read from source

[knaaptime]

random question @martinfleis... any idea why i can read these datasets directly from the web using pandas but not geopandas? e.g. i can do

import pandas as pd
# vanilla pandas
df = pd.read_parquet("https://github.com/oturns/example_datasets/blob/main/acs/ca_tracts_2021.pq")

just fine, but

import geopandas as gpd

df = gpd.read_parquet("https://github.com/oturns/example_datasets/blob/main/acs/ca_tracts_2021.pq")

gives ArrowInvalid: Unrecognized filesystem type in URI: https://github.com/oturns/example_datasets/raw/refs/heads/main/acs/ca_tracts_2021.pq

[knaaptime]

ah, ok, i can get it to work with

import geopandas as gpd
from fsspec import filesystem
fs = filesystem('https')

df = gpd.read_parquet("https://github.com/oturns/example_datasets/raw/refs/heads/main/acs/ca_tracts_2021.pq", filesystem=fs)

but i'm curious if geopandas should handle that automatically like pandas does?

[martinfleis]

Pandas does some custom handling here while we directly call pyarrow, which needs fs if it is remote. Ideally we should deal with it under the hood, yeah.