Dave.

It is an experiment to write a module to allow you to write tests like this:

sub test_someting($tempdir) {
    ...
}
[download]

The test harness runs all the individual test_something functions. If the function has arguments that it will prepare the appropriate object and call the function passing those objects into the function. The example I created was injecting a $tempdir object, but, once the system is ready, you could write and register your own objects. In Pythest these are called fixtures, and they mostly act test fixtures. (In Pytest there are a few other ready-made such fixtures, for example one that captures stdout/stderr, and one that can mock methods, attributes, etc.)

If you look at the modules in the t/ directory, those are all examples on how to use the module to write tests.

Is there a reason why they need to be available as parameters at all? These fixtures could just be global variables, couldn't they?

I may have trouble understanding the full intention.... please help me:

You want to parse

sub test_someting($tempdir) {
    # do some tests
}
[download]

And because you see the (registered) keyword "tempdir" your test system is supposed to call this test with a prepared temporary directory?

ok(
    test_something( create_tempdir() ),
    "something",
)
[download]

Kind of a naming convention?

Did I get it right?

If yes, I could think of some alternative (and more perlish) ways to achieve this without needing to parse the signature.

PS: not sure why you call this "dependency injection" or do you plan testruns where the injected $tempdir is not a temporary dir? :)

Cheers Rolf
_{(addicted to the Perl Programming Language :)
Wikisyntax for the Monastery}