huggingface URLs don't work

[alexkreidler]

What happens?

I can't use the Huggingface URL Scheme supported by the HTTPFS extension: SELECT * FROM 'hf://datasets/<username>/<dataset>/<file>.csv|parquet|jsonl;

When running these in the DuckDB-WASM Shell, I get:

CREATE VIEW train AS (SELECT * FROM read_parquet('hf://datasets/harvard-lil/cold-cases@~parquet/default/train/*.parquet'));
IO Error: No files found that match the pattern "hf://datasets/harvard-lil/cold-cases@~parquet/default/train/*.parquet"

SELECT count(*) FROM 'hf://datasets/harvard-lil/cold-cases/*.parquet';
IO Error: No files found that match the pattern "hf://datasets/harvard-lil/cold-cases/*.parquet"

Both of these queries work in the DuckDB CLI on my local machine.

To Reproduce

https://shell.duckdb.org/#queries=v0,CREATE-VIEW-train-AS-(SELECT-*-FROM-read_parquet('hf%3A%2F%2Fdatasets%2Fharvard%20lil%2Fcold%20cases%40%3Bparquet%2Fdefault%2Ftrain%2F*.parquet'))~,SELECT-count(*)-FROM-'hf%3A%2F%2Fdatasets%2Fharvard%20lil%2Fcold%20cases%2F*.parquet'~

Browser/Environment:

Firefox

Device:

Windows laptop

DuckDB-Wasm Version:

1.1.2

DuckDB-Wasm Deployment:

shell.duckdb.org

Full Name:

Alex Kreidler

Affiliation:

independent