Hi Team! This might be more of a Pylance question, but run into it while using Metaflow Extensions so wanted to ask it here. I'm creating a new Metaflow Extension for our internal company's use. I was able to create the extension and import from metaflow, but Pylance cannot resolve the dependency. Took a glance at how Metaflow adds extensions to the main library, and I believe it has to do with the dependencies not being in the actual module, but rather getting loaded later. While this is not an actual issue, I wanted to ask if anyone has run into this before and if so, how did you solve it? 😢

you can generate stubs that are specific for your extension. Once you have everything installed, you should be able to do:

metaflow develop stubs --help

to see available options.

It worked! Thank you for your help! 🙂

great!

Do you have an example of how you package the pre-built stubs together? 🤔 The stubs get stored into the virtual env that I use to run the command, so I'm unsure on how to best package them with the extension

we just build the stubs package when releasing (something like this;

${my_python} -c "from metaflow.cmd.develop.stub_generator import StubGenerator; StubGenerator('./metaflow-stubs').write_out()"

That is then used to create a package

my-mf-extension-stubs

and then we have something like this in

my-mf-extension

extras_require={
        "stubs": ["my-mf-extension-stubs==%s" % version],
    },

So then you can install

my-mf-extension[stubs]

to get the optional stubs. You could also choose to make it a part of the requirements (and not an

extra_require

)

Awesome! Thank you for your help! 🙂