Can you add a test for this within src/reactpy/tests/test_web/test_module.py?

You can directly use the GridLayout component if you'd like.

@Archmonger I'm fresh off of vacation and will try to get that test incorporated ASAP. Thanks for already taking a look at this.

So I never actually set up a proper development environment for reactpy. Thus, I've never ran tests or anything. I just barely tried to follow the Contribute Guide to get the test environment set up, and am running into issues.

I'm on a Windows 11 machine and used conda to create a new virtual environment with Python 3.13.2. I installed all of the prerequisites. Then when running hatch test I get the following error:

ImportError while loading conftest 'C:\Users\shawn\Code\reactpy\tests\conftest.py'.
tests\conftest.py:11: in <module>
    from reactpy.config import (
E   ModuleNotFoundError: No module named 'reactpy'

I tried to install reactpy from the cloned repo using pip install -e . and although that seems to work without failing, I continue to get the ModuleNotFoundError above...

Any help would be appreciated. Thanks!

Have you tried using standard CPython venv instead of using conda (ex python -m venv .venv).

I haven't tried our toolchain with Conda yet, so there might be some deficiencies there.

Got it working! And I have the test written. My first stab at writing a test within this framework. Unlike the other examples in test_module.py I had to actually use react and react-dom instead of preact for it to work as expected. With preact I would get an error in the test browser console - something about objects not being able to be passed as children to GridLayout (i.e. the reactpy.html.div children elements were not being properly converted to the expected objects). Let me know if you see any issues with this.

I'm okay with using React for that test. I've seen oddball scenarios where mixed preact/react contexts are hit or miss. That might be the cause of issues with using a parent preact component with react-grid-layout children.

Out of curiosity, was the issue with setting up your development environment related to Conda?

I'm okay with using React for that test. I've seen oddball scenarios where mixed preact/react contexts are hit or miss. That might be the cause of issues with using a parent preact component with react-grid-layout children.

Out of curiosity, was the issue with setting up your development environment related to Conda?

I'm quite sure it was not related to conda. There were two indisputable issues, however. The first was a permission denied error that occurred when the underlying hatch run build:javascript command was running and attempting to cache some packages. I had to run my Command Prompt as Administrator to get past that.

The second issue was much more difficult to track down. I got that same permission error above whether I used conda or the system Python. In both cases, I immediately re-opened the Command Prompt as Administrator and re-executed the hatch test command. However, this time it would no longer fail on the permission denied error, but would instead throw the reactpy ModuleNotFoundError. Which is where I was at when I messaged you about it above.

I eventually determined - my best guess at least - that the underlying isolated test environment must have been getting cached and treated as built even though it had failed with that first permissions error, if that makes sense. All I know is that to finally get it to work, I had to execute the hatch env prune command, after which the hatch test command (as Admin) finally worked.

So again, apparently if something in the underlying build fails when running hatch test for the first time, a manual hatch env prune must be done to recover before rebuilding after fixing the issue with the failed build. That seems like a bug with hatch to me... In fact, after typing that last sentence, I went and searched for it and found this open issue that seems to match what I've described - and I commented on it as well. So there you have it. Way bigger pain than it should have been for sure.

Hatch will always caches environments it creates, which is expected. But Hatch really should have auto-deleted the initial environment if the build command failed.

I agree that issue you linked seems to be the culprit. Hopefully they fix it soon.

There we go. I have the server-side implementation complete with tests updated and passing. Let me know if you prefer the original client-side way I had done it, and I can go ahead and revert.

Very confused as to why a test just failed after my tweak... Is that some random fluke of a fail?