Drop in the accuracy of the graph with large time series

[DDorch]

First, thank you for this very amazing package which give us the availability to interact with graphs!

I met an issue with very large (several thousands of data) where I noticed, when I zoom on the graph, that all the data are not present on it.

length.out = 10000
dygraph(as.xts(data.frame(x = cos((1:length.out)/100), row.names = seq.POSIXt(from = ISOdate(2018,1,1), by = "5 min", length.out = length.out))))

This gives you this:

And now if you try with more data

length.out = 30000
dygraph(as.xts(data.frame(x = cos((1:length.out)/100), row.names = seq.POSIXt(from = ISOdate(2018,1,1), by = "5 min", length.out = length.out))))

You have this, with only one data per day instead of one every 5 minutes which is a bit crappy:

Is there a way to avoiding this limitation? I noticed that the threshold where this limitation appears depends on the example. Here it's between 20000 and 30000 points, on other example it is at 68330...

Thanks in advance for your help as I would like to figure large time series with the availability to zoom in details at 5 minutes time step.

[przmv]

It looks like an upstream bug, I'll try to prepare pure (non-R) steps to reproduce and submit an issue to https://github.com/danvk/dygraphs