Error plot distribution of categorical feature

[ReinhardSellmair]

Describe the bug A ValueError is thrown when plotting distribution of categorical feature.

To Reproduce I'm using version 0.10.3 Running following code:

import nannyml as nml
from IPython.display import display

reference_df, analysis_df, _ = nml.load_synthetic_car_loan_dataset()
display(reference_df.head())

column_names = ['car_value', 'salary_range', 'debt_to_income_ratio', 'loan_length', 'repaid_loan_on_prev_car', 'size_of_downpayment', 'driver_tenure', 'y_pred_proba', 'y_pred']

calc = nml.UnivariateDriftCalculator(
    column_names=column_names,
    treat_as_categorical=['y_pred'],
    timestamp_column_name='timestamp',
    continuous_methods=['kolmogorov_smirnov', 'jensen_shannon'],
    categorical_methods=['chi2', 'jensen_shannon'],
)

calc.fit(reference_df)
results = calc.calculate(analysis_df)

figure = results.filter(column_names=results.categorical_column_names, methods=['chi2']).plot(kind='distribution')
figure.show()

raises following error: ValueError: all input arrays must have the same shape

[nnansters]

Hey @ReinhardSellmair , thanks for submitting your report!

This looks like a local issue, since it didn't trip any tests in the automated build and I also couldn't reproduce this on a fresh installation.

Are you working in a fresh environment as well, or does this occur after updating NannyML?

Would you happen to have some more logging, so we can check where the issue arises?

[ReinhardSellmair]

Thanks for your quick response. I installed nannyML version 0.10.3 to my an existing environment. Here is the full error log:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
File <command-3447662672388356>, line 20
     17 calc.fit(reference_df)
     18 results = calc.calculate(analysis_df)
---> 20 figure = results.filter(column_names=results.categorical_column_names, methods=['chi2']).plot(kind='distribution')
     21 figure.show()

File python/lib/python3.10/site-packages/nannyml/usage_logging.py:238, in log_usage.<locals>.logging_decorator.<locals>.logging_wrapper(*args, **kwargs)
    236 finally:
    237     if runtime_exception is not None:
--> 238         raise runtime_exception
    239     else:
    240         return res

File python/lib/python3.10/site-packages/nannyml/usage_logging.py:187, in log_usage.<locals>.logging_decorator.<locals>.logging_wrapper(*args, **kwargs)
    184 runtime_exception = None
    185 try:
    186     # run original function
--> 187     res = func(*args, **kwargs)
    188 except BaseException as exc:
    189     runtime_exception = exc

File python3.10/site-packages/nannyml/drift/univariate/result.py:249, in Result.plot(self, kind, *args, **kwargs)
    234     return plot_metrics(
    235         self,
    236         title='Univariate drift metrics',
   (...)
    246         metric_name='Method',
    247     )
    248 elif kind == 'distribution':
--> 249     return plot_distributions(
    250         self,
    251         reference_data=self.reference_data,
    252         analysis_data=self.analysis_data,
    253         chunker=self.chunker,
    254     )
    255 else:
    256     raise InvalidArgumentsException(
    257         f"unknown plot kind '{kind}'. " f"Please provide on of: ['drift', 'distribution']."
    258     )

File python/lib/python3.10/site-packages/nannyml/plots/blueprints/distributions.py:83, in plot_distributions(result, reference_data, analysis_data, chunker, title, figure, x_axis_time_title, x_axis_chunk_title, y_axis_title, figure_args, subplot_title_format, number_of_columns)
     80 x_axis_is_time_based = is_time_based_x_axis(analysis_chunk_start_dates, analysis_chunk_end_dates)
     82 if column_name in result.categorical_column_names and method in result.categorical_method_names:
---> 83     figure = _plot_stacked_bar(
     84         figure=figure,
     85         row=row,
     86         col=col,
     87         chunker=chunker,
     88         column_name=column_name,
     89         metric_display_name=method,
     90         reference_data=reference_data[column_name],
     91         reference_data_timestamps=reference_data[result.timestamp_column_name]
     92         if x_axis_is_time_based
     93         else None,
     94         reference_alerts=reference_result.alerts(key),
     95         reference_chunk_keys=reference_result.chunk_keys,
     96         reference_chunk_periods=reference_result.chunk_periods,
     97         reference_chunk_indices=reference_result.chunk_indices,
     98         reference_chunk_start_dates=reference_result.chunk_start_dates,
     99         reference_chunk_end_dates=reference_result.chunk_end_dates,
    100         analysis_data=analysis_data[column_name],
    101         analysis_data_timestamps=analysis_data[result.timestamp_column_name] if x_axis_is_time_based else None,
    102         analysis_alerts=analysis_result.alerts(key),
    103         analysis_chunk_keys=analysis_result.chunk_keys,
    104         analysis_chunk_periods=analysis_result.chunk_periods,
    105         analysis_chunk_indices=analysis_result.chunk_indices,
    106         analysis_chunk_start_dates=analysis_chunk_start_dates,
    107         analysis_chunk_end_dates=analysis_chunk_end_dates,
    108     )
    109 elif column_name in result.continuous_column_names and method in result.continuous_method_names:
    110     figure = _plot_joyplot(
    111         figure=figure,
    112         row=row,
   (...)
    133         analysis_chunk_end_dates=analysis_chunk_end_dates,
    134     )

