Hi, I have two standalone html bokeh plots that present different kinds of data in different ways.
The first is 58 MB, the plot has 7 panels, each with one figure than has ~ 30 scatter plots with tens of thousands of points. Figures in different panels are not linked. I use Columndatasource objects in the figure.scatter() lines and I also use webgl.
The second is 7.5 MB, 6 panels, each with 16 figures that each have 12 lines drawn with 70 points each, all figures across all panels are linked in x/y axis. I don’t use Columndatasource objects and each line has x/y arrays specified directly in the figure.line() lines. Each line has an error band drawn with figure.patch().
all the data is in numpy arrays.
The 58 MB standalone html takes ~10 - 15 sec to load and loads all the time.
The 7.5 MB one can be stuck for ~1 min and sometimes the browser (Chrome Version 58.0.3029.96 (64-bit), or Explorer 11.162.10586.0) asks me to wait or kill the page. If I increase further the number of figures per panel I get the kill message every time.
I use bokeh 0.12.4
What could cause a smaller standalone html file to take longer to load?
I’m experience the same extremely lame page load speed with a version 0.12.5. https://drive.google.com/file/d/0BygWfuhlw-2fYU5jalU0dVZRejQ/view?usp=sharing - 6MB - takes 1.7min to load on Chrome / Ubuntu on i7-4690X. I’m guessing it would take less than a second, or somewhere close to init and render with PaperJS.
Actually, I’ve just reloaded jupyter notebook’s kernel and it started generating 400kb files instead which load as expected - split second. Looks like Bokeh was accumulating some stuff over the runs. The more I re-run the script in a notebook the more stuff gets accumulated. My code is close to that of chart / Area example. I’ve attached it just in case.