Brian McCallister wrote: > How hard would it be to abstract the filesystem access completely and go > directly against in-memory implementations, rather than "getting sloppy" > with disk i/o?
A file system abstraction is already there, package org.apache.derby.io, just waiting for an in-memory implementation. As Dain pointed out the issues with a memory based database is the ability to do post test analysis on the data. Though one could normally run the test with in-memory and switch to disk for analysis of failures. > 6 seconds per test is still pretty expensive time wise. Well, that depends on what is happening in the test :-), Derby tests tend not to be a single test case, but a number of test cases. Dan.