File python/lib/python3.10/site-packages/nannyml/plots/blueprints/distributions.py:285, in _plot_stacked_bar(figure, column_name, metric_display_name, reference_data, reference_data_timestamps, analysis_data, analysis_data_timestamps, chunker, reference_alerts, reference_chunk_keys, reference_chunk_periods, reference_chunk_indices, reference_chunk_start_dates, reference_chunk_end_dates, analysis_alerts, analysis_chunk_keys, analysis_chunk_periods, analysis_chunk_indices, analysis_chunk_start_dates, analysis_chunk_end_dates, row, col, hover)
    276 if has_reference_results:
    277     reference_value_counts = calculate_value_counts(
    278         data=reference_data,
    279         chunker=chunker,
   (...)
    282         missing_category_label='Missing',
    283     )
--> 285     figure = stacked_bar(
    286         figure=figure,
    287         stacked_bar_table=reference_value_counts,
    288         color=Colors.BLUE_SKY_CRAYOLA,
    289         chunk_indices=reference_chunk_indices,
    290         chunk_start_dates=reference_chunk_start_dates,
    291         chunk_end_dates=reference_chunk_end_dates,
    292         annotation='Reference',
    293         showlegend=True,
    294         legendgrouptitle_text=f'<b>{column_name}</b>',
    295         legendgroup=column_name,
    296         subplot_args=subplot_args,
    297     )
    299     assert reference_chunk_indices is not None
    300     analysis_chunk_indices = analysis_chunk_indices + (max(reference_chunk_indices) + 1)

File python/lib/python3.10/site-packages/nannyml/plots/components/stacked_bar_plot.py:143, in stacked_bar(figure, stacked_bar_table, color, chunk_start_dates, chunk_end_dates, chunk_indices, subplot_args, annotation, **kwargs)
    131     hover.add(data['value_counts_normalised'], name='value_counts_normalised')
    132     hover.add(data['value_counts'], name='value_counts')
    134     figure.add_trace(
    135         Bar(
    136             name=category,
    137             x=x,
    138             y=data['value_counts_normalised'],
    139             orientation='v',
    140             marker=dict(line_color=color, color=category_colors_transparent[i], line_width=0),
    141             yperiodalignment="start",
    142             offset=0,
--> 143             customdata=hover.get_custom_data(),
    144             hovertemplate=hover.get_template(),
    145             hoverlabel=dict(bgcolor=category_colors_transparent[i], font=dict(color='white')),
    146             **kwargs,
    147         ),
    148         **subplot_args,
    149     )
    151 # Shade chunk type
    152 x0 = chunk_start_dates.min() if is_time_based_x_axis(chunk_start_dates, chunk_end_dates) else chunk_indices.min()

File python3.10/site-packages/nannyml/plots/components/hover.py:60, in Hover.get_custom_data(self)
     57 if not isinstance(self.custom_data[0], (List, np.ndarray)):
     58     return np.asarray([self.custom_data, self.custom_data])
---> 60 return np.stack(self.custom_data, axis=-1)

File python3.10/site-packages/numpy/core/shape_base.py:449, in stack(arrays, axis, out, dtype, casting)
    447 shapes = {arr.shape for arr in arrays}
    448 if len(shapes) != 1:
--> 449     raise ValueError('all input arrays must have the same shape')
    451 result_ndim = arrays[0].ndim + 1
    452 axis = normalize_axis_index(axis, result_ndim)

ValueError: all input arrays must have the same shape```

[michael-nml]

Hi @ReinhardSellmair, I'm also unable to reproduce this using NannyML 0.10.3 in my environment. There's probably a different dependency version somewhere that is causing this.

Would you be able to share the output of pip freeze in the environment where you're seeing this issue?

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[anopsy]

I'm experiencing exactly the same problem: This code figure = results.filter(column_names=results.categorical_column_names, methods=['jensen_shannon']).plot(kind='distribution') leads to this error coming from nannyml Hover, get_custom_data(), and it's call to np.stack result in this message ValueError: all input arrays must have the same shape My versions: {'nannyml': '0.12.1', 'pandas': '2.2.3', 'polars': '0.20.31', 'pyarrow': '14.0.2', 'numpy': '1.24.4'} python:3.10.12

Btw, maybe for the debugging purposes (and better communication with our users) we could implement .show_version() and ask for output from it in the "report a bug"-type of New Issue?

EDIT: Fixed it with using the right environment : {'nannyml': '0.12.1', 'pandas': '1.5.3', , 'pyarrow': '14.0.2', 'numpy': '1.24.4'}, pandas >2 could be the culprit