Dependency tracking requires save / inconsistent version display between header output and nbproject.__version__

[sunnyosun]

See here:

[Koncopd]

Hi, thanks, this should be fixed on master.

[sunnyosun]

I think the issue was because vscode doesn't immediately autosave the notebook.

If I run Header() twice it displays the new dependencies.

[sunnyosun]

@Koncopd I'm already using the latest version 0.1.6, but somehow the header still says 0.0.3... Is this a bug?

[falexwolf]

The auto-inference of dependencies only works if the current notebook buffer is saved on disk.

Jupyter Lab automatically saves, hence there is no quirks. VS Code requires to hit save before re-running.

I think we have to spell this out in the tutorial or print a logging warning or something like this. 🤔 Otherwise, users will probably get confused.

[falexwolf]

@sunnyosun , I think that also Jupyter Lab doesn't auto-save. I needed to hit save before pandas is detected as a dependency also on Jupyter Lab.

I think that's OK to expect people to hit save. Let's just document it.

[sunnyosun]

Looks like nbproject also didn't detect bionty version correctly. Did I do something wrong?

[Koncopd]

@sunnyosun and what do you have in conda list and pip freeze? I think this happens when you use develop install (for example flit install -s or pip install -e), it freezes the version then in the environment and doesn't update it when the package code changes. The version is updated only when you reinstall the package.

From my conda env

But from python it shows 0.2.0.

[sunnyosun]

Hmm, I have 0.2.0 in both conda and pip, also I've tried to reinstall the package 🤔

[Koncopd]

Hm, interesting, i need to check with bionty.