Harden release boundaries and fix test contracts

Author: YapengWei

.github/workflows/python-app.yml

@@ -501,6 +501,7 @@ jobs:
           #    所以必须在 shell 命令内临时设置，delocate 才能找到 libomp.dylib。
           # 2. 添加 --require-archs {delocate_archs} 确保打包的 dylib 架构
           #    与 .so 文件一致（防止 arm64/x86_64 混淆导致用户端 crash）。
+          # 3. wheel repair 失败时必须让 CI 失败，不能发布未打包 libomp 的 macOS wheel。
           CIBW_REPAIR_WHEEL_COMMAND_MACOS: |
             echo "=== Repairing macOS wheel ==="
             echo "Current architecture: $(uname -m)"
@@ -510,13 +511,8 @@ jobs:
             # 在 repair 命令内部设置 DYLD_LIBRARY_PATH，绕过 macOS SIP 的剥离
             DYLD_LIBRARY_PATH="${REPAIR_LIBRARY_PATH}:${DYLD_LIBRARY_PATH:-}" \
               delocate-wheel --require-archs {delocate_archs} -w {dest_dir} -v {wheel} || {
-              echo "[WARNING] delocate with --require-archs failed, retrying without arch check..."
-              DYLD_LIBRARY_PATH="${REPAIR_LIBRARY_PATH}:${DYLD_LIBRARY_PATH:-}" \
-                delocate-wheel -w {dest_dir} -v {wheel} || {
-                  echo "[WARNING] delocate failed completely, copying wheel as-is"
-                  mkdir -p {dest_dir}
-                  cp {wheel} {dest_dir}/
-                }
+                echo "[ERROR] delocate failed; refusing to publish an unrepaired macOS wheel."
+                exit 1
             }
 
             # 验证修复结果：确认 libomp 是否真的打包进去了
@@ -528,7 +524,8 @@ jobs:
               if delocate-listdeps --all "$REPAIRED_WHEEL" 2>&1 | grep -q "libomp"; then
                 echo "[OK] libomp successfully bundled into wheel"
               else
-                echo "[WARNING] libomp may not be bundled - check wheel manually"
+                echo "[ERROR] libomp is not bundled into the macOS wheel."
+                exit 1
               fi
             fi
 
@@ -544,9 +541,15 @@ jobs:
             python -m pip install --upgrade pip
             echo "Ensuring clean build environment..."
             
-            # 再次清理，确保没有残留的构建文件
-            find . -name "*.so" -delete 2>/dev/null || true
-            find . -name "*.c" -delete 2>/dev/null || true
+            # 再次清理，确保没有残留的项目构建文件；不要全仓库扫描依赖/实验目录。
+            find ./sequenzo -name "*.so" -type f -delete 2>/dev/null || true
+            find . -maxdepth 1 -name "*.so" -type f -delete 2>/dev/null || true
+            find ./sequenzo -path "*/utils/*.c" -type f -print0 | while IFS= read -r -d '' cfile; do
+              pyxfile="${cfile%.c}.pyx"
+              if [ -f "$pyxfile" ]; then
+                rm -f "$cfile"
+              fi
+            done
             rm -rf build/ dist/ *.egg-info
             
             echo "Clean build environment ready"
@@ -567,11 +570,10 @@ jobs:
             echo "=== Attempting wheel repair ==="
             echo "Wheel to repair: {wheel}"
             
-            # 尝试修复，如果失败则跳过修复
+            # wheel repair 失败时必须让 CI 失败，不能发布未修复 wheel。
             auditwheel repair -w {dest_dir} {wheel} || {
-              echo "[WARNING] auditwheel repair failed, copying wheel as-is"
-              mkdir -p {dest_dir}
-              cp {wheel} {dest_dir}/
+              echo "[ERROR] auditwheel repair failed; refusing to publish an unrepaired Linux wheel."
+              exit 1
             }
             
             # 检查修复结果
@@ -1289,4 +1291,4 @@ jobs:
         uses: actions/upload-artifact@v4
         with:
           name: wheels-${{ matrix.os }}-${{ matrix.python-version }}
-          path: dist/
+          path: dist/

MANIFEST.in

@@ -13,6 +13,9 @@ recursive-include sequenzo/dissimilarity_measures/src *.hpp
 recursive-include sequenzo/clustering/src *.cpp
 recursive-include sequenzo/clustering/src *.h
 recursive-include sequenzo/clustering/src *.hpp
+recursive-include sequenzo/clustering/fuzzy_clustering/src *.cpp
+recursive-include sequenzo/clustering/fuzzy_clustering/src *.h
+recursive-include sequenzo/clustering/fuzzy_clustering/src *.hpp
 
 # Include all header files and external libraries
 recursive-include sequenzo/dissimilarity_measures/src/xsimd/include *.hpp

pyproject.toml

@@ -80,7 +80,9 @@ include-package-data = true
 ]
 "sequenzo.clustering" = [
   "src/**/*.cpp",
-  "src/**/*.h"
+  "src/**/*.h",
+  "fuzzy_clustering/src/**/*.cpp",
+  "fuzzy_clustering/src/**/*.h"
 ]
 "sequenzo.clustering.sequenzo_fastcluster" = [
   "src/**/*.cpp",
@@ -102,6 +104,9 @@ include = ["sequenzo*"]
 include = ["pybind11"]
 
