fix: include overlapping fragments in core_ids for diff context

When a new function is added at the end of a file, the hunk may start
on an empty line before the function definition. Previously, this caused
the function fragment to be excluded because it didn't fully cover the
hunk. Now we check for overlapping fragments when full coverage fails.

Added regression test: test_sel_core_004_new_function_at_end_of_file

Author: nikolay-e

demo

@@ -0,0 +1 @@
+Subproject commit 4046d4c78425e02b931ecb8cc38d643bbb7c3b7a

src/treemapper/diffctx/__init__.py

@@ -78,8 +78,8 @@ def build_diff_context(
     if not is_git_repo(root_dir):
         raise GitError(f"'{root_dir}' is not a git repository")
 
-    if not (0.0 <= alpha < 1.0):
-        raise ValueError(f"alpha must be in [0, 1), got {alpha}")
+    if not (0.0 < alpha < 1.0):
+        raise ValueError(f"alpha must be in (0, 1), got {alpha}")
     if tau < 0.0:
         raise ValueError(f"tau must be >= 0, got {tau}")
 
@@ -150,9 +150,14 @@ def build_diff_context(
             best = min(covering, key=lambda f: f.line_count)
             core_ids.add(best.id)
         else:
-            # Fallback: use enclosing fragment
-            enc = enclosing_fragment(frags, h_start)
-            if enc is not None:
+            # Check for fragments that OVERLAP with the hunk (partial coverage)
+            overlapping = [f for f in frags if f.start_line <= h_end and f.end_line >= h_start]
+            if overlapping:
+                # Add all overlapping fragments as core
+                for f in overlapping:
+                    core_ids.add(f.id)
+            elif (enc := enclosing_fragment(frags, h_start)) is not None:
+                # Fallback: use enclosing fragment
                 core_ids.add(enc.id)
             else:
                 # For hunks in gaps between fragments, find nearest adjacent fragments

src/treemapper/diffctx/graph.py

@@ -622,13 +622,7 @@ def _build_config_code_edges(fragments: list[Fragment]) -> dict[tuple[FragmentId
     if not config_frags or not code_frags:
         return edges
 
-    # Build inverted index: key -> list of code fragments containing the key
     key_to_code_frags: dict[str, list[FragmentId]] = defaultdict(list)
-    for code_frag in code_frags:
-        content_lower = code_frag.content.lower()
-        for code_frag_inner in code_frags:
-            if code_frag_inner.id == code_frag.id:
-                continue
 
     # Extract keys from config files
     config_keys_by_frag: dict[FragmentId, set[str]] = {}

tests/test_diff_edge_cases.py

@@ -638,6 +638,21 @@ def test_invalid_alpha_negative(self, diff_project):
                 alpha=-0.1,
             )
 
+    def test_invalid_alpha_zero(self, diff_project):
+        """alpha=0 should raise ValueError (must be in (0, 1))"""
+        diff_project.add_file("test.py", "x = 1")
+        diff_project.commit("Initial")
+        diff_project.add_file("test.py", "x = 2")
+        diff_project.commit("Modify")
+
+        with pytest.raises(ValueError):
+            build_diff_context(
+                root_dir=diff_project.repo,
+                diff_range="HEAD~1..HEAD",
+                budget_tokens=1000,
+                alpha=0.0,
+            )
+
     def test_invalid_alpha_greater_than_one(self, diff_project):
         """alpha >= 1 should raise ValueError"""
         diff_project.add_file("test.py", "x = 1")
@@ -669,18 +684,17 @@ def test_invalid_tau_negative(self, diff_project):
             )
 
     def test_valid_alpha_boundaries(self, diff_project):
-        """alpha=0.0 and alpha=0.99 should work"""
+        """alpha=0.01 and alpha=0.99 should work (0 and 1 are excluded)"""
         diff_project.add_file("test.py", "x = 1")
         diff_project.commit("Initial")
         diff_project.add_file("test.py", "x = 2")
         diff_project.commit("Modify")
 
-        # Should not raise
         tree1 = build_diff_context(
             root_dir=diff_project.repo,
             diff_range="HEAD~1..HEAD",
             budget_tokens=1000,
-            alpha=0.0,
+            alpha=0.01,
         )
         assert tree1 is not None
 

tests/test_diff_selection_regression.py

@@ -158,6 +158,61 @@ def func2():
         assert new_func is not None
 
 
+    def test_sel_core_004_new_function_at_end_of_file(self, diff_project):
+        """Regression test: new function added at end of file must be included.
+
+        The hunk may start on empty line before function definition,
+        but the function fragment must still be detected as core.
+        """
+        diff_project.add_file(
+            "calculator.py",
+            """\
+def add(a, b):
+    return a + b
+
+def sub(a, b):
+    return a - b
+
+def mul(a, b):
+    return a * b
+""",
+        )
+        diff_project.commit("Initial calculator")
+
+        diff_project.add_file(
+            "calculator.py",
+            """\
+def add(a, b):
+    return a + b
+
+def sub(a, b):
+    return a - b
+
+def mul(a, b):
+    return a * b
+
+def div(a, b):
+    if b == 0:
+        raise ValueError("division by zero")
+    return a / b
+""",
+        )
+        diff_project.commit("Add div function")
+
+        tree = build_diff_context(
+            root_dir=diff_project.repo,
+            diff_range="HEAD~1..HEAD",
+            budget_tokens=10000,
+        )
+
+        fragments = _extract_fragments_from_tree(tree)
+        div_func = next((f for f in fragments if "def div" in f.get("content", "")), None)
+        assert div_func is not None, "div function must be included in diff context"
+
+        mul_func = next((f for f in fragments if "def mul" in f.get("content", "")), None)
+        assert mul_func is None, "mul function should NOT be included (not changed)"
+
+
 class TestBudgetBehavior:
     def test_sel_budget_001_respects_token_limit(self):
         fragments = [