dataframe is not recognized as input in Colab

[danielsparing]

The Colab version linked from the Jupyter Notebooks example currently fails at:

%sql output_df << SELECT sum(i) as total_i FROM input_df
output_df
# Returns:
# RuntimeError: If using snippets, you may pass the --with argument explicitly.
# For more details please refer: https://jupysql.ploomber.io/en/latest/compose.html#with-argument
# 
# 
# Original error message from DB driver:
# (duckdb.duckdb.CatalogException) Catalog Error: Table with name input_df does not exist!
D# id you mean "pg_attrdef"?
# 
# LINE 1: SELECT sum(i) as total_i FROM input_df
#                                       ^
# [SQL: SELECT sum(i) as total_i FROM input_df]
# (Background on this error at: https://sqlalche.me/e/20/f405)
# 
# If you need help solving this issue, send us a message: https://ploomber.io/community

[danielsparing]

Some more details:

This seems to only happen with SQLAlchemy. If I replace %sql duckdb:///:memory: with:

conn = duckdb.connect()
%sql conn --alias duckdb

, then it works¹.

Footnotes

1. well, it then runs into https://github.com/ploomber/jupysql/issues/1038 which is already fixed in the latest release, so it works if I remove the version pinning !pip install --quiet jupysql # ==0.10.12. But maybe that was there for a reason? :) And in general, SQLAlchemy was avoided for a reason? ↩

[danielsparing]

fyi I also wrote up a similar but different issue.