmachine]

Pinging @elastic/kibana-data-discovery (Team:DataDiscovery)

[mattkime]

@davismcphee How do you define a big data set?

[davismcphee]

@mattkime I guess I don't have a hard definition, but in the GIF I originally posted I was working with the entirety of the nyc_taxis data set, which is pretty huge and probably an extreme example at 165m records. But taking a couple week chunk of that set with 5.5m records still results in significant enough loading times, approaching 5s for some, that a fallback might be helpful: