Bivariate filters can be really useful and we should consider ways to make these easy to specify in the UI. It would be part of the "Filter" sidebar, where currently univariate filters are activated by clicking on a button for the variable we want to filter on. How should 2 variables be specified to be used together for a bivariate filter?
Options:
After selecting one variable and activating its univariate filter UI, have an option to add an additional variable which will create a bivariate filtering UI
Click and drag one button onto another button to active a bivariate filtering UI for the two variables
Others?
But there are more questions:
How to specify which variable goes on which axis?
How to handle categorical / continuous / date / datetime variable combinations?
What does the filter closing button in the footer cause to happen to a variable that's currently in a bivariate visual filter? (We will keep brush selection in 2d mode to be square so that filters are still orthogonal so removing the filter for one variable won't effect any others, so this is more of a UI question).
Other technical challenges here will be:
Creating a useful bivariate visual filter UI using d3 / brushes
Using hexbins (we don't want to show every single data point in our 2d scatterplot) which we will need to be able to do on the fly with crossfilter (somehow...)
Bivariate filters can be really useful and we should consider ways to make these easy to specify in the UI. It would be part of the "Filter" sidebar, where currently univariate filters are activated by clicking on a button for the variable we want to filter on. How should 2 variables be specified to be used together for a bivariate filter?
Options:
But there are more questions:
Other technical challenges here will be: