Github user xuchuanyin commented on the issue:
https://github.com/apache/carbondata/pull/2670
Problems that may be ignored are that during loading
1. We use a buffer to store one row and the row is 2MB fow now
2. For a column page, we compress it as a byte array and its length is of
'INT' datatype, which means that a page at most 2GB.
So do we have any restrictions on the binary data type?---
