Large integers are truncated

[rossviljoen]

Large Int32s in the plot data seem to get truncated, resulting in an unwanted loss of precision when plotting with PlutoPlot(Plot(...)), this is fixed when setting PlutoPlotly.floatval() = Val(false), so I think it's related to the preprocessing.

[disberd]

Indeed by default all numbers are converted to Float32 for saving memory and speeding up transfer between JS and julia.
You can control this with PlutoPlotly.FORCE_FLOAT32[] = false but I guess overwriting the method achieves the same result.

I can probably skip converting Int32s into Float32 but not sure whether converting Int64 to Int32 or Float32. Any opinion on this @rossviljoen?

I also wanted this to be a setting that could be controlled on a plot-by-plot basis (hence the ScopedValue for the flag) but unfortunately the processing is done during the show method which is happening outside of the ScopedValue block, so at the moment it still needs to happen globally for all plots in a notebook...

[rossviljoen]

I don't believe PlutoPlotly.FORCE_FLOAT32[] = false will work with a ScopedValue, as there's no setindex! method for it, hence why it's necessary to override the method. I'm not sure there's that much benefit to it being a ScopedValue anyway given the issue with show you mentioned.

I can probably skip converting Int32s into Float32 but not sure whether converting Int64 to Int32 or Float32

If it's necessary for performance, converting Int64 to Float32 likely makes more sense to avoid overflow issues. I think it would be good to document this behaviour (and how to disable it) in general though.

(PS. thanks for the excellent package!)