Optimised datashader install

Datashader
,
1

Hi,

In the docs, it is recommended to install datashader using conda to make sure the right libs are installed for best performance on the targeted platform.

I’d like to know more about this. Is it only OS-related, and if so in what way? I use locked conda environments created on one Linux distribution, installed within a container that’s potentially based on another (and deployed on a host thay may also differ from either but this may be irrelevant.)

Are there specific points to pay attention to in this situation? Everything works fine but if it can be improved in any way I’m interested.

Also, datashader is the main–if not only–reason why I use conda for my environment. I’d like to have more details on how to install it with pip while still making sure the install is optimised. I may create an issue on GitHub regarding this.

2

Datashader runs on numba and numpy AFAIK https://github.com/holoviz/datashader/blob/c77a2e6eab8c7d75e28cec97c744aecbc6b549cc/pyproject.toml#L30-L43

Thus, if you get those correct, Datashader should work well (if my understanding is correct)

3

Ok, if that’s all there is to it, then it’s fairly straightforward.

Thank you!

4

Please note this is my understanding only :slight_smile: I haven’t worked on the Datashader codebase. Perhaps you can throw this question on Discord as well.