You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Because datashader allows automatic ranging and a flexible pipeline that can recompute all plot parameters for every pan and zoom, it is both essential and difficult to provide meaningful keys and legends. For instance, users could insert any arbitrarily nonlinear (including non-monotonic) transfer function into the pipeline, without the downstream client being able to detect that it was added, what its properties are, or how it should affect the values shown in any color key. We'll need an API that makes suitable information available from the core library, as long as the programmer sets things up according to our recommendations, while knowing that in principle there's no way to solve this problem for the fully general case of arbitrary transformations. Hopefully it will be practical to at least semi-automate this process, such that colors can be shown for a set of test values provided to the front of the pipeline, if only for the maximum and minimum values from the viewport.
Tasks (will probably need expanding):
Add a colormap key to one of our Bokeh plots as an example, using fixed values if necessary
Auto-range colormap values to show at least the minimum and maximum values automatically
Support nonlinear sampling for the colormap, working out intermediate values
Add a key to plots of categorical data, showing categories only
Possibly add a values-based key for categorical data, showing numerical ranges in each category
The text was updated successfully, but these errors were encountered:
Because datashader allows automatic ranging and a flexible pipeline that can recompute all plot parameters for every pan and zoom, it is both essential and difficult to provide meaningful keys and legends. For instance, users could insert any arbitrarily nonlinear (including non-monotonic) transfer function into the pipeline, without the downstream client being able to detect that it was added, what its properties are, or how it should affect the values shown in any color key. We'll need an API that makes suitable information available from the core library, as long as the programmer sets things up according to our recommendations, while knowing that in principle there's no way to solve this problem for the fully general case of arbitrary transformations. Hopefully it will be practical to at least semi-automate this process, such that colors can be shown for a set of test values provided to the front of the pipeline, if only for the maximum and minimum values from the viewport.
Tasks (will probably need expanding):
The text was updated successfully, but these errors were encountered: