GraphScope
GraphScope copied to clipboard
refactor(interactive): Refactor the `AbstractArrowFragmentLoader` for better concurrency
Currently, graph loading is done concurrently at the granularity of files. However, there is a significant size difference between files. We should perform concurrency at the granularity of RecordBatch.
Also, ODPSFragment can refactored to allow parallel stream reading, since it allows splitting table by rows.
Closed since already implemented.