wasm published notebook with pandas read csv error

Question

Trying to change the penguins sample to use pandas instead of polars, results in This cell raised an exception: BadGzipFile('Not a gzipped file (b'sp')')

Myles Scolnick · Accepted Answer

We aren’t doing anything fancy, so might just be a bug in GH pages headers or pandas

Myles Scolnick · Answer

Does the is work locally or just fails in wasm? Looks like pandas could be failing on incorrect headers. Maybe file an issue in their repo?

lovuikeng · Answer

works normal other envs except when serving on gh pages

lovuikeng · Answer

sorry, could have been clear that the issue with following the gh template

lovuikeng · Answer

i mean when trying to use pandas with this sample https://github.com/marimo-team/marimo-gh-pages-template/blob/main/notebooks/penguins.py#L30

lovuikeng · Answer

ok, would have to be using requests to get the csv contents then pass to pd to read as string, looks like pandas and polars have different opinion on url 😦

lovuikeng · Answer

it appears to be that github pages returning the CSV as content-type: text/csv causing the issue on pandas although polars is happy to accept as is, so, another option is to use the "raw" url of the csv file on master branch, of course, this certainly "breaks" the handy utils mo.notebook_location()

Myles Scolnick · Answer

isn't the content type content-type: text/csv correct?

Myles Scolnick · Answer

why would pandas not like that?

lovuikeng · Answer

ya, beats me 😦

lovuikeng · Answer

btw, @Myles Scolnick how do we export wasm of python scripts file used in notebook?

Myles Scolnick · Answer

Can you explain that a bit more?

lovuikeng · Answer

suppose we move that build df into a local python package itself, and import the function in notebook to build the df. it would cause "package not found" on wasm notebook because the local package not being "packed" along with the wasm export

Myles Scolnick · Answer

Ah I see. Marimo doesn’t handle this for you

Myles Scolnick · Answer

We do somewhat if you build on our community cloud, but not when using “marimo export” from the cli

lovuikeng · Answer

i see, meaning need to produce wheel for wasm notebook to do micropip import, or something else?

lovuikeng · Answer

here's one way to build local wheel and install it using micropipbuild wheel, $ uv buildcopy the local wheel to the public folder of the wasm public folderon notebook, micropip install and use the wheel package as normalwhl = str(mo.notebook_location() / "public" / "utils-0.1.0-py3-none-any.whl")
import micropip
await micropip.install(whl) from utils import (...)export wasm of notebook as normal

lovuikeng · Answer

is this what is expected with wheel install of local package in the context of marimo wasm, @Myles Scolnick ?

Myles Scolnick · Answer

Yea that seems like a good solution to me

Myles Scolnick · Answer

I’m not sure of another way

Get help from the marimo community

wasm published notebook with pandas read csv error