No Sub-Packages of scipy (or numpy) can be Imported
#619
Comments
|
Thanks for opening an issue, I'm looking into it
I am unable to replicate this, is there an import from Numpy in particular that you see failing? |
Run in my test repo results in: 
|
|
Nice test 👍🏽 That's a one we should add in the Numpy recipe |
|
I read a bit too fast. Those "subpackages" it cannot import are not subpackages. Those are attributes. e.g. there is no |
|
I guess that's what you wanted to do: import pkgutil
import numpy
for submodule in pkgutil.iter_modules(numpy.__path__):
try:
__import__(f"numpy.{submodule.name}")
except ImportError:
print(f"failed to import {submodule.name}") |
|
That's weird, I can import Using the same package versions as you do. |
|
...I can't confirm. On your JupyterLite site, I can't even import 
This is on Chrome Version 116.0.5845.96 (Official Build) (arm64) under macOS 13.5.1 (22G90). |
|
This is just a warning, in your screenshot, scipy was imported correctly. EDIT In order to get rid of the warning, we need to update the scipy recipe to use the latest Scipy version, a PR is opened for this but it needs fixing |
|
It looks like the following results in But the following works: |
|
Ok, so from my side:
I just tried this - it works for me as well. Unfortunately, I cannot even reproduce the original error in our environment (https://brightway-lca.github.io/brightway-hub/): 
The kernel is always 'busy', but won't execute a single cell. I tried:
I am not sure if this should be a separate issue? I couldn't find any relevant issues in the JupyterLite repo. I'm not sure how to proceed - I can't even get the JupyterLite site to even execute a cell with |
|
This looks very much like a cache issue. Can you try running JupyterLite in a new private window? |
|
After some more testing. It looks like the It may be because matplotlib and scipy are vendoring a different version of qhull? |
This worked - now I can at least reproduce the original issue (thank you 🙏). |
|
Bear with me, we'll get there 😄 |
|
Hello, hello! Can we help with this issue in any way, @martinRenou? |
|
...as far as I can see, the error messages (eg. here for #527) are rather non-descript. However, the configuration in |
|
Hey, @martinRenou - while we're still waiting for pyodide 0.24.0 to drop, could you update us on the scipy/matplotlib issue? Do you have any idea what the underlying issue could be? Could we help speed up the process? As I mentioned above, the error-messages are too non-descript for me to help you without further guidance 🙈 |
|
This will be fixed by #643 |
|
Amazing, thank you! |
|
You should be able to test it with |
|
I tried to test this just now - works fine. Many thanks for the fix! |
If I try to import a subpackage from
scipy:in a JupyterLite environment built with
jupyter_server=2.7.1
jupyterlite-core=0.1.2
jupyterlite-xeus-python=0.9.4
I get the error:
This affects our attempt to run Brightway in JupyterLite, since we are using functions from
scipy.interpolate.This works fine in the 0.23.4 version of Pyodide:
The text was updated successfully, but these errors were encountered: