Grand-prix

Tests, stress-test and performance-tests for Common Lisp object-databases and persistence libraries.

Why?

For evaluating object databases and comparing syntax and features.
Makes implementation of new object databases easier, since you don't have to invent all the new testcases.
A wishlist for new features, add a test and see if someone implements it.
Maybe improves the quality of object databases over time?
Writing tests is a good way to learn the object-databases.

Some inspiration, such as the name Grand-Prix and the name of some classes, come from a similar project in Java called PolePosition

Current status:

2007-feb-07, the first upload, codename "basic". Contains:

Code to run tests
Poor mans multiprocessing, makes it possible to write tests that run over parallel lisp processes.
Only one test, named Anderstorp after the Swedish racetrack, checks if several processes can write at the same time.
A driver for rucksack (expected to fail on the Anderstorp test).
A driver for elephant (expected to work on the Anderstorp test).
A driver for cl-perec (expected to work on the Anderstorp test).
Some documentation

History

No history yet.

Project members

Henrik Hjelte

Mailing list

The mailing list is grand-prix-devel.

License

MIT (BSD without advertising clause)

Download

Please be aware that the code will currently probably not work out of the box. If you are not interested in coding your own tests, downloading it is probably a waste of time.

darcs get http://common-lisp.net/project/grand-prix/darcs/grand-prix

Contribute

Pick anything on the todo list and do it. Update the todo list with new todos. Just because you put it there doesn't mean you have to do it...

Send critique, comments and ideas to the mailing list.

Todo/ideas

More tests

More backend drivers.

Clean up things

Document things

Now the rucksack-anderstorp testcase fails, which is expected. Maybe this should be communicated when running the test, otherwise people might think that rucksack is not working, when the fact is that rucksack does not have multi-processing capabilities.

Add time measurements.

Add ad-hoc metrics (see the comment in the boavista sample).

Think of a nice way to vary the specific backend, for example allow the tests easily to be run both for elephant-bdb and elephant-clsql. Can be done now with inheritance, but is that nice?

Make an easy but nicer configuration for hardcoded stuff like paths, connection strings and other things that vary between users/computers.

Make a uniform syntax layer for rucksack-elephant-allegrocache-perec. That way you can could write tests that fit all these databases using the same syntax.

Anything else you can think of that improves grand-prix. Send in patches!