At the time of import, user has an option to choose attributes of interest.
In such case there is this set of attributes(here we call them Must Include
Attributes) which are primarily needed for map imported data.
Similarly there are few attributes which are redundant and we decided to
chop them off during export.
Currently, data is loaded onto memory and then goes for further processing.
Hence comes the problem with large BLOB files. We had a discussion to
overcome it, I think @dylan can has better say on this.