Bug: histogram silently ignores data, when bin size is too small

[cnowzohour]

When plotting histograms with very fine grained bins, some data will be ignored. See:

https://codepen.io/cnowzohour/pen/qmYmoM

vs

https://codepen.io/cnowzohour/pen/vmjmjM

Both histograms should show 10 bars, but the second one only shows 5 - the only difference being the bin size of 0.005 vs 0.001.

I realized this when visualizing timestamps over one day with a bin size of 10 seconds:

plotly/plotly.R#1001

[etpinard]

Thanks for the report. Looks like this hard-coded 5000 bin limit is the culprit. For example, bumping that number to 50000 yields:

as desired.

We should come up with a better upper bound. Maybe something like x.length / bins.size?

[cnowzohour]

Yeah I guess 5000 can be a bit conservative in cases when the data is sparse and only a small subset of the bins are actually drawn. Also, it would be great if it was somehow visible if this limit is exceeded (and data is omitted). Maybe via a big last bin, containing all the remaining data?

[etpinard]

Update: changing the while clause from

while(i < binend && pos.length < 5000) {}

to

while(i < binend && pos.length < (pos0.length / bins.size)) {}

breaks some of our tests. We'll have to dig deeper to find a solution.