CHIA-2090 Add a section about monkey-patching in tests to PRETTY_GOOD_PRACTICES.md #19053
Conversation
AmineKhaldi
added
Changed
AmineKhaldi
force-pushed
the
pgp_tests_monkeypatch
branch
from
b543aab to
6016876
Compare
AmineKhaldi
force-pushed
the
pgp_tests_monkeypatch
branch
from
6016876 to
6e0cfbf
Compare
AmineKhaldi
changed the title
| @@ -619,6 +619,9 @@ class SomeWallet: | |||
| pytest does not. | |||
| Our tests are fully dependent on pytest so there's no use in retaining `unittest` compatibility for this single point. | |||
| Making a few tests `unittest` compatible is not useful compared with the cost of inconsistency. | |||
| - Avoid monkey-patching (assigning to a method on a class object or instance) as much as possible. | |||
| If necessary, opt for it in an explicit and scoped manner using the `monkeypatch: pytest.MonkeyPatch` fixture with a context manager, instead of silencing mypy using `# type: ignore[assignment]`. | |||
| While in both cases we're signaling to the reader that something smelly/shady is going on, and type safety is lost, the monkeypatch approach with a context manager allows us to make the scope of the change controlled consistently. | |||
There was a problem hiding this comment.
I fail to see the benefit of using monkey patch instead of mypy suppressions.
using monkey patch is much more verbose, and risk diluting the important logic.
what kind of "scope" is more consistently controlled with monkey patch compared to a suppression?
do you mean "consistent" in the sense that the code base should only use money patch and no mypy suppressions?
There was a problem hiding this comment.
fIrst, i didn't approve this because i wasn't quite comfortable with it, but I'll have to go back and review discussions with amine to remind myself what i had considered.
but, directly to the scope question, we certainly do have cases where we just assign to stuff and the effects just live on forever. being explicit about the scope with a context manager clarifies when the patch is intended to be removed. this certainly is less relevant when patching attributes of instances and more relevant when patching classes, modules, and functions.
(p.s. go ahead and raise a concern :])
No description provided.