fix(validation): catch bad input at the public boundary with specific error messages #592

Activity

1. murray-ds

   added

   type:enhancementImprovement to existing functionalityImprovement to existing functionality

   on Oct 22, 2026
2. murray-ds commented on Oct 22, 2026

   

   murray-ds

   on Oct 22, 2026

   ContributorAuthor

   More actions

   Failing tests for the error-message cases above. All 12 fail on main @ c5f4b2a.

   The two RFMSegmentation date cases don't raise at construction at all: the DuckDB ConversionException only appears when .df is read, which is a further reason to validate at the boundary.

   The bugs these checks turned up are filed separately: #593, #594, #595, #600, #601. The suspected SegTransactionStats failure with no customer column did not reproduce: without a customer column it works and drops the customer metrics. Only an integer segment_col with the default rollup_value fails.

   import pandas as pd
   import pytest
   
   from openretailscience.analysis.cohort import CohortAnalysis
   from openretailscience.analysis.gain_loss import GainLoss
   from openretailscience.options import ColumnHelper
   from openretailscience.segmentation.rfm import RFMSegmentation
   from openretailscience.segmentation.segstats import SegTransactionStats
   from openretailscience.segmentation.threshold import ThresholdSegmentation
   
   cols = ColumnHelper()
   
   
   def _transactions(dates: list) -> pd.DataFrame:
       return pd.DataFrame(
           {
               cols.customer_id: [101, 101, 102, 103],
               cols.transaction_id: [9001, 9002, 9003, 9004],
               cols.unit_spend: [24.99, 61.50, 12.75, 149.00],
               cols.store_id: [11, 11, 24, 24],
               cols.transaction_date: dates,
           },
       )
   
   
   DATES = pd.to_datetime(["2024-12-30", "2024-11-02", "2024-06-15", "2024-12-31"])
   
   
   class TestDateColumnErrors:
       """A non-date transaction_date is rejected with an error naming the column."""
   
       @pytest.mark.parametrize(
           "dates",
           [
               pytest.param(["2024-01-05", "2024-02-11", "2024-02-20", "2024-03-02"], id="iso_string"),
               pytest.param([20240105, 20240211, 20240220, 20240302], id="integer_yyyymmdd"),
           ],
       )
       def test_cohort_analysis(self, dates):
           with pytest.raises((TypeError, ValueError), match=cols.transaction_date):
               CohortAnalysis(_transactions(dates), aggregation_column=cols.customer_id, period="month")
   
       @pytest.mark.parametrize(
           "dates",
           [
               pytest.param([20241230, 20241102, 20240615, 20241231], id="integer_yyyymmdd"),
               pytest.param(["12/30/2024", "11/02/2024", "06/15/2024", "12/31/2024"], id="non_iso_string"),
           ],
       )
       def test_rfm_segmentation(self, dates):
           with pytest.raises((TypeError, ValueError), match=cols.transaction_date):
               RFMSegmentation(_transactions(dates), current_date="2024-12-31", r_segments=2, f_segments=2, m_segments=2)
   
   
   class TestParameterNamedInError:
       """Errors name the parameter the caller got wrong."""
   
       @pytest.mark.parametrize("current_date", ["31/12/2024", "2024-13-01"])
       def test_rfm_invalid_current_date(self, current_date):
           with pytest.raises(ValueError, match="current_date"):
               RFMSegmentation(_transactions(DATES), current_date=current_date, r_segments=2, f_segments=2, m_segments=2)
   
       def test_threshold_segmentation_invalid_agg_func(self):
           with pytest.raises(ValueError, match="agg_func"):
               ThresholdSegmentation(_transactions(DATES), thresholds=[0.5, 1.0], segments=["Light", "Heavy"], agg_func="summ")
   
       def test_cohort_analysis_invalid_agg_func(self):
           with pytest.raises(ValueError, match="agg_func"):
               CohortAnalysis(_transactions(DATES), aggregation_column=cols.unit_spend, agg_func="summ")
   
       @pytest.mark.parametrize("segment_col", [cols.store_id, cols.customer_id])
       def test_seg_transaction_stats_integer_segment_default_rollup(self, segment_col):
           with pytest.raises(ValueError, match="rollup_value"):
               SegTransactionStats(_transactions(DATES), segment_col=segment_col, calc_total=True)
   
       def test_gain_loss_group_col_is_customer(self):
           df = _transactions(DATES).assign(brand=["Coke", "Pepsi", "Coke", "Pepsi"])
           with pytest.raises(ValueError, match="group_col"):
               GainLoss(
                   df=df,
                   p1_index=df[cols.transaction_date] < "2024-07-01",
                   p2_index=df[cols.transaction_date] >= "2024-07-01",
                   focus_group_index=df["brand"] == "Coke",
                   focus_group_name="Coke",
                   comparison_group_index=df["brand"] == "Pepsi",
                   comparison_group_name="Pepsi",
                   group_col=cols.customer_id,
               )
   
   
   class TestInputTypeErrors:
       def test_cohort_analysis_rejects_non_dataframe(self):
           with pytest.raises(TypeError, match="DataFrame"):
               CohortAnalysis([1, 2, 3], aggregation_column=cols.unit_spend)
3. murray-ds

   closed this as completedin #625on Oct 29, 2026
4. murray-ds

   added a commit that references this issue on Oct 29, 2026

   fix(validation)!: catch bad input at the public boundary with specifi…

   ...

   This issue will close once commit 9fe8f14 is merged into the 'main' branch.

   9fe8f14