I’m building a plotting library which basically aims to quickly plot hundreds of parameters on a figure (or say, 6 figures on a page with hundreds of timeseries parameters on each), and each series has 10k to 1m samples. Datashader seems like the obvious piece to do the heavy lifting, and I’m trying to use Holoviews to provide the zoom/pan/hover capability. But what are my best options for hosting/storing the data, to optimize performance?
Holoviews/Datashader give many examples using Pandas or Dask Dataframes, which I think are good for “synchronous” timeseries, but my data is very “asyncronous”. Currently I’m trying to use a dictionary of Pandas Series objects, creating a Holoviews Curve for each, and then plotting with Holoviews/Datashader.
I’m wondering if there’s another approach, which would optimize performance? Perhaps a dictionary of Dask Arrays? Or maybe Dask Dataframes would efficiently handle numerous asynchronous signals?