feat(sql): respect display.dataframes config setting

[chrisflath]

Summary

When display.dataframes is set to "plain", mo.sql() now uses plain() to format output instead of the rich interactive table viewer.

This makes SQL cell output consistent with regular DataFrame display behavior when the user has configured plain table output.

Changes

• Added check for include_opinionated() in marimo/_sql/sql.py before creating table output
• When include_opinionated() returns False (plain mode), uses plain(df) instead of table.table()

Testing

Added tests in tests/_sql/test_sql.py:

• test_sql_plain_output_when_not_opinionated: Verifies that Plain object is used when include_opinionated() returns False
• test_sql_rich_output_when_opinionated: Verifies that table is used when include_opinionated() returns True

Both tests mock include_opinionated() and verify the correct output type is passed to replace().

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
marimo-docs	Ready	Preview, Comment	Jan 30, 2026 11:21pm

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[mscolnick]

there is some duplicate code from below:

            if SQLAlchemyEngine.is_cursor_result(df):
                display_df = SQLAlchemyEngine.get_cursor_metadata(df)
            elif DBAPIEngine.is_dbapi_cursor(df):
                display_df = DBAPIEngine.get_cursor_metadata(df)

you might be able to do to t = plain(display_df) (plain will run back through our formatters, but skip making them pretty dataframes)

[chrisflath]

Good suggestion! A minimal elif instead of the long manual handling. The pull request has been updated with a simplified implementation.

[chrisflath]

I have read the CLA Document and I hereby sign the CLA

[Light2Dark]

Would you be able to add some tests? Probably under test_sql.py, may need to mock different return values of include_opinionated()

[Light2Dark]

I think it'd be better to move this import to below if output for consistency. In general I'd prefer moving all imports to top of the file, but for lazy loading and consistency, could move it below output imo

[chrisflath]

sure

[chrisflath]

Would you be able to add some tests? Probably under test_sql.py, may need to mock different return values of include_opinionated()

Forcepushed tests.

[Light2Dark]

just a few touch ups, i think it looks good!

[Light2Dark]

better to move this import to the top of file. It's in a test file so it's fine.

[Light2Dark]

Is it possible to just do sql("SELECT 1"). We don't need to create & drop a new table for this.

[Light2Dark]

Hmm, this test seems to fail for me. The fixture cannot be found (not sure if/how CI passes)
You can patch it in the code itself for simplicity

    @patch("marimo._sql.sql.replace")
    @patch("marimo._output.formatters.df_formatters.include_opinionated")
    @pytest.mark.skipif(
        not HAS_POLARS or not HAS_DUCKDB, reason="polars and duckdb required"
    )
    def test_sql_plain_output_when_not_opinionated(
        self, mock_include_opinionated, mock_replace
    ):
        """Test that SQL uses plain() when include_opinionated returns False."""
        mock_include_opinionated.return_value = False
        import polars as pl

        from marimo._output.formatting import Plain

        # Test query
        result = sql("SELECT 1")
        assert isinstance(result, pl.DataFrame)

        # Should call replace with Plain object
        mock_replace.assert_called_once()
        call_args = mock_replace.call_args[0][0]

        # The call should be a Plain object (not a table)
        assert isinstance(call_args, Plain)