There doesn't seem a way to import NULL values in the shell using the .import command. Missing values are treated as empty strings. When I import into a table with constraints this might fail if either NULL or a valid value is enforced, with an empty string not being valid. Furthermore, I have to replace empty strings in nullable fields with NULL using UPDATE commands. I think that .import is much faster than .read with lots of INSERT statements. So, I suggest that .import should provide a new option --nullvalue or so, just like the .nullvalue command is used for exporting. Another idea is that a missing value is NULL, and '' is an empty string. I could probably provide a patch unless there is some objection or a better idea.