Hi!
Is there any documentation on how to use a custom data backend or at least a custom pandas dataframe shape?
I have semi-discretized data in the form of:
x y intensity (for color)
1.0 100.0, 200.0, 300.0,.. 34.0, 59.0, 0.0, ...
4.0 120.0, 213.0, 305.0,.. 37.0, 0.0, 60.0, ...
...
with its own backend to make fast queries (also directly from disk) given an x,y range.
Can I somehow let datashader use this backend? If so, how? Just visualizing Points will be fine.
Thanks for any hints!
No, there is no documentation available on this. Currently Canvas.points accepts x and y columns that are single-valued. It would be possible to add support for x to remain single-valued by allow y to be multi-valued and reuse the same x values, but this isn’t trivial. Datashader is written in a specific way to take advantage of numba's performance and this usually means that adding new functionality is quite a lot of work.
The easiest way to use datashader with your example dataset would be to separate out the multiple y values per row into multiple rows each so that each new row only has a single y value.
I see. What a pity! Thank you for the explanation. I am indeed using your exact workaround for now.