Merge commit '9e45dadcf6d8dbab36f83d9df94a706c0b4f9207' into release-1.16

Author: JukkaL

.pre-commit-config.yaml

@@ -11,12 +11,12 @@ repos:
       - id: black
         exclude: '^(test-data/)'
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.8.6
+    rev: v0.11.4
     hooks:
       - id: ruff
         args: [--exit-non-zero-on-fix]
   - repo: https://github.com/python-jsonschema/check-jsonschema
-    rev: 0.31.0
+    rev: 0.32.1
     hooks:
       - id: check-github-workflows
       - id: check-github-actions
@@ -43,7 +43,7 @@ repos:
           # but the integration only works if shellcheck is installed
           - "github.com/wasilibs/go-shellcheck/cmd/[email protected]"
   - repo: https://github.com/woodruffw/zizmor-pre-commit
-    rev: v1.0.1
+    rev: v1.5.2
     hooks:
       - id: zizmor
   - repo: local

CONTRIBUTING.md

@@ -76,10 +76,14 @@ python runtests.py self
 # or equivalently:
 python -m mypy --config-file mypy_self_check.ini -p mypy
 
-# Run a single test from the test suite
-pytest -n0 -k 'test_name'
+# Run a single test from the test suite (uses pytest substring expression matching)
+python runtests.py test_name
+# or equivalently:
+pytest -n0 -k test_name
 
 # Run all test cases in the "test-data/unit/check-dataclasses.test" file
+python runtests.py check-dataclasses.test
+# or equivalently:
 pytest mypy/test/testcheck.py::TypeCheckSuite::check-dataclasses.test
 
 # Run the formatters and linters

docs/source/command_line.rst

@@ -749,8 +749,19 @@ of the above sections.
 
 .. option:: --strict
 
-    This flag mode enables all optional error checking flags.  You can see the
-    list of flags enabled by strict mode in the full :option:`mypy --help` output.
+    This flag mode enables a defined subset of optional error-checking flags.
+    This subset primarily includes checks for inadvertent type unsoundness (i.e
+    strict will catch type errors as long as intentional methods like type ignore
+    or casting were not used.)
+
+    Note: the :option:`--warn-unreachable` flag
+    is not automatically enabled by the strict flag.
+
+    The strict flag does not take precedence over other strict-related flags.
+    Directly specifying a flag of alternate behavior will override the
+    behavior of strict, regardless of the order in which they are passed.
+    You can see the list of flags enabled by strict mode in the full
+    :option:`mypy --help` output.
 
     Note: the exact list of flags enabled by running :option:`--strict` may change
     over time.

docs/source/config_file.rst

@@ -432,7 +432,7 @@ Platform configuration
 
     Specifies the Python version used to parse and check the target
     program.  The string should be in the format ``MAJOR.MINOR`` --
-    for example ``2.7``.  The default is the version of the Python
