feat: date sanity

.gitignore

@@ -12,6 +12,7 @@ wheels/
 
 # Output files
 *.csv
+*.xlsx
 tests/output/
 *.jsonl
 *.jsonl.gz

nuh_helper/date_shift/__init__.py

@@ -20,6 +20,20 @@
 logger = logging.getLogger(__name__)
 
 
+class DateTooFarBack(Exception):
+    def __init__(self, value: pd.Timestamp) -> None:
+        message = f"the date {value} is too far in the past"
+        super().__init__(message)
+        self._message = message
+
+
+class DateTooFarAhead(Exception):
+    def __init__(self, value: pd.Timestamp) -> None:
+        message = f"the date {value} is too far in the future"
+        super().__init__(message)
+        self._message = message
+
+
 def _get_patient_ids_and_shift_mappings(
     input_file: str,
     patient_sheet: str,
@@ -336,6 +350,8 @@ def shift_excel_dates_inplace(
     seed: int | None = None,
     patient_header_row: int = 0,
     patient_skip_rows: list[int] | None = None,
+    sanity_date_latest: datetime | pd.Timestamp | None = None,
+    sanity_date_earliest: datetime | pd.Timestamp | None = None,
 ) -> None:
     """
     Shift dates in an Excel file, preserving all cell formatting.
@@ -364,7 +380,20 @@ def shift_excel_dates_inplace(
         seed: Optional random seed for generating shifts.
         patient_header_row: Zero-based header row index for the patient sheet (default: 0).
         patient_skip_rows: Optional zero-based row indices to exclude from patient data.
+        sanity_date_latest, sanity_date_earliest: latest and earliest dates allowed in the data. used as a sanity check
+
     """  # noqa: E501
+
+    if sanity_date_latest is None:
+        sanity_date_latest = datetime.now()
+    if not isinstance(sanity_date_latest, pd.Timestamp):
+        sanity_date_latest = pd.Timestamp(sanity_date_latest)
+
+    if sanity_date_earliest is None:
+        sanity_date_earliest = datetime(1900, 1, 1)
+    if not isinstance(sanity_date_earliest, pd.Timestamp):
+        sanity_date_earliest = pd.Timestamp(sanity_date_earliest)
+
     logger.info("Shifting dates in-place: '%s' → '%s'", input_file, output_file)
     logger.debug(
         "Shift range: %d to %d days, seed=%s",
@@ -491,6 +520,13 @@ def shift_excel_dates_inplace(
                         cell.value = None
                     continue
 
+                # only check the dates we're shifting
+                # ... other branches handle non-dates and dates we're not shifting
+                if parsed <= sanity_date_earliest:
+                    raise DateTooFarBack(parsed)
+                if parsed >= sanity_date_latest:
+                    raise DateTooFarAhead(parsed)
+
                 if shift_days is None:
                     continue
 

tests/data/.gitignore

@@ -0,0 +1,2 @@
+!*.csv
+!*.xlsx

tests/test_date_sanity.py

@@ -0,0 +1,79 @@
+from pathlib import Path
+
+import pytest
+
+from nuh_helper import shift_excel_dates_inplace
+from nuh_helper.date_shift import (
+    DateTooFarAhead,
+    DateTooFarBack,
+)
+
+
+def test_too_far_ahead(tmp_path: Path) -> None:
+
+    source_file = Path(__file__).parent / "data/date_sanity/too_far_ahead.xlsx"
+    output_path = tmp_path / "target.xlsx"
+    linking_table_old = tmp_path / "linking_table_old.csv"
+    linking_table_out = tmp_path / "linking_table_out.csv"
+
+    sheet_configs = {
+        "tofarr": {
+            "patient_id_col": "patient",
+            "date_columns": [
+                "dobirth",
+            ],
+            "header_row": 0,
+            "skip_rows_after_header": [],
+        },
+    }
+    with pytest.raises(DateTooFarAhead) as info:
+        shift_excel_dates_inplace(
+            input_file=str(source_file),
+            output_file=str(output_path),
+            patient_sheet="tofarr",
+            patient_id_col="patient",
+            sheet_configs=sheet_configs,
+            min_shift_days=-20,
+            max_shift_days=-1,
+            seed=14333,
+            linking_table_path=str(linking_table_old),
+            linking_table_output=str(linking_table_out),
+        )
+
+    assert (
+        info.value._message == "the date 2072-12-10 00:00:00 is too far in the future"
+    )
+
+
+def test_too_far_back(tmp_path: Path) -> None:
+
+    source_file = Path(__file__).parent / "data/date_sanity/too_far_back.xlsx"
+    output_path = tmp_path / "target.xlsx"
+    linking_table_old = tmp_path / "linking_table_old.csv"
+    linking_table_out = tmp_path / "linking_table_out.csv"
+
+    sheet_configs = {
+        "Sheet1": {
+            "patient_id_col": "patient",
+            "date_columns": [
+                "dob",
+            ],
+            "header_row": 0,
+            "skip_rows_after_header": [],
+        },
+    }
+    with pytest.raises(DateTooFarBack) as info:
+        shift_excel_dates_inplace(
+            input_file=str(source_file),
+            output_file=str(output_path),
+            patient_sheet="Sheet1",
+            patient_id_col="patient",
+            sheet_configs=sheet_configs,
+            min_shift_days=-20,
+            max_shift_days=-1,
+            seed=14333,
+            linking_table_path=str(linking_table_old),
+            linking_table_output=str(linking_table_out),
+        )
+
+    assert info.value._message == "the date 1007-02-05 00:00:00 is too far in the past"