java work

Good response...

I was just wondering what database you guys are using. We are trying to adopt DBUnit but are having a hard time managing referential integrity when reloading the database with the test data (mostly when deleting old test data).

Could you share a pointer as to how you accomplish that?

We are doing this on multiple oracle schemas.

As for referential integrity there are two ways
i) drop all constraints on the test databases. This is yucky but makes life easier
ii) keep track of the order in which your tables must be loaded. You then know the order in which they must be removed. Although we don't do it this way, this is definatly the way I want to do it the next time I do setup automated tests. My current idea is to keep track of table dependencies and compute all necessary tables for seting up any table. Am I making any sense to you ?

I don't think option 1 will be usable in practice since you wouldn't be able to detect errors that occur from bad data being inserted.

I have also thought along the lines of your second suggestion but when you have a fair number of tables it becomes almost impossible to figure the right sequence out. It can of cause be that there is not easy way and since is simply the price to pay for leveraging dbunit in your setup.

One think I have been wondering - do you guys use one "master" dataset for testing or do you have multiple smaller ones for specific tests ? I thought of splitting my datasets into more manageble chunks but discarded the approach since it would be unmanagable to run the tests from the build file. Any thoughts ?

lekkim

I don't think so

Mikael,

Do you think there cannot be a quick algorithm to figure out the correct sequence in which tables can be inserted if their dependencies are tracked ? I haven't given it any thought but really cannot find any reason why not, that's "just" graph theory isn't it ?

As for our test data, yes, we have a set of files that describes a coherant database. But we do have some test-specific files that we sometimes use. Or we just insert plain Strings that describe the XML for the table in some other tests... it's just a matter of how many lines there are.
I must confess that we have dropped all constraints on the test DB though which makes this easier :(

Here are some strategies I use to circumvent aforementioned issues.

1. Computing insert sequence automatically
There is a new class in DbUnit 2.0, DatabaseSequenceFilter, which does that. I’m using it to export my DTD and make sure that all my datasets conform to it when editing them. Using it once prevent adding overheads when executing my tests.
http://www.dbunit.org/faq.html#tableseq

2. Deleting old data
I’m using a strategy similar to CLEAN_INSERT but in two steps and with two different datasets. Note that I’m not extending the DatabaseTestCase class but implemented my own. First, I’m doing a DELETE_ALL with the above DTD as dataset to clean the whole database. I’m filtering out configuration, read-only and rarely used tables from the DTD. Second, I perform an INSERT with my test class dataset.

3. One large dataset vs multiple small datasets
I’m personally using multiple small datasets. One dataset per test class, having the same name and automatically loaded by my base class. Datasets are relatively easy to manage because there is a one to one relationship between datasets and test classes. This way I’m very confident to not break other test classes when I add or modify my test class dataset. This also allows more developers to work in parallel without conflict since not editing the same global dataset.

I kind of already discussed these strategies before in the dbunit-user mailing list:
http://sourceforge.net/mailarchive/message.php?msg_id=7063085

I’m currently setting up a DbUnit Wiki. My intention is to allow people to capture, organize and discuss such patterns, best practices and recurrent issues about DbUnit and DatabaseTesting in general. I believe that this Wiki will be a more appropriate medium than the existing mailing list.
http://www.dbunit.org/wiki/

Hoping to read you there!

Your site is very good.

cody

Googl

Ok, I will sign your blog. I really love your site.

Online Casino Gambling Guide with Cross-media Online Casinos

hi, just popped in here through a random link. cool site, keep this good work up :-)

best online casinos

enjoyed reading your posts.

best online casinos

great site and great information.

online casinos

congrats mate! fine job and fine site!

online casinos

Is it worth using DbUnit in a reporting project.
In reporting project you generally read selective data from various tables and display it to the user. So does it make sense to use DbUnit here ?

Thanks for great info

Thanks for good info!

very useful comments - good to read

business grants

Thanks for your useful information!

I agree with you :) Thanks!