The original implementation only supported a simple graph with a single output, but now that we support multiple outputs it is probably a good idea to have a way to identify the task from which an output batch was created. Upon writeout these may need to go to distinct tables or partitions.
The original implementation only supported a simple graph with a single output, but now that we support multiple outputs it is probably a good idea to have a way to identify the task from which an output batch was created. Upon writeout these may need to go to distinct tables or partitions.
Run could have this signature:
where the tuple is
(task_name, dataframe)
added options to writeto include partitioning on task or having outdir map from task to outdir