So far I'm really impressed by marimo. I've been able to put together a dynamic GUI faster than with other tools I tried in the past while hitting less bugs. I have a use case where I want to let the user:

1. choose a number of datasets
2. pick a directory for each dataset
3. for each dataset select channels (e.g. R,G,B). The options to select from are dynamic, determined by content of files selected at 2 and independent for each dataset.

I was able to get something working but the user experience is very clunky. E.g. if the user picks a different directory at 2), the selection made at 3) is reset for all datasets, etc. I might be abusing the intended use case of marimo but I'm wondering if I'm hitting a fundamental limit or if I'm just using it wrong.

Below is a simplified example that works around the problem by defining sel1 and sel2 in different cells. If I don't, a change to either n1 or n2 resets both selections already made.

import marimo

__generated_with = "0.13.4"
app = marimo.App(width="full")


@app.cell
def _():
    import marimo as mo
    return (mo,)


@app.cell
def _(mo):
    def gen_options(n):
        return ["a", "b", "c"] if n % 2 == 0 else ["m", "n", "o"]

    n1 = mo.ui.number(1)
    n2 = mo.ui.number(2)
    return gen_options, n1, n2


@app.cell
def _(gen_options, mo, n1):
    sel1 = mo.ui.multiselect(options=gen_options(n1.value))
    return (sel1,)


@app.cell
def _(gen_options, mo, n2):
    sel2 = mo.ui.multiselect(options=gen_options(n2.value))
    return (sel2,)


@app.cell
def _(n1, n2, sel1, sel2):
    [n1, n2], [sel1, sel2]
    return


if __name__ == "__main__":
    app.run()

This hack however doesn't really work in the practical use case I described above.

• the number of mo.ui.multiselect can be larger than 2
• It's only known at runtime
• 3 steps are chained instead of 2.

Trying to apply a similar workaround becomes really messy...