Tooltips don't display for most datapoints on graph and timeseries visualization

[cameronwaterman]

What happened:

Tooltips don't display for most datapoints on graph and timeseries visualization. I've also seen a related issue where the plot doesn't display at all. See gifs below for both issues.

What you expected to happen:

Tooltip would show for every datapoints and plot would show.

How to reproduce it (as minimally and precisely as possible):

Apply timeseries data to either of the above-mentioned visualizations and attempt to hover datapoints.

Anything else we need to know?:

Environment:

• Grafana version: 8.4.5 (Fork)
• Data source type & version: Custom datasource
• OS Grafana is installed on: Alpine 3.15
• User OS & Browser: Windows 10 / Chrome
• Grafana plugins: Above-mentioned datasource
• Others: N/A

[joshhunt]

I'm unable to reproduce on https://play.grafana.org/d/igIv8Ww7k/47795-time-series-tooltips

Do you have a sample data CSV or dashboard that replicates the issue?

[grafanabot]

Thanks for creating this issue! We think it's missing some basic information.

Follow the issue template and add additional information that will help us replicate the problem.
For data visualization issues:

• Query results from the inspect drawer (data tab & query inspector)
• Panel settings can be extracted in the panel inspect drawer JSON tab

For dashboard related issues:

• Dashboard JSON can be found in the dashboard settings JSON model view

For authentication, provisioning and alerting issues, Grafana server logs are useful.

Happy graphing!

[grafanabot]

Thanks for creating this issue! We think it's missing some basic information.

Follow the issue template and add additional information that will help us replicate the problem.
For data visualization issues:

• Query results from the inspect drawer (data tab & query inspector)
• Panel settings can be extracted in the panel inspect drawer JSON tab

For dashboard related issues:

• Dashboard JSON can be found in the dashboard settings JSON model view

For authentication, provisioning and alerting issues, Grafana server logs are useful.

Happy graphing!

[cameronwaterman]

Here is some additional info as requested. Inside the zip file you'll find the dashboard JSON and the data CSV.
Grafana_Issue_47795.zip

[mure]

I was able to reproduce this on play.grafana using the TestData DB and CSV content. I can't save the dashboard, but here's the CSV I used.
Panel Title-data-2022-04-19 11_38_13.csv
I had to add a transform to convert the time column to be a correctly typed time field.

[leeoniya]

the data must be ordered time/ASC, but the data in the CSV is ordered time/DESC.

if you add a Sort by -> time transform, it works as expected, i think. here's the CSV content i used:

time,value
2022-04-19T11:23:07.504Z,89.5
2022-04-19T11:22:07.504Z,13.3
2022-04-19T11:21:07.504Z,42.5
2022-04-19T11:20:07.504Z,8.78
2022-04-19T11:19:07.504Z,7.85
2022-04-19T11:18:07.504Z,57.5
2022-04-19T11:17:07.504Z,9.78
2022-04-19T11:16:07.504Z,91.3
2022-04-19T11:15:07.504Z,84.8
2022-04-19T11:14:07.504Z,16.8
2022-04-19T11:13:07.504Z,1.65
2022-04-19T11:12:07.504Z,2.80
2022-04-19T11:11:07.504Z,25.2
2022-04-19T11:10:07.504Z,70.6
2022-04-19T11:09:07.504Z,11.6
2022-04-19T11:08:07.504Z,61.2
2022-04-19T11:07:07.504Z,38.2
2022-04-19T11:06:07.504Z,90.6
2022-04-19T11:05:07.504Z,92.8
2022-04-19T11:04:07.504Z,9.63
2022-04-19T11:03:07.504Z,59.0
2022-04-19T11:02:07.504Z,25.8
2022-04-19T11:01:07.504Z,46.3
2022-04-19T11:00:07.504Z,29.3
2022-04-19T10:59:07.504Z,88.4
2022-04-19T10:58:07.504Z,82.3
2022-04-19T10:57:07.504Z,51.1
2022-04-19T10:56:07.504Z,76.9
2022-04-19T10:55:07.504Z,15.5
2022-04-19T10:54:07.504Z,73.3
2022-04-19T10:53:07.504Z,43.4
2022-04-19T10:52:07.504Z,34.6
2022-04-19T10:51:07.504Z,60.3
2022-04-19T10:50:07.504Z,83.1
2022-04-19T10:49:07.504Z,67.8
2022-04-19T10:48:07.504Z,1.90
2022-04-19T10:47:07.504Z,26.7
2022-04-19T10:46:07.504Z,64.5
2022-04-19T10:45:07.504Z,77.7
2022-04-19T10:44:07.504Z,77.1
2022-04-19T10:43:07.504Z,22.6
2022-04-19T10:42:07.504Z,27.6
2022-04-19T10:41:07.504Z,55.7
2022-04-19T10:40:07.504Z,15.5
2022-04-19T10:39:07.504Z,92.8
2022-04-19T10:38:07.504Z,82.2
2022-04-19T10:37:07.504Z,50.4
2022-04-19T10:36:07.504Z,72.4

[mure]

Yep, the "Prepare time series" transform also fixes it.
I think this should still be an issue, though, because the behavior is extremely confusing. The graph should either handle "incorrectly" ordered data or throw some sort of error.

[leeoniya]

we do have some handling for this, but need to look into why it's not working here.

grafana/packages/grafana-data/src/transformations/transformers/joinDataFrames.ts

Lines 112 to 118 in 785145c

	if (options.enforceSort) {
	if (joinIndex >= 0) {
	if (!isLikelyAscendingVector(frameCopy.fields[joinIndex].values)) {
	frameCopy = sortDataFrame(frameCopy, joinIndex);
	}
	}
	}

it looks like options.enforceSort is not set by default. the reason for this is isLikelyAscendingVector() must scan the entire dataset to be sure everything is sorted. this would impose a performance penalty for the 95% case. maybe we can make this assertion much cheaper and simply test just the first and last datapoints 🤔.

[cameronwaterman]

maybe we can make this assertion much cheaper and simply test just the first and last datapoints 🤔.

Seems like an improvement over the current behavior at least. I would hope that we could give randomly sorted data as well, but it sounds like that would break interpolation. We could potentially take a decimated (or random) sample of the dataset and perform the comparison on those data points - this would help avoid the case where the first and last are sorted but intermediate points are not.

[leeoniya]

the linked PR should cheaply handle when the first and last values indicate the set is not sorted ascending (e.g. reverse sorted).

I would hope that we could give randomly sorted data as well,

it's an edge case i think, since we have few issues that get opened about random data. in this scenario it's reasonable to expect a manual Sort by transform to be applied instead of incurring the cost of assertion for the 99.5% case.