Use left-stitched instead of patched columns for target clusters (#139)

* Update solver.py

* Update test_solver.py to match changes

* ran black, isort, etc.

* Update .gitignore

* Delete my_debugger.py

* better variable name

.gitignore

@@ -25,4 +25,5 @@ __pycache__/
 *.log
 
 # Ignore vscode workspace virtual environment
-/venv/
+/venv/
+/tests/my_debugger.py

syndiffix/clustering/solver.py

@@ -249,10 +249,10 @@ def new_cluster() -> MutableCluster:
     derived_clusters = [(StitchOwner.SHARED, [main_column], list(cluster.columns)) for cluster in clusters[1:]]
 
     ml_columns = [main_column] + main_features
-    patch_columns: list[DerivedCluster] = (
+    non_feature_columns: list[DerivedCluster] = (
         []
         if drop_non_features
-        else [(StitchOwner.SHARED, [], [ColumnId(c)]) for c in range(num_columns) if c not in ml_columns]
+        else [(StitchOwner.LEFT, [main_column], [ColumnId(c)]) for c in range(num_columns) if c not in ml_columns]
     )
 
-    return Clusters(initial_cluster=initial_cluster, derived_clusters=derived_clusters + patch_columns)
+    return Clusters(initial_cluster=initial_cluster, derived_clusters=derived_clusters + non_feature_columns)

syndiffix/tree.py

@@ -207,9 +207,11 @@ def _create_child_leaf(self, child_index: int, initial_row: RowId) -> Leaf:
 
         # Set child's subnodes to the matching-interval children of the parent's subnodes.
         subnodes = tuple(
-            subnode.children.get(Branch._remove_dimension_from_index(dim_index, child_index))
-            if isinstance(subnode, Branch)
-            else None
+            (
+                subnode.children.get(Branch._remove_dimension_from_index(dim_index, child_index))
+                if isinstance(subnode, Branch)
+                else None
+            )
             for dim_index, subnode in enumerate(self.subnodes)
         )
 

tests/clustering/test_solver.py

@@ -36,8 +36,8 @@ def test_solve_with_features() -> None:
     assert clusters_patches.derived_clusters == [
         (StitchOwner.SHARED, [Col(0)], [Col(3), Col(4)]),
         (StitchOwner.SHARED, [Col(0)], [Col(5)]),
-        (StitchOwner.SHARED, [], [Col(6)]),
-        (StitchOwner.SHARED, [], [Col(7)]),
+        (StitchOwner.LEFT, [Col(0)], [Col(6)]),
+        (StitchOwner.LEFT, [Col(0)], [Col(7)]),
     ]
 
     clusters_no_patches = solver.solve_with_features(