qcheck_gleeunit_utils
This package provides utility functions for working with Gleam’s gleeunit test framework.
While it may be more broadly useful, this library is mainly developed for internal use in the qcheck library. As such, no guarantees about API stability will be made until qcheck itself is more stable.
Contents
Compilation target
The functions provided in this package only work properly on the Erlang target. They will not manage your tests properly on the JavaScript target.
Usage
The following sections show some common use cases. More info can be found in the docs.
(Click the tiny arrows to expand the sections containing example code.)
Run all tests in parallel
If you want to keep things as simple as possible and simply run all your tests in parallel, you may replace the call to gleeunit.main
with run.run_gleeunit
.
Click to expand the example code!
import gleeunit/should
import qcheck_gleeunit_utils/run
pub fn main() {
run.run_gleeunit()
}
pub fn example_1_should_pass__test() {
do_work()
should.equal(1, 1)
}
pub fn example_2_should_fail__test() {
do_work()
should.equal(1, 2)
}
pub fn example_3_should_pass__test() {
do_work()
should.equal(1, 1)
}
pub fn example_4_should_fail__test() {
do_work()
should.equal(100, 200)
}
import gleam/list
// A small function simulating some CPU bound work.
fn do_work() {
let _l = list.range(0, 10_000_000)
Nil
}
Be aware that any individual test that takes longer than 5 seconds (the default EUnit timeout), will cause the whole test suite to come crashing down.
Run tests with long timeouts
Here is an example of a test that will timeout (…at least, it will timeout on my laptop).
Click to expand the example code!
import gleeunit
import gleeunit/should
pub fn main() {
gleeunit.main()
}
pub fn hello__test() {
do_work(25)
should.equal(1, 1)
}
import gleam/list
// A small function simulating some CPU bound work.
fn do_work(i) {
case i >= 0 {
True -> {
let _l = list.range(0, 10_000_000)
do_work(i - 1)
}
False -> Nil
}
}
If you run that with gleam test
, (and you’re computer isn’t too much faster than mine), then you should get a timeout error that looks sort of like this:
Pending:
qcheck_gleeunit_utils_test.hello__test: module 'qcheck_gleeunit_utils_test'
%% Unknown error: {timeout,
... lots more stack trace ...
Okay, so that test is taking too long, so it crashes with the timeout error.
We can avoid that by using a test generating function along with the test_spec.make
function.
Click to expand the example code!
import gleeunit
import gleeunit/should
// Add this import statement.
import qcheck_gleeunit_utils/test_spec
pub fn main() {
gleeunit.main()
}
// Add a trailing `_` (underscore) character to the test name to specify that it
// is a function for generating rather than a test itself.
pub fn hello__test_() {
// And use the `test_spec.make` function here.
test_spec.make(fn() {
do_work(25)
should.equal(1, 1)
})
}
import gleam/list
// A small function simulating some CPU bound work.
fn do_work(i) {
case i >= 0 {
True -> {
let _l = list.range(0, 10_000_000)
do_work(i - 1)
}
False -> Nil
}
}
You could also write the hello__test_()
function with the use
syntax if you prefer.
pub fn hello__test_() {
// And use the `test_spec.make` function here.
use <- test_spec.make
do_work(25)
should.equal(1, 1)
}
Now, run that with gleam test
and you will get a passing test. (On my computer, it takes about 7 seconds.)
*Note that this example can also be used with the run.run_gleeunit
as opposed to the gleeunit.main
if you want to parallelize all the tests.
Run parallel test groups
Sometimes, you may want only certain tests to be run in parallel, perhaps because you have some tests that interact with some external state or involve some other tricky setup or teardown that may interact poorly with other tests running in parallel.
For this, we use the TestGroup
s, which are created with the test_spec.run_in_parallel
and test_spec.run_in_order
functions. For this use case, we will use test_spec.run_in_parallel
to create a TestGroup
that will specify tests to be run in parallel.
Click to expand the example code!
import gleam/list
import gleeunit
import gleeunit/should
import qcheck_gleeunit_utils/test_spec
pub fn main() {
gleeunit.main()
}
pub fn in_order_1__test() {
do_work(5)
should.equal(1, 1)
}
pub fn in_order_2__test() {
do_work(5)
should.equal(1, 1)
}
pub fn in_order_3__should_fail__test() {
do_work(5)
should.equal(1, 3)
}
fn in_parallel_1() {
do_work(5)
should.equal(1, 1)
}
fn in_parallel_2() {
do_work(5)
should.equal(1, 1)
}
fn in_parallel_3__should_fail() {
do_work(5)
should.equal(1, 11)
}
pub fn in_parallel__test_() {
[in_parallel_1, in_parallel_2, in_parallel_3__should_fail]
// Note the use of `test_spec.make` here.
|> list.map(test_spec.make)
|> test_spec.run_in_parallel
}
// A small function simulating some CPU bound work.
fn do_work(i) {
case i >= 0 {
True -> {
let _l = list.range(0, 10_000_000)
do_work(i - 1)
}
False -> Nil
}
}
Gotchas
There is a tricky gotcha that you need to be aware of with test groups. It occurs when you give different timeouts for functions in a test group.
Click to expand the example code!
import gleam/list
import gleeunit
import gleeunit/should
import qcheck_gleeunit_utils/test_spec
pub fn main() {
gleeunit.main()
}
// All of these tests should fail, and they are all too long for the default
// timeout.
fn in_parallel_1__should_fail() {
do_work(25)
should.equal(1, 1)
}
fn in_parallel_2__should_fail() {
do_work(25)
should.equal(1, 10)
}
fn in_parallel_3__should_fail() {
do_work(25)
should.equal(1, 100)
}
pub fn in_parallel__test_() {
[
// The `make` calls by default use a really long timeout.
test_spec.make(in_parallel_1__should_fail),
// For this one, we set a timeout of 1 second.
test_spec.make_with_timeout(1, in_parallel_2__should_fail),
test_spec.make(in_parallel_3__should_fail),
]
|> test_spec.run_in_parallel
}
// A small function simulating some CPU bound work.
fn do_work(i) {
case i >= 0 {
True -> {
let _l = list.range(0, 10_000_000)
do_work(i - 1)
}
False -> Nil
}
}
If you run that with gleam test
, while you should see three test failures, you will only see the first test failure. I suspect this is because the timeout error triggered by the second test silently brings down the whole test group.
If you change the timeout to 100
seconds (or really any long enough value), and rerun gleam test
, you will see all three tests failing as expected.
To summarize, be careful with test groups when one or more of the tests may timeout–it will silently crash the whole test group, swallowing any failures (or passes) for some (or all) of the tests.
Run in-order tests in a parallel context
If what you really want is all tests to be run in parallel, it can be a real drag to specify all your tests in parallel test groups. So it is easy enough to use the run.run_gleeunit
parallel helper. However, what if you need some of the tests to be run in order?
You can do that by creating a TestGroup
with the test_spec.run_in_order
function.
Click to expand the example code!
import gleam/list
import gleeunit/should
import qcheck_gleeunit_utils/run
import qcheck_gleeunit_utils/test_spec
pub fn main() {
run.run_gleeunit()
}
pub fn in_parallel_1__test() {
do_work(5)
should.equal(1, 1)
}
pub fn in_parallel_2__test() {
do_work(5)
should.equal(1, 1)
}
pub fn in_parallel_3__should_fail__test() {
do_work(5)
should.equal(1, 3)
}
pub fn in_parallel_4__test() {
do_work(5)
should.equal(1, 1)
}
pub fn in_parallel_5__test() {
do_work(5)
should.equal(1, 1)
}
pub fn in_parallel_6__test() {
do_work(5)
should.equal(1, 1)
}
fn in_order_1() {
do_work(5)
should.equal(1, 1)
}
fn in_order_2() {
do_work(5)
should.equal(1, 1)
}
fn in_order_3__should_fail() {
do_work(5)
should.equal(1, 11)
}
fn in_order_4() {
do_work(5)
should.equal(1, 1)
}
fn in_order_5() {
do_work(5)
should.equal(1, 1)
}
fn in_order_6() {
do_work(5)
should.equal(1, 1)
}
pub fn in_parallel__test_() {
[
in_order_1,
in_order_2,
in_order_3__should_fail,
in_order_4,
in_order_5,
in_order_6,
]
|> list.map(test_spec.make)
|> test_spec.run_in_order
}
// A small function simulating some CPU bound work.
fn do_work(i) {
case i >= 0 {
True -> {
let _l = list.range(0, 10_000_000)
do_work(i - 1)
}
False -> Nil
}
}
If you run that with gleam test
, you will see the first 6 test complete more or less all at once, then the following six tests complete one-by-one.
License
Copyright (c) 2024 Ryan M. Moore
Licensed under the Apache License, Version 2.0 or the MIT license, at your option. This program may not be copied, modified, or distributed except according to those terms.
See licenses/gleeunit_license.txt
for the original gleeunit license.