Property-based fuzz test

.github/workflows/fuzz.yml

@@ -0,0 +1,31 @@
+name: Fuzz
+
+on: [push, pull_request]
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: [3.6, 3.7, 3.8]
+
+    steps:
+      - uses: actions/checkout@v2
+
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v2
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install --upgrade coverage
+          python -m pip install --upgrade hypothesmith
+          python -m pip install -e ".[d]"
+
+      - name: Run fuzz tests
+        run: |
+          coverage run fuzz.py
+          coverage report

.gitignore

@@ -14,3 +14,4 @@ src/_black_version.py
 .eggs
 .dmypy.json
 *.swp
+.hypothesis/

README.md

@@ -684,3 +684,4 @@ Multiple contributions by:
 - Yazdan
 - [Yngve Høiseth](mailto:[email protected])
 - [Yurii Karabas](mailto:[email protected])
+- [Zac Hatfield-Dodds](mailto:[email protected])

fuzz.py

@@ -0,0 +1,59 @@
+"""Property-based tests for Black.
+
+By Zac Hatfield-Dodds, based on my Hypothesmith tool for source code
+generation.  You can run this file with `python`, `pytest`, or (soon)
+a coverage-guided fuzzer I'm working on.
+"""
+
+import hypothesmith
+from hypothesis import HealthCheck, given, settings, strategies as st
+
+import black
+
+
+# This test uses the Hypothesis and Hypothesmith libraries to generate random
+# syntatically-valid Python source code and run Black in odd modes.
+@settings(
+    max_examples=1000,  # roughly 1k tests/minute, or half that under coverage
+    derandomize=True,  # deterministic mode to avoid CI flakiness
+    deadline=None,  # ignore Hypothesis' health checks; we already know that
+    suppress_health_check=HealthCheck.all(),  # this is slow and filter-heavy.
+)
+@given(
+    # Note that while Hypothesmith might generate code unlike that written by
+    # humans, it's a general test that should pass for any *valid* source code.
+    # (so e.g. running it against code scraped of the internet might also help)
+    src_contents=hypothesmith.from_grammar() | hypothesmith.from_node(),
+    # Using randomly-varied modes helps us to exercise less common code paths.
+    mode=st.builds(
+        black.FileMode,
+        line_length=st.just(88) | st.integers(0, 200),
+        string_normalization=st.booleans(),
+        is_pyi=st.booleans(),
+    ),
+)
+def test_idempotent_any_syntatically_valid_python(
+    src_contents: str, mode: black.FileMode
+) -> None:
+    # Before starting, let's confirm that the input string is valid Python:
+    compile(src_contents, "<string>", "exec")  # else the bug is in hypothesmith
+
+    # Then format the code...
+    try:
+        dst_contents = black.format_str(src_contents, mode=mode)
+    except black.InvalidInput:
+        # This is a bug - if it's valid Python code, as above, black should be
+        # able to code with it.  See issues #970, #1012, #1358, and #1557.
+        # TODO: remove this try-except block when issues are resolved.
+        return
+
+    # And check that we got equivalent and stable output.
+    black.assert_equivalent(src_contents, dst_contents)
+    black.assert_stable(src_contents, dst_contents, mode=mode)
+
+    # Future test: check that pure-python and mypyc versions of black
+    # give identical output for identical input?
+
+
+if __name__ == "__main__":
+    test_idempotent_any_syntatically_valid_python()