Merge pull request #22 from longieirl/fix/rfc-17-inject-row-classifier

feat: inject shared RowClassifier chain (RFC #17)

Author: longieirl

packages/parser-core/src/bankstatements_core/extraction/boundary_detector.py

@@ -9,7 +9,10 @@
 import logging
 from dataclasses import dataclass
 
-from bankstatements_core.extraction.row_classifiers import create_row_classifier_chain
+from bankstatements_core.extraction.row_classifiers import (
+    RowClassifier,
+    create_row_classifier_chain,
+)
 
 logger = logging.getLogger(__name__)
 
@@ -49,6 +52,7 @@ def __init__(
         min_section_gap: int = 50,
         structure_breakdown_threshold: int = 8,
         dynamic_boundary_threshold: int = 15,
+        row_classifier: RowClassifier | None = None,
     ):
         """
         Initialize boundary detector.
@@ -60,11 +64,16 @@ def __init__(
             min_section_gap: Minimum gap in pixels to consider a section boundary
             structure_breakdown_threshold: Number of empty columns to consider structure broken
             dynamic_boundary_threshold: Consecutive non-transaction rows before ending extraction
+            row_classifier: Optional RowClassifier chain; creates default if not provided
         """
         self.columns = columns
         self.fallback_bottom_y = fallback_bottom_y
         self.table_top_y = table_top_y
-        self._row_classifier = create_row_classifier_chain()
+        self._row_classifier = (
+            row_classifier
+            if row_classifier is not None
+            else create_row_classifier_chain()
+        )
 
         # Configuration parameters
         self.min_gap_threshold = min_section_gap

packages/parser-core/src/bankstatements_core/extraction/extraction_facade.py

@@ -14,6 +14,7 @@
 from bankstatements_core.extraction.extraction_params import TABLE_BOTTOM_Y, TABLE_TOP_Y
 
 if TYPE_CHECKING:
+    from bankstatements_core.extraction.row_classifiers import RowClassifier
     from bankstatements_core.templates.template_model import BankTemplate
 
 
@@ -25,6 +26,7 @@ def detect_table_end_boundary_smart(
     min_section_gap: int = 50,
     structure_breakdown_threshold: int = 8,
     dynamic_boundary_threshold: int = 15,
+    row_classifier: "RowClassifier | None" = None,
 ) -> int:
     """
     Detect table end intelligently (facade).
@@ -39,6 +41,7 @@ def detect_table_end_boundary_smart(
         min_section_gap: Minimum gap in pixels to consider a section boundary
         structure_breakdown_threshold: Number of empty columns to consider structure broken
         dynamic_boundary_threshold: Consecutive non-transaction rows before ending extraction
+        row_classifier: Optional RowClassifier chain; creates default if not provided
 
     Returns:
         Detected bottom Y coordinate
@@ -52,6 +55,7 @@ def detect_table_end_boundary_smart(
         min_section_gap=min_section_gap,
         structure_breakdown_threshold=structure_breakdown_threshold,
         dynamic_boundary_threshold=dynamic_boundary_threshold,
+        row_classifier=row_classifier,
     )
 
     return detector.detect_boundary(words)

packages/parser-core/src/bankstatements_core/extraction/pdf_extractor.py

@@ -248,7 +248,11 @@ def _determine_boundaries_and_extract(
             )
 
             dynamic_bottom_y = detect_table_end_boundary_smart(
-                all_words, table_top_y, self.columns, table_bottom_y
+                all_words,
+                table_top_y,
+                self.columns,
+                table_bottom_y,
+                row_classifier=self._row_classifier,
             )
 
             # Safety check: Cap dynamic boundary at static boundary to prevent over-extraction

packages/parser-core/tests/extraction/test_boundary_detector.py

@@ -379,3 +379,21 @@ def test_configuration_via_constructor(self):
         assert detector.min_gap_threshold == 100
         assert detector.structure_breakdown_threshold == 10
         assert detector.consecutive_threshold == 20
+
+    def test_injected_classifier_is_used(self):
+        from unittest.mock import Mock
+
+        from bankstatements_core.extraction.row_classifiers import RowClassifier
+
+        mock_chain = Mock(spec=RowClassifier)
+        mock_chain.classify.return_value = "metadata"
+        detector = TableBoundaryDetector(
+            columns=TEST_COLUMNS, row_classifier=mock_chain
+        )
+        words = [{"text": "Footer", "x0": 60, "top": 350}]
+        detector.detect_boundary(words)
+        mock_chain.classify.assert_called()
+
+    def test_default_classifier_created_when_not_injected(self):
+        detector = TableBoundaryDetector(columns=TEST_COLUMNS)
+        assert detector._row_classifier is not None