A proof-of-concept snapshot testing crate for Bevy.
This crate began as a simple end-to-end test runner forked from the ideas in https://chadnauseam.com/coding/gamedev/automated-testing-in-bevy/.
After a while, I found it useful to capture the inputs of my test runs, so that I did not have to repeatedly run the same inputs to evaluate test behavior. Long term, I'd like to update this to add the capacity to capture and compare the visual output of these test runs in some way, similar to end-to-end test suites for JS/web ecosystem. For the time being, that's somewhat out of scope.
This crate is not published anywhere because it is not really ready for wide
use, and does very little other than attach common Bevy plugins. If you are
interested in using it, feel free to import it via git or let me know by
adding an issue.
Tests run with bevy_geppetto cannot run on the default test harness since they
must be run on the main thread to satisfy winit. One approach is to add test
declarations in an isolated integration or e2e folder, with the following
declaration to Cargo.toml:
[[test]]
name = "<name>"
path = "e2e/<name>.rs"Such a test could then be run with cargo test --test <name>. This runs the
given Bevy program with the following behaviors attached to App:
- DefaultPlugins added, with
return_from_runspecified - bevy_inspector_egui::WorldInspectorPlugin added
bevy::window::close_on_escadded in theUpdateschedule
Inputs are saved in the snapshots directory under a subdirectory inputs. The
root of the snapshot directory can be changed with the SNAPSHOTS_DIR env
variable. The filename of the snapshot will be a kebab-case version of the
string passed as the test's label, and contains one serialized RON object with
all input events per line.
To capture inputs during a test run, pass the -c flag and a rate for registering
the screen-captures:
cargo test --test $TEST_NAME -- -c $FRAMES_PER_SECTo replay inputs from a previous snapshot, pass the -r flag:
cargo test --test $TEST_NAME -- -r