Test::Async::Base
– this test bundle contains all the base test tools
use Test::Async::Base;
use Test::Async;
plan 1;
pass "Hello world!";
done-testing
This bundle is supposed to provide same test tools, as the standard Raku Test
. So that
use Test::Async;
plan ... ;
... ;
done-testing
would be the same as:
use Test;
plan ... ;
... ;
done-testing
For this reason this document only tells about differences between the two.
Test tools resulting in either ok or not ok messages return either True or False depending on test outcome. skip
always considered to be successful and thus returns True.
The message set with test-flunks
.
Number of tests expected to flunk. Reduces with each next test completing.
See take-FLUNK
.
If test-flunks
is in effect then method returns its message and decreases $.FLUNK-count
.
Method produces standardized "expected ... but got ..." messages.
The second candidate is used for non-string values. It stringifies them using Test::Async::Utils
stringify
routine and then passes over to the first candidate for formatting alongside with named parameters captured in %c
.
Named parameters:
-
:$exp-sfx
- suffix for "expected", a string which will be inserted after it. -
:$got-sfx
– suffix for "got" -
:$gist
- enforces use of methodgist
to stringify values -
:$quote
- enforces use of quotes around the stringified values
Handler for Event::Cmd::SetTestFlunk
defined by this bundle.
Unlike the standard Test
diag
, accepts a list too allowing similar usage as with say
and note
.
Skips all remaining tests in current suite. If $global
is set then it's the same as invoking skip-remaining
on all suite parents, including the topmost suite.
Mark all remaining tests of the current suite as TODO.
multi subtest(Pair $what, Bool:D :$async=False, Bool:D :$instant=False, :$hidden=False, *%plan)
multi subtest(Str:D $message, Callable:D \code, Bool:D :$async=False, Bool:D :$instant=False, :$hidden=False, *%plan)
multi subtest(Callable:D \code, Bool:D :$async=False, Bool:D :$instant=False, :$hidden=False, *%plan)
subtest
is a way to logically group a number of tests together. The default subtest
behaviour is no different from what is described in Test
. But additionally we can invoke it:
-
asynchronously
-
in random order with other
subtest
s of the same nesting level -
randomly and asynchronously at the same time
A subtest
could also kind of hide itself behind another test tool.
subtest
returns a Promise
kept with True or False depending on subtest
pass/flunk status.
The asynchronous invocation means that a subtest
will be run in a new dedicated thread. The random invocation means that subtest
invocation is postponed until the suite code ends. Then all postponed subtests will be pulled and invoked in a random order.
It is possible to combine both async and random modes which might add even more stress to the code tested.
Some more information about Test::Async
job management can be found in Test::Async::Manual
, Test::Async::Hub
, Test::Async::JobMgr
The particular mode of operation is defined either by plan
keys parallel
or random
, or by subtest named parameters async
or instant
. The named parameters take precedence over plan parameters:
-
if
instant
is set thenplan
'srandom
is ignored -
if
async
is set thenplan
'sparallel
is ignored
For example, let's assume that our current suite is configured for random execution of subtest. Then
subtest "foo", :instant, {
...
}
would result in the subtest
be invoked right away, where it's declaration is encountered, without postponing. Similarly, if parallel
plan parameter is in effect, :instant
will overrule it so it will run right here, right now!
Adding :async
named parameter too will invoke the subtest instantly and asynchronously. And this also means that a subtest invoked this way won't be counted as a job by Test::Async::JobMgr
. In other words, we treat :instant
as: bypass any queue, just do it here and now!
Another edge case is using :async
with random
. In this case the subtest will be postponed. But when time to invoke subtests comes this particular one will get his dedicated thread no matter what parallel
is set to.
Any other named parameters passed to a subtest
are treated as plan keys.
Subset topic variable is set to the backing suite object. For example, this is an excerpt from t/060-subtest.t:
subtest "subtest topic" => {
.plan: 1;
.cmp-ok: $_, '===', test-suite, "topic is set to the test suite object";
}
The example is the recommended mode of operation when a subtest is invoked in a module. In other words, the above example could be written as:
Test::Async::Hub.test-suite.subtest "subtest topic" => {
.plan: 1;
.cmp-ok: $_, '===', test-suite, "topic is set to the test suite object";
}
and this is the way it must be used in a module. See Test::Async
and Test::Async::CookBook
for more details.
Hidden subtest
:hidden
named parameter doesn't change how a subtest runs but rather how it reports itself. A hidden subtest pretends to be integral part of test tool method which invoked it. It means two things:
-
flunked test tools called by subtest code won't report their location (file and line) (implemented by
Test::Async::Reporter::TAP
and might not be supported by 3rd party reporters) -
flunked subtest would report location of the test tool method which invoked it
The primary purpose of this mode is to provide means of implementing compound test tools. I.e. tools which consist of two or more tests which outcomes are to be reported back to the user. The most common implementation of such tool method would look like:
method compound-tool(..., Str:D $message) is test-tool {
subtest $message, :hidden, :instant, :!async, {
plan 2;
my ($result1, $result2) = (False, False);
...;
ok $result1, "result1";
ok $result2, "result2";
}
}
Note that we're using explicit :instant
and :!async
modes to prevent possible side effect related to use of :parallel
and :random
in parent suite's plan. Besides, it is normal for a user to expect a test tool to be semi-atomic operation being done here and now.
This test is similar to is-deeply
as it compares complex structure in depth. The difference is that cmp-deeply
traverses deep into the structure is reports any difference found at the point where it is found. For example:
my @got = [1, 2, %( foo => Foo.new(:foo('13'), :fubar(11)) )];
my @expected = [1, 2, %( foo => Foo.new(:foo(13), :fubar(12)) )];
cmp-deeply @got, @expected, "class instance deep withing an array";
This test would result in a diagnostic message like this:
Which tells us that a difference has been found in an instance of a class (Object) located in a key foo
of an Associative
which is located in the second index of a Positional
. Differences are reported for each attribute where they are found.
Another difference of this test to is-deeply
is that it disrespect containerization status and focuses on structure alone.
This test tool is not provided by the standard Test
framework, but in slightly different forms it is defined in helper modules included in Rakudo and roast tests.
is-run
tests $code
by executing it in a child compiler process. In a way, it is like doing:
Takes the following named parameters (%params
from the first candidate is passed to the second candidate as a capture):
-
:$in
– data to be sent to the compiler input -
:$out?
– expected standard output -
:%env = %*ENV
- environment to be passed to the child process -
:@compiler-args
– command line arguments for the compiler process -
:@args
- command line arguments for$code
-
:$err?
– expected error output -
:$exitcode = 0
– expected process exit code. -
:$timeout
- time in second to wait for the process to complete
This test tool informs the bundle that the following tests are expected to flunk and this is exactly what we expect of them to do! Or we can say that it inverts next $count
tests results. It can be considered as a meta-tool as it operates over other test tools.
The primary purpose is to allow testing other test tools. For example, test t/080-is-approx.t uses it to make sure that tests are failing when they have to fail:
test-flunks 2;
is-approx 5, 6;
is-approx 5, 6, 'test desc three';
Setting $count
to Inf
is the same as using :remaining
named parameter and means: all remaining tests in the current suite are expected to flunk.
(c) 2020-2023, Vadim Belman [email protected]
Artistic License 2.0
See the LICENSE file in this d