What is best practice for null values in CSV files?

General Discussion
1

What is the best practice for handling null values in CSV files that is to be imported with <loadData>?

I googled on beforehand and used an empty field, i.e. nothing between the commas.

But that gave me problems when I tried to use a dumped CSV to restore a table in a JUnit test, using derby as the in-memory database:
The data type '0' is not supported

I managed to fix this problem for numeric and boolean columns by setting nullPlaceHolder to the empty string
<column name="CONTENTLENGTH" type="NUMERIC" nullPlaceholder=""/> <column name="REQUIRE_LOGIN" type="BOOLEAN" nullPlaceholder=""/> <column name="GROUP_BY_YEAR" type="BOOLEAN" nullPlaceholder=""/>

But trying to set this on a string column gives me the derby failure The data type '0' is not supported again
<column name="CONTENTTYPE" type="STRING" nullPlaceholder=""/>

I would like my dumped CSV files to work as well out of the box as possible with whatever tools uses them, be it excel or a database import or setting up a mockrunner result set.
But my primary usage will be using liquibase to load the CSV into a table.
So, what is liquibase best practice for saving null values to CSV files?
Is best practice different depending on the column type?
Is best practice different depending on the target RDBMS?

Thanks!

  • Steinar