half lion
What's the best way to make a test depend on the result of another?
One test parses a file and tests if no errors happened.
If no errors happened, I want to use the result in another test to see if it parsed correctly.
Otherwise both tests should fail I guess
Using JUnit btw
floral flickerBOT
<@&987246399047479336> please have a look, thanks.
hard condor
I think this is not possible
Test methods are intended to be independent
But i think u can do
A variable
lean dome
junit itself has no mechanism for test dependencies
u can use before/after/beforeeach/aftereach
half lion
A variable
I can't guarantee the order so that might fail
lean dome
and create a fixed test order
lean dome
its sth like @FixMethodOrder(MethodSorters.NAME_ASCENDING)
or similar
its ofc a bit hacky overall
lean dome
yeah. u can probably make it somehowdo what u need. but firas is right that it smells and that test cases should be independent
so for example u could isolate the setup code
from that one test
and then call that at the beginning of both tests
half lion
and then call that at the beginning of both tests
so just in a separate method you mean?
lean dome
thing is, people are supposed to also be able to run tests individually
instead of only the entire suite
half lion
Yeah that's fair
Okay, I think I'll do that then
I can imagine a test needing some expensive operation you don't wanna do twice
lean dome
yeah. in this case u can try to cache it
half lion
But maybe in that case you could have something lazy initialize a variable
lean dome
in the class
hard condor
Very good expl from zabu
lean dome
gotta be careful with caching things in tests though. u can easily run into situations where u influence tests by what u have saved
in a way that it would perhaps pass if executed in a different order
half lion
oh also 0_0
half lion
so for example u could isolate the setup code
I guess the problem is here
that my setup code needs a test as well
maybe that doesn't matter, they just all fail
but the other tests utilizing the setup code may fail for a weird reason
e.g. if the setup code doesn't throw
but the return value is invalid (and that would need to be tested)
but the other tests don't test that return value
lean dome
are u sure ur actually doing unit tests and not for example an integration test
in which case u might just be "abusing junit"
hard condor
U gotta know the diffs
lean dome
in a nutshell, unit tests are supposed to be really small in terms of what they test. usually they only involve a single method and merely check its edge cases
like "what if i call this method with null or with a negative value or ..."
they wouldnt include complex setups and dependencies
integration tests are higher level and would represent for example an entire flow of an user through ur app
merry shuttle
and if they do use dependencies, mockito time
lean dome
such as "user wants to purchase a product and expects it to work"
which includes many steps and touches multiple classes in ur code
lean dome
in which case there might be better test frameworks then, giving u the flexibility u need
although a lot of people just try their way with junit anways in practice, lol
turbid bough
Tests should also run in isolation, and ideally the order should not matter.
To go back to your initial question @half lion you could use JUnit pioneer => https://junit-pioneer.org/docs/disable-if-test-fails/ and then assume the expected output is the next tests input.
merry shuttle
I didn't ask the question, but thanks for this, I did not know this exists
