[BUG]: floating-point artefacts in tick labels

[jendrikw]

Description

Plotly shows a tick label of −0.0000000000000000888178, which is just silly.

Screenshots/Video

Steps to reproduce

https://codepen.io/Shack-1-Spoilage/pen/MYjdYoR

[alexcjohnson]

Thanks for the report! The default tickformat works fine, but ~r and some of the other d3-format specs show this behavior.

In this case the behavior is clearly a bug, since you didn't specify tick0 so it defaults to 0 therefore it's unambiguous that the tick should be exactly at zero. And we could make a narrow fix that says "if the tick value is close to tick0 (much closer than whatever we've calculated as dtick), set it to exactly tick0. That would cover the 99% use case.

The one case it wouldn't cover is if someone sets a nonzero tick0 but there's still a tick that should fall exactly on zero. Like tick0=1 and dtick=0.1. If we wanted to ensure we're robust in that situation, but still allow things that are not quite that to honor the user's intent like tick0=1.0001 and dtick=0.1, we'd need some sort of heuristic for how close to zero, as a fraction of dtick, counts as zero. One millionth? One billionth?

[camdecoster]

Thanks for the report! This does indeed look like a bug. Our team likely won't have time to address this for a while, but can prioritize a PR review if you want to look into it.