[BUG] st_read fails to read Excel file from an http URL

[tboddyspargo]

In duckdb v0.9.2, the following code fails (where it succeeded in 0.9.1):

import duckdb

conn = duckdb.connect(database=':memory:')
conn.execute('SET s3_endpoint="localhost:8333";')
conn.execute('SET s3_url_style="path";')
conn.execute('SET s3_use_ssl=false;')
conn.execute('INSTALL spatial; LOAD spatial;')
r = conn.execute(
    "SELECT * FROM st_read('http://localhost:8333/appdata/global/sba_sample.xlsx', layer='sba_sample')"
)

print(r.fetchdf())

In this case, http://localhost:8333 is a local seaweedfs container compliant with the S3 API (and sba_sample.xlsx exists at that path).

With this exception:

Traceback (most recent call last):
  File "tmp/excel_example.py", line 7, in <module>
    r = conn.execute(
duckdb.duckdb.IOException: IO Error: Unknown file type

[tboddyspargo]

I'll note that this was working in duckdb 0.9.1 and it has been our work-around due to the other failure mentioned in #182.

Thanks in advance for your help!

[tboddyspargo]

I suspect this may have something to do with #168, given that the error message matches.

[Maxxen]

Hi! I think this has been fixed in #187.

Note: for XLSX in particular using DuckDB's filesystem with httpfs has a slower initial load time than /vsis3/ or /vsicurl/. Im aware of the issue and will look into improving it more in the future.

In a couple of hours (when the CI finishes) you should be able to install the extension with this fix for 0.9.2 by running:

FORCE INSTALL spatial FROM 'http://nightly-extensions.duckdb.org';

[tboddyspargo]

Thanks @Maxxen - Can fix be made available in the default 0.9.2 spatial extension or in a 0.9.3 fix release?

[Maxxen]

We may eventually replace the default 0.9.2 install version with this fixed version, but I can't give a timeline. If more issues occur it may be delayed further.