fix: background steps, rglob scan, __init__.py, lowercased function names (v0.4.0) (#17)

Four bug fixes discovered during pytest-beehave e2e integration:

1. ScenarioInfo.steps now includes background steps (was only scenario steps)
2. check_all uses rglob() to find features in subdirectories
3. generate_stubs creates __init__.py in generated directories
4. Function names are lowercased, matching path slug behavior

Adds 9 new tests. Updates spec and README to document lowercasing.

Author: nullhack

README.md

@@ -192,7 +192,7 @@ Warnings exit 0. Errors exit 1. Stubs (bodies with only `pass` or `...`) skip bo
 
 ## How it maps
 
-- **Scenario title → function name:** `Honey Production From Nectar` → `test_honey_production_from_nectar`. Globally unique across all features.
+- **Scenario title → function name:** `Honey Production From Nectar` → `test_honey_production_from_nectar`. Lowercased. Globally unique across all features.
 - **Rule → test file:** Top-level scenarios go to `default_test.py`. Scenarios inside a Rule go to `<rule>_test.py`.
 - **Feature title → directory:** `Hive Activity` → `tests/features/hive_activity/`.
 - **Strategy inference:** Examples table column values are typed — all integers → `st.integers()`, all floats → `st.floats()`, all booleans → `st.booleans()`, else → `st.text()`.

beehave/check.py

@@ -230,7 +230,7 @@ def check_all(config: Config) -> list[Violation]:
     feature_paths: dict[str, Path] = {}
 
     seen_fn: dict[str, str] = {}
-    for feature_file in sorted(features_dir.glob("*.feature")):
+    for feature_file in sorted(features_dir.rglob("*.feature")):
         try:
             scenarios = parse_feature(feature_file, config, seen_function_names=seen_fn)
         except GherkinError as e:

beehave/generate.py

@@ -173,6 +173,9 @@ def _write_file(
     config: Config,
 ) -> None:
     test_file.parent.mkdir(parents=True, exist_ok=True)
+    init_file = test_file.parent / "__init__.py"
+    if not init_file.exists():
+        init_file.touch()
 
     existing_functions: set[str] = set()
     existing_strategies: set[str] = set()

beehave/gherkin.py

@@ -43,7 +43,7 @@ def _derive_path_slug(title: str) -> str:
 
 def _derive_function_name(title: str) -> str:
     trimmed = title.strip()
-    collapsed = re.sub(r"\s+", "_", trimmed)
+    collapsed = re.sub(r"\s+", "_", trimmed).lower()
     name = f"test_{collapsed}"
     if not name.isidentifier():
         raise GherkinError(
@@ -192,7 +192,7 @@ def _build_scenario(
     return ScenarioInfo(
         title=title,
         function_name=function_name,
-        steps=tuple(steps),
+        steps=tuple(merged),
         placeholders=_collect_placeholders(merged),
         literals=_collect_literals(
             steps, feature_bg + rule_bg, check_numeric, check_string

docs/spec/v3/beehave_v3_spec.md

@@ -66,12 +66,14 @@ Every `Scenario` and `Scenario Outline` maps to exactly one test function. The f
 1. **Trim** leading and trailing whitespace.
 2. **Collapse** consecutive internal spaces to a single space.
 3. **Replace** each space with an underscore (`_`).
-4. **Prepend** `test_`.
-5. **Validate** that the result is a valid Python identifier (`str.isidentifier()` returns `True`). If not, raise a parse error.
+4. **Lowercase** the result.
+5. **Prepend** `test_`.
+6. **Validate** that the result is a valid Python identifier (`str.isidentifier()` returns `True`). If not, raise a parse error.
 
 ```
 Scenario: deposit increases balance       →  test_deposit_increases_balance
 Scenario:  extra   spaces   here          →  test_extra_spaces_here
+Scenario: Add Single Item                 →  test_add_single_item
 ```
 
 No `@scenario` decorator. No `@id` tags. No cache file. At collection time, beehave re-parses all `.feature` files and AST-parses all test files, then joins on function name.
@@ -80,7 +82,7 @@ No `@scenario` decorator. No `@id` tags. No cache file. At collection time, beeh
 
 - **Characters:** Unicode letters, digits, and spaces only. Applies to Scenario, Scenario Outline, Feature, and Rule titles equally. Special characters would break generated file paths or Python identifiers.
 - **Non-empty:** The title must be non-empty after trimming.
-- **Scenario titles:** Globally unique across all features. Two scenario titles that collapse to the same function name produce a parse error.
+- **Scenario titles:** Globally unique across all features. Two scenario titles that collapse to the same function name (case-insensitive) produce a parse error.
 - **Rule titles:** Unique within their parent Feature. Rule titles are used internally as keys for background lookup and in error messages — duplicate rule titles within a Feature produce a parse error. Per the Gherkin specification, rule names must be unique within their parent feature.
 - **Feature titles:** Globally unique across all features. Feature titles determine the generated folder structure — `Feature: Bank` generates `tests/features/bank/`. Duplicate or special-character feature titles would create path collisions or invalid directories.
 

