Lens area chart appears empty

[AlonaNadler]

Version 7.6.1
I dropped a few fields in Lens and then changed to the area chart using the chart switcher. The chart preview appears empty. In a closer look, it seems that instead of area Lens plots dots but you can hardly see it

[elasticmachine]

Pinging @elastic/kibana-app (Team:KibanaApp)

[wylieconlon]

@AlonaNadler This looks like what we expect to happen when the interval chosen by the auto-interval is smaller than the rate at which data is collected: for example, if data is collected every hour, but the interval is 30 minutes, then this would be expected.

There are two options here:

• Increase the time interval so that every bucket has data
• Prioritize working on the fitting functions work in Lens: Allow users to configure fitting/filling function for missing/null buckets #17717

But I don't think this is a bug, even though it's unexpected.

[wylieconlon]

Closing as I think the technical work is already tracked on the fitting functions issue.

[AlonaNadler]

The reason I think it is a bug :

• you can hardly see there is something in the preview, for a few seconds I thought it was empty.
• this is how the same chart look in visualize, this behavior make much more sense
  . Even if you take 4 hours intervals and the data ingested once a day and it is an area chart. It should show one bucket a day with the accurate number
  

[wylieconlon]

If this is a bug, then we still have some choices to make about how to define the solution. I think the best solution to this is to let users define their fitting functions, which you can test on the charts demos.

For example, the fitting function that is closest to the behavior you see in the vislib area charts is possible, but I don't think this is the right choice for this data set:

Because your dataset is "spiky", it might be better to let the user choose something like a linear fitting function, which will smooth out the spikiness.

The reason I closed this issue is that we are already tracking the need for user-selected fitting functions, which I think is the right solution to the problem.

[timroes]

I agree with Wylie here. The behavior in visualize is actually the broken one, since we make assumptions about how we will be able to fill null values for the user, which does not make sense, since we can't, without knowing the semantics of the data and whether filling with 0 actually makes sense for that specific dataset.

So I'll close this with respect to letting the user specify their fitting function which would then allow them to chose the right approach for those missing data: #17717