Describe the enhancement requested
Currently, in the Parquet library there is a significant infrastructure for computing the byte ranges that are necessary to read for a given set of columns and row groups in order to pre-buffer the file data via various caching mechanisms. Unfortunately, none of this infrastructure is externally exposed via the API.
There's no way for a consumer to simply query what byte ranges of the file were going to be necessary to read specific columns and row groups without actually implementing their own ReadRangeCache and the FileReader assuming that they are actually performing the caching.
We can add a simple function to the FileReader which returns the computed column chunk ranges, coalescing them as necessary, to provide this functionality information.
Component(s)
C++, Parquet
Describe the enhancement requested
Currently, in the Parquet library there is a significant infrastructure for computing the byte ranges that are necessary to read for a given set of columns and row groups in order to pre-buffer the file data via various caching mechanisms. Unfortunately, none of this infrastructure is externally exposed via the API.
There's no way for a consumer to simply query what byte ranges of the file were going to be necessary to read specific columns and row groups without actually implementing their own
ReadRangeCacheand the FileReader assuming that they are actually performing the caching.We can add a simple function to the
FileReaderwhich returns the computed column chunk ranges, coalescing them as necessary, to provide this functionality information.Component(s)
C++, Parquet