Also it seems to use a large amount of memory when reading the data: 
...isn't that an order of magnitude larger than the actual data itself?  [ Full content available at: https://github.com/apache/arrow/issues/2614 ] This message was relayed via gitbox.apache.org for [email protected]