 [tool.pytest.ini_options]
+testpaths = [
+    "tests",
+]
 norecursedirs = [
     ".git",
     ".venv",

sequenzo/decomposition/sa_kob.py

@@ -1,5 +1,5 @@
 """
-@Author  : Yuqi Liang 梁彧祺
+@Author  : Yuqi Liang 梁彧祺; Yapeng Wei 卫亚鹏
 @File    : sa_kob.py
 @Time    : 2026-04-20 16:30
 @Desc    :
@@ -343,10 +343,12 @@ def detect_cluster_coefficient_owners(
                 override_key = category_id
 
         if override_key is not None:
-            owner = _validate_coefficient_owner(
-                int(owner_overrides[override_key]),
-                context="[detect_cluster_coefficient_owners] owner_overrides",
-            )
+            owner = int(owner_overrides[override_key])
+            if owner not in (-1, 0, 1):
+                raise ValueError(
+                    "[detect_cluster_coefficient_owners] owner_overrides must map "
+                    "cluster labels or category ids to -1, 0, or 1."
+                )
             classification = "override"
         elif gap_percent > majority_gap_threshold:
             owner = 0 if share0 > share1 else 1

sequenzo/seqhmm/formulas.py

@@ -264,12 +264,13 @@ def create_model_matrix_time_constant(
                 f"Variable '{term}' not found in data columns: {list(data.columns)}"
             )
 
-        covar_values = data[term].values
+        series = data[term]
+        covar_values = series.values
 
         # Check if this is a categorical variable
-        if pd.api.types.is_categorical_dtype(data[term]) or \
-           pd.api.types.is_object_dtype(data[term]) or \
-           (data[term].dtype == 'object'):
+        if isinstance(series.dtype, pd.CategoricalDtype) or \
+           pd.api.types.is_object_dtype(series) or \
+           pd.api.types.is_string_dtype(series):
             # Categorical variable: create dummy variables
             # Use pandas get_dummies to create dummies, drop first level as reference
             dummies = pd.get_dummies(data[[term]], prefix=term, drop_first=True)

tests/compare_differences/test_sa_kob_decomposition.py

@@ -309,7 +309,7 @@ def test_invalid_cluster_coefficient_reference_raises():
             cluster_labels=cluster_labels,
             k=4,
             categories=[1, 2, 3, 4],
-            cluster_coefficient_reference="pooled",
+            cluster_coefficient_reference="invalid",
         )
 
 

tests/seqHMM/advanced/test_seqhmm_advanced.py

@@ -331,9 +331,9 @@ def test_create_matrix_time_constant_categorical(self):
             "color": ["red", "blue", "green", "red", "blue", "green"],
         })
         X = create_model_matrix_time_constant("~ color", data, n_seq)
-        # color has 3 levels, drop_first=True -> 2 dummy columns
-        # (no separate intercept column added by this function)
-        assert X.shape == (n_seq, 2)  # 2 dummies
+        # intercept + color's 3 levels with drop_first=True -> 3 columns
+        assert X.shape == (n_seq, 3)
+        assert np.allclose(X[:, 0], 1.0)
 
 
 # ============================================================================

tests/test_build_release_boundaries.py

@@ -0,0 +1,57 @@
+from pathlib import Path
+
+try:
+    import tomllib
+except ModuleNotFoundError:  # Python 3.9/3.10
+    import tomli as tomllib
+
+
+def test_setup_does_not_clone_unpinned_xsimd_from_network():
+    setup_py = Path(__file__).resolve().parents[1] / "setup.py"
+    source = setup_py.read_text()
+
+    assert "git\", \"clone" not in source
+    assert "https://github.com/xtensor-stack/xsimd.git" not in source
+    assert "include/xsimd/xsimd.hpp" in source
+
+
+def test_macos_wheel_repair_fails_instead_of_copying_unrepaired_wheel():
+    workflow = Path(__file__).resolve().parents[1] / ".github" / "workflows" / "python-app.yml"
+    source = workflow.read_text()
+
+    assert "copying wheel as-is" not in source
+    assert "cp {wheel} {dest_dir}/" not in source
+    assert 'find . -name "*.c" -delete' not in source
+    assert 'find . -name "*.so" -delete' not in source
+    assert "delocate-wheel --require-archs {delocate_archs}" in source
+    assert "exit 1" in source
+
+
+def test_setup_builds_fanny_cpp_binding_without_fast_math():
+    setup_py = Path(__file__).resolve().parents[1] / "setup.py"
+    source = setup_py.read_text()
+
+    assert "'sequenzo/clustering/fuzzy_clustering/src/'" in source
+    assert "'sequenzo/clustering/fuzzy_clustering/src/fanny.cpp'" in source
+    assert "or 'fanny.cpp' in src" in source
+
+
+def test_sdist_includes_fanny_cpp_sources_required_by_setup():
+    repo_root = Path(__file__).resolve().parents[1]
+    manifest = (repo_root / "MANIFEST.in").read_text()
+    config = tomllib.loads((repo_root / "pyproject.toml").read_text())
+
+    assert "recursive-include sequenzo/clustering/fuzzy_clustering/src *.cpp" in manifest
+    assert "recursive-include sequenzo/clustering/fuzzy_clustering/src *.h" in manifest
+
+    package_data = config["tool"]["setuptools"]["package-data"]["sequenzo.clustering"]
+    assert "fuzzy_clustering/src/**/*.cpp" in package_data
+    assert "fuzzy_clustering/src/**/*.h" in package_data
+
+
+def test_pytest_collects_only_project_tests_by_default():
+    pyproject = Path(__file__).resolve().parents[1] / "pyproject.toml"
+    config = tomllib.loads(pyproject.read_text())
+
+    pytest_config = config["tool"]["pytest"]["ini_options"]
+    assert pytest_config["testpaths"] == ["tests"]