Execute UniFFI bindings via a simple Python test #87
Conversation
|
@mulmarta, I think this more or less implements your idea of testing using Python. You're right: it's super easy to get going with just a few lines of code! This also very quickly shows that we need a lot more fields or getters on the types to actually make them usable from any of the high-level target langauges 😄 |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #87 +/- ##
=======================================
Coverage 88.88% 88.89%
=======================================
Files 171 172 +1
Lines 31194 31217 +23
=======================================
+ Hits 27727 27750 +23
Misses 3467 3467 ☔ View full report in Codecov by Sentry. |
It would be amazing to have this integrate with the coverage reports — I think that would require making the Python code call into the coverage-instrumented |
This uses the UniFFI test helpers to make the `libmls_rs_uniffi.so` file available to the Python script. I would have liked to use `$CARGO_TARGET_TMPDIR` to place the temporary files in a stable place under `target/`, but this variable is only set for integration tests, not unit tests. So the script now ends up somewhere in your system temp directory. We could decide to only use integration tests for these tests, let me know what you think! The test is very simple here: it simply executes the Python script, which then has to return with a zero exit code. If it fails, the output is printed and you then have to debug the Python code by hand. Related to awslabs#81.
tomleavy
force-pushed
the
test-uniffi-with-python
branch
from
4c90b3e to
39944ea
Compare
Issues:
Addresses #81
Description of changes:
This uses the UniFFI test helpers to compile and make the
libmls_rs_uniffi.sofile available to the Python script.The test is very simple here: it simply executes the Python script, which then has to return with a zero exit code. If it fails, the output is printed and you then have to debug the Python code by hand.
There is no facility to pass data back to the Rust unit test — it would be great to have this, but right now it's a one-way communication only.
Call-outs:
I would have liked to use
$CARGO_TARGET_TMPDIRto place the temporary files in a stable place undertarget/. However, this variable is only set for integration tests, not unit tests. So the script now ends up somewhere in your system temp directory.We could decide to only use integration tests for these tests, let me know what you think! We could also write a bit more code to find a good place under
target/without relying on the environment variable.Testing:
Use
cargo test!By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 and MIT license.