Add special ParquetBlobStorage allowing filtering
Parquet is de facto standard for storing big data.
However we usually do not need all date stored in parquet hive/files.
It would be great if there were ParquetBlobStorage - extension of BlobStorage that:
a) would understand parquet format i.e. in order to read schema it will read only end of the file
b) allow selective column/row reading - no need to copy full blob when we need just couple columns or group of row.
Pliszka, Jacek commented
c) allow for type casts like string to string(10) decimal to int