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "beehave"
-version = "0.3.1"
+version = "0.4.0"
 description = "A thin layer on Hypothesis for Gherkin-style BDD testing with vocabulary enforcement"
 readme = "README.md"
 requires-python = ">=3.14"

tests/test_check.py

@@ -377,3 +377,30 @@ def test_unmapped():
         violations = check_all(config)
         types = [v.error_type for v in violations]
         assert "unmapped-test" in types
+
+    def test_subdirectory_features_found(
+        self, tmp_project: Path, config: Config
+    ) -> None:
+        write_feature(
+            tmp_project,
+            "cart/shopping",
+            """\
+            Feature: Shopping
+              Scenario: add item
+                Given stuff
+            """,
+        )
+        write_feature(
+            tmp_project,
+            "smoke",
+            """\
+            Feature: Smoke
+              Scenario: everything is fine
+                Given stuff
+            """,
+        )
+
+        generate_stubs("cart/shopping", config)
+        generate_stubs("smoke", config)
+        violations = check_all(config)
+        assert violations == []

tests/test_generate.py

@@ -252,3 +252,36 @@ def test_scenario_outline_with_examples(
     def test_nonexistent_feature_exits(self, tmp_project: Path, config: Config) -> None:
         with pytest.raises(SystemExit):
             generate_stubs("nonexistent", config)
+
+    def test_creates_init_py(self, tmp_project: Path, config: Config) -> None:
+        write_feature(
+            tmp_project,
+            "initcheck",
+            """\
+            Feature: Initcheck
+              Scenario: hello
+                Given stuff
+            """,
+        )
+        generate_stubs("initcheck", config)
+        init_file = tmp_project / "tests" / "features" / "initcheck" / "__init__.py"
+        assert init_file.exists()
+
+    def test_does_not_overwrite_existing_init_py(
+        self, tmp_project: Path, config: Config
+    ) -> None:
+        write_feature(
+            tmp_project,
+            "initexist",
+            """\
+            Feature: Initexist
+              Scenario: hello
+                Given stuff
+            """,
+        )
+        test_dir = tmp_project / "tests" / "features" / "initexist"
+        test_dir.mkdir(parents=True, exist_ok=True)
+        init_file = test_dir / "__init__.py"
+        init_file.write_text("# custom init\n", encoding="utf-8")
+        generate_stubs("initexist", config)
+        assert init_file.read_text() == "# custom init\n"

tests/test_gherkin.py

@@ -59,12 +59,22 @@ def test_invalid_identifier_raises(self) -> None:
         with pytest.raises(GherkinError, match="not a valid Python identifier"):
             _derive_function_name("hello-world")
 
+    def test_uppercase_lowered(self) -> None:
+        assert _derive_function_name("Add Single Item") == "test_add_single_item"
+
+    def test_mixed_case_lowered(self) -> None:
+        assert _derive_function_name("EvErYtHiNg") == "test_everything"
+
+    def test_case_insensitive_collision(self) -> None:
+        assert _derive_function_name("Test") == _derive_function_name("tEsT")
+
     @given(st.from_regex(r"[a-zA-Z_][a-zA-Z0-9_]*", fullmatch=True))
     @settings(max_examples=50)
     def test_single_word_always_valid(self, word: str) -> None:
         result = _derive_function_name(word)
         assert result.startswith("test_")
         assert result.isidentifier()
+        assert result == result.lower()
 
 
 class TestDeriveFeaturePath:
@@ -392,6 +402,57 @@ def test_background_literals_configurable(self, tmp_project: Path) -> None:
         assert all(not isinstance(lit.value, int) for lit in si.literals)
         assert all(not isinstance(lit.value, str) for lit in si.literals)
 
+    def test_background_steps_included_in_scenario_info(
+        self, tmp_project: Path, config: Config
+    ) -> None:
+        fp = write_feature(
+            tmp_project,
+            "bgsteps",
+            """\
+            Feature: BG Steps
+              Background:
+                Given a user exists
+                And the user has an empty cart
+
+              Scenario: add item
+                When the user adds "Widget" to the cart
+                Then the cart contains 1 item
+            """,
+        )
+        result = parse_feature(fp, config)
+        si = result["test_add_item"]
+        assert len(si.steps) == 4
+        assert si.steps[0].text == "a user exists"
+        assert si.steps[1].text == "the user has an empty cart"
+        assert si.steps[2].text == 'the user adds "Widget" to the cart'
+        assert si.steps[3].text == "the cart contains 1 item"
+
+    def test_rule_background_steps_included_in_scenario_info(
+        self, tmp_project: Path, config: Config
+    ) -> None:
+        fp = write_feature(
+            tmp_project,
+            "rulebgsteps",
+            """\
+            Feature: Rule BG Steps
+              Background:
+                Given feature bg step
+
+              Rule: Sub
+                Background:
+                  Given rule bg step
+
+                Scenario: combined
+                  Given scenario step
+            """,
+        )
+        result = parse_feature(fp, config)
+        si = result["test_combined"]
+        assert len(si.steps) == 3
+        assert si.steps[0].text == "feature bg step"
+        assert si.steps[1].text == "rule bg step"
+        assert si.steps[2].text == "scenario step"
+
     def test_rule_background_composes(self, tmp_project: Path, config: Config) -> None:
         fp = write_feature(
             tmp_project,