as·say /ˈaˌsā,aˈsā/ noun - the testing of a metal or ore to determine its ingredients and quality.
Rust is great, but the testing leaves much to be desired sometimes. With custom
test frameworks being unstable and only an eRFC since 2018 there's not much we
can do to expand the abilities of our tests right? Well that's where assay
enters the picture. It seeks to solve a few problems when testing in rust:
- Tests all run in the same process which means setting env vars or changing the
working dir affects all of the tests meaning you have to resort to things like
cargo test -- --test-threads=1
or using some kind of mutex whereby you lose the parallelization of running the test suite - Setting up a temporary file system to run things in for a test and having the
test run inside it is a pain to setup and being relative to it by using
std::env::set_working_dir
is prone to the above issues - Including fixtures in a test, let alone multiple, can get a bit verbose
- Setting up and tearing down the same thing for each test can be a lot
- Want to run
async
tests? There's no runtime and you have to do setup just to run it. - Using
?
in your test means putting-> Result<(), Box<dyn std::error::Error>>
on every test and it can be tedious assert_eq
/assert_ne
output can be hard to grok and see why something is equal/not equal
assay
fixes these issues by:
- Running each test as it's own process for you automatically if you use
cargo test
or if you usecargo nextest
then it let's that handle the processes being in parallel in their own process for you. This means you can mutate per process state as much as you want without affecting other tests and always have your tests run in parallel - Setting per process env vars
- Setting up a temporary directory that the test runs in (sort of like
chroot
without the jail aspect and no need forsudo
privileges) - Including files you want into the temp directory by specifying them
- Letting you run async tests by simply adding
async
to the test function - Making all of your tests act as if they returned
Result<(), Box<dyn std::error::Error>>
. Use the?
to your hearts content and no need to add the Eye of Sauron (Ok(())
) to each test - Automatically importing the
pretty_assertions_sorted
crate so that you can have pretty output forassert_eq
,assert_eq_sorted
, andassert_ne
. - Allowing you to define setup and teardown functions to call for the test
assay
was born out of personal frustration with the way things are and wanting
to handle the boilerplate without needing to write a whole test framework, while
also pushing the bounds of what we could have today on stable Rust.
While assay
is capable of a lot right now it's not without issues:
- Tests run in their own process and so getting the output available in a good way is still kind of an open problem
- Sometimes tests that shouldn't pass do, at least when having developed
assay
, because they run in another process. You should intentionally crash your test to make sure it's actually working, because you'll have tests pass that really shouldn't which frankly isn't great - Rust Analyzer gets tripped up sometimes and the error propagates to each
invocation making it harder to track down. In these cases
cargo test
will let you know where the issue actually is - No work on spans yet! This macro just slaps things in and so error messages
are much to be desired without much in the way to tell you why an invocation
of
assay
fails. assay
does not work inside doc tests!
Take a look at HOW_TO_USE.md
(which is included in the crate
documentation) or tests/integration_tests.rs
.
We do not have a Minimum Supported Rust Version and only track stable
. Given
this crate uses 2021 edition rustc
>= 1.56 for now, but that's not always
guaranteed to be the case and later versions might require a greater version
than 1.56.
All files within this project are distributed under the Mozilla Public License
version 2.0. You can read the terms of the license in LICENSE.txt
.