On 14/06/12 15:04, Ali Lown wrote:
>>>> Is anyone against having an instance of Mongo DB instantiated and
>>>> destroyed when tests are run? I'm already assuming the developer doesn't
>>>> need to have Mongo DB already installed, as well.
>>> How much of an overhead is that going to add to the running of the
>>> tests? Inutition suggests that instantiating a DB is going to be a
>>> pretty heavy job...
>> It will only take 30 seconds or a little less/more (depending on your
>> internet connection) to download the Mongo distribution for your
>> OS/arch. After that, database population, tests and shutdown take around
>> 15 seconds (Core 2 Duo 2.4 GHz, 4GB DDR2, SSD disk).
> That sounds reasonable for testing. (Though I don't think fetching
> outside of distro-specific package managers is a good idea. An error
> saying please install MongoDB from your distro's package repository
> before running these tests would be ok.)
It's not going to install Mongo. It's just going to extract it to a
local directory and use it. When the test JVM finishes, the instance
will be shutdown.
If you ever delete that local directory, then it will download again and
so on.
>
>> Btw, just to clarify, I have only one Mongo DB instance for all tests
>> and not an instance per test.
> Can you be sure of full isolation if you setup one instance and then
> run each test? You would need to revert all changes at the end of each
> test before you could run the next...
This *should* be enforced by each test and not the tools we're using,
being it Mongo or any other.
And this has been done before, since every test already implements the
following:
@Override
protected void tearDown() throws Exception {
super.tearDown();
database.dropDatabase();
}
--
Paulo Pires
