Parallel computations #723
Labels
help wanted
Extra attention is needed, feel free to help :)
research
This requires a deeper dive to gather a better understanding
Milestone
Object properties sometimes can be heavy or lazily computed and overall conversion take minutes for somewhat big lists
One can write this fairly simple code to speed up the conversion
Although code is simple, it seems hard to properly make this parallelism part of the
toDataFrame
implementation.Only
list.toDataFrame(maxDepth = int)
andlist.toDataFrame { properties(maxDepth = int) { } }
are side effect free, and it's (mostly) safe to split the list in chunks, run conversion in parallel and concat results. But even computation of the properties can be not parallel friendly. And then there is a question how workload is split and so on.add
andconvert
can be heavy and involve IO too. For this i have something like this in mindUsage:
These two approaches can speed up dataframe code significantly in certain scenarios, so we can give them some visibility in the documentation.
The text was updated successfully, but these errors were encountered: