pyhere in Google Collab?

[chillerb]

Hi! Thanks for pyhere - I like it a lot, especially for getting the project root in Jupyter Notebooks.

Unfortunately, I just noticed that issues might arise, if one of your notebooks is launched interactively in Google Collab - for example, via the UI from a Jupyter Book.

Since the notebook is mounted independently from the main project into /content/ inside the Collab, pyhere fails to detect any root indicators.

A quick monkey patch might be to add .config or sample_data to root_indicators (which also seem to be present in the Collab working directory), but this is probably not reasonable, since .config files might also appear elsewhere.

Are there other recommendations on how to use pyhere in (Collab) notebooks? Thanks!

[joshpsawyer]

Hi! I haven't tried it myself but give me a bit and I'll see if I can fix it. I don't do a lot with Jupyter / Collab.

What are the minimal steps to reproduce the issue?

[chillerb]

Hi Josh, thanks for your reply. As mentioned, it is not really an issue with pyhere, since everything is working as intended, just an unfortunate interaction between Jupyter Books/Colab and pyhere.

For demonstration, I created this minimal repository:
https://github.com/chillerb/jb-colab-pyhere

Now that I had some time to think about it, I think what I'm really asking for is just a new feature in pyhere to configure the fallback option, if no root indicators are found.

For example, something like this:

# default behavior as is
here(path, fail="warning", fallback="/")

# raises an error if no root indicator was found
here(path, fail="error")

# on fail, provides the fallback path (would solve the specific Colab issue for me)
here(path, fail="ignore", fallback=".")

Maybe it would also be nice to be able to specify some absolute paths, where here stops traversing the file tree back if encountered:

# would also solve the issue in colab
here(path, stop_on=["/content"])

[joshpsawyer]

those are easy enough to provide - give me a couple days to update it

[chillerb]

Thank you! Maybe just one last thought from my side: From a user's perspective, it would probably be nice if I could just call a function to configure it once, so I don't have to provide all arguments again on each here call. E.g., something like:

pyhere.set_fallback(".")
pyhere.set_on_fail("error")