The average parse time on a sample of about 40 DB files is around
7 seconds per file (on a very fast G4 with plenty of RAM). Most of
this time is because the coding schemes are normalized on
import.
Some possible approaches:
1) Make the coding scheme implementation faster
2) Allow deferred scheme normalization in some use cases (needs
further use case analysis)
3) Accept slowness of importing -- side benefit is it discourages
importing the same files repeatedly, which should be seriously
discouraged in general.