Open aleqx opened 9 months ago
Hi. You are correct that the graphs just divide the time into 10 segments and calculates the date at each of those points. This definitely needs improving. It needs to be a little more broad than your suggestion there just because it should adapt to whether the range is best represented in blocks of minutes, hours, days, months etc. But yes this needs doing really.
My suggestion was only an example. You'd use whatever scale the data span has (days, hours, minutes, etc).
Consider the CSV data below. Note how the first entry is close to 00:00 UTC (01:00 UTC+1) on 30 Sept and the last entry is close to 12:30 UTC (13:30 UTC+1) on 7 Oct. Screenshot below. Also included reference (correct) plot of the same data underneath.
It seems your plot takes the date range, and divides that into a fixed 10 segments, and then places tick marks (vertigal grid lines) at values which are NOT multiples of 24h ... which makes it look wrong when interpretting the graph (*). It would really be better to make sure tick marks are at day boundaries for the graph to make good sense. If your limit is max 10 grid segments, then if the date range is N days, then place grid lines every k days where k=ceil(N/10), where the first and last grid segments would be smaller than 24h, which is fine.
(*) Visual bugs: