Details
-
New Feature
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
-
None
Description
For Parquet file reading that is not IO-bound, we can go faster by reading columns in multiple threads (assuming underlying the IO source is threadsafe). The code will be very similar to that in https://github.com/apache/arrow/blob/master/python/src/pyarrow/adapters/pandas.cc#L1193