How can I get a dataflow graph (DAG) from a notebook via CLI or API?

[Meng6]

I checked the PyPI page and it mentioned that we could convert a Jupyter notebook to a marimo notebook with the CLI:

marimo convert your_notebook.ipynb > your_notebook.py

Similarly, is it possible to get a dataflow graph (DAG) from a notebook via CLI or API? For example, the input is a Jupyter notebook and the output is a JSON string:

{
    "cell1": {
        "usesVariable": [],
        "declaresVariable": ["v1", "v2"],
    },
    "cell2"{
        "usesVariable": ["v1"],
        "declaresVariable": ["v3"]
   },
    ...
}

Any suggestions? Thank you!

[mscolnick]

this is not currently possible, but seems like a useful artifact. we definitely have this knowledge.

what is the use case? do you need the DAG by itself? if you are coming from a jupyter notebook, it is not guaranteed to be a DAG (could have cycles and re-declarations)

[mscolnick]

cell's are zero indexed. so it is used by cell1 and cell3, which is correct

[Meng6]

Isn't it used by val = test(1, 3)?

[mscolnick]

transitively, yes, but the DAG only shows "parent/children" not "ancestors/descendants". the later can be inferred

[Meng6]

I see, thank you!

[Meng6]

Thanks for all the help! I need to think about my project carefully & might contribute to Marimo in the future :)
Thanks for developing this fancy notebook!