Repository navigation
Don't connect dots unless they are adjacent in the data #339
Description
This is a rare issue that can make affected charts somewhat confusing. Imagine that a particular population is only in the data on two days: January 1, 2017 and January 1, 2019. Ensemble will show a line connecting those two points, even though that population isn't actually in the dataset for most of that time.
This can happen on the Top Languages chart. The chart shows only the top 5 languages. For some time, zh-CN and es-ES were fighting for 5th place. zh-CN overtook es-ES exactly twice: once in December 2018 and once in April 2019. It wasn't in 5th place in January, February, or March. Nonetheless, Ensemble drew a line between December 2018 and April 2019 for zh-CN, implying that it was one of the top 5 languages that whole time when it was not.
One solution might be to show separate dots in these situations and only draw lines between dots that represent consecutive dates in the dataset.
Activity
stephaniehobson commentedon Nov 6, 2026
Still valid after #481, which replaces metrics-graphics with Observable Plot. The new chart joins each population's points in date order, as the old one did, so a population that drops out of the data for a few weeks still gets a straight line across the gap.
It's more common than the Top Languages example suggests. In today's live data, 28 populations on 8 charts have weeks missing within their date range. The worst:
zh-CN on Top Languages still has 6 missing weeks. In every case the missing weeks are absent from the data, not stored as nulls.
Observable Plot makes the fix simpler. A Plot line leaves a gap wherever a point's value is missing, so the chart can:
This should be done after #481 lands, as a separate change.
stephaniehobson commentedon Nov 6, 2026
Current:
Proposed: