add_analysis no longer works

[paddymul]

Checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of Buckaroo.

What type of jupyter notebook were you using (VSCode notebook, google colab, Jupyter Lab, Jupyter notebook). Select multiple if you can reproduce this in multiple environments. If other, please add to description.

Other (please describe in Issue Description)

Reproducible example

class Quin99Analysis(StylingAnalysis):
    provides_defaults = {'quin99': None}

    @staticmethod
    def series_summary(sampled_ser, ser):
        if pd.api.types.is_numeric_dtype(ser) and not pd.api.types.is_bool_dtype(ser):
            return dict(
                quin99=ser.quantile(.99))
        return {}
    
    pinned_rows = [{'primary_key_val': 'quin99', 'displayer_args': {'displayer': 'obj' }},
                  {'primary_key_val': 'dtype', 'displayer_args': {'displayer': 'obj'}},
                  ]
    df_display_name = 'quin99'
    data_key = "empty"  # the non pinned rows will pull from the empty dataframe
pbw = BuckarooWidget(df)
pbw.add_analysis(Quin99Analysis)

Issue description

bw.add_analysis() should work

Expected behavior

Installed versions

Details

``` Replace this line with the output of buckaroo.debug_packages(). Leave the backticks in place. ```

Jupyter Log output

[paddymul]

Implementation Plan

Overview

Fix the broken add_analysis public API so users can add custom analysis classes.

Steps

1. Reproduce — try calling BuckarooWidget.add_analysis(MyCustomAnalysis) and capture the error/traceback.

2. Diagnose — the analysis registration was likely refactored when the pluggable analysis framework was updated. Check:

   • Does add_analysis still exist as a method?
   • Does it correctly append to analysis_klasses?
   • Is the analysis graph rebuilt after adding?
   • Are there import-time vs runtime ordering issues?

3. Fix — depending on root cause:

   • If method was removed: re-add it with the correct signature
   • If it adds to the list but the graph isn't rebuilt: trigger graph rebuild
   • If there's a class hierarchy issue: ensure custom analysis classes can extend the base properly

4. Verify related issue class based add_analysis #424 (class-based add_analysis) — determine if this fix should also implement the improved API proposed there, or keep them separate.

5. Tests:

   • Unit test: define a custom ColAnalysis subclass, call add_analysis, verify it appears in summary stats
   • Test with both pandas and polars widgets

Files to modify

• buckaroo/buckaroo_widget.py — add_analysis method
• buckaroo/pluggable_analysis_framework/analysis_management.py — graph rebuild
• Possibly buckaroo/polars_buckaroo.py — polars variant

@codex

[chatgpt-codex-connector]

Summary

• Kept the widget-facing analysis list synchronized after dynamic registration by updating BuckarooWidgetBase.add_analysis(...) to mirror dataflow.analysis_klasses after a successful add. This helps preserve the expected public API behavior for class-based additions. buckaroo/buckaroo_widget.pyL223-L227
• Hardened CustomizableDataflow.add_analysis(...) with explicit type validation so the API clearly accepts only ColAnalysis subclasses before attempting pipeline updates/rebuild. buckaroo/dataflow/dataflow.pyL428-L435
• Added an explicit graph-rebuild hook in the legacy analysis pipeline and invoked it during add_analysis(...), ensuring dependency/order recomputation is explicit when new analyses are registered. buckaroo/pluggable_analysis_framework/analysis_management.pyL230-L238 buckaroo/pluggable_analysis_framework/analysis_management.pyL262-L275
• Added regression coverage for class-based add_analysis on pandas widgets via a custom Quin99Analysis class and assertion that the new metric/display appears after registration. tests/unit/basic_widget_test.pyL43-L57 tests/unit/basic_widget_test.pyL87-L94
• Added parallel regression coverage for polars widgets via Quin99PolarsAnalysis, asserting the custom metric row is present in serialized all-stats output after add_analysis(...). tests/unit/polars_basic_widget_test.pyL63-L66 tests/unit/polars_basic_widget_test.pyL80-L88

Testing

• ⚠️ pytest -q tests/unit/basic_widget_test.py -k add_analysis_with_custom_class (fails during collection because pandas is unavailable in this environment).
• ✅ python -m compileall buckaroo/buckaroo_widget.py buckaroo/dataflow/dataflow.py buckaroo/pluggable_analysis_framework/analysis_management.py tests/unit/basic_widget_test.py tests/unit/polars_basic_widget_test.py

Committed on the current branch and created a follow-up PR via the make_pr tool.

View task →

[paddymul]

Status: Appears to be working on main

add_analysis() is implemented and functional:

• buckaroo/buckaroo_widget.py line 223: add_analysis() method delegates to dataflow
• buckaroo/dataflow/dataflow.py line 428: full implementation
• buckaroo/lazy_infinite_polars_widget.py line 856: polars implementation with pinned_row_configs support

This may have been fixed as part of the PAF v2 refactor in PR #515. If the original reporter can confirm it works on the latest release, recommend closing this issue. Related: #424 (class-based add_analysis improvements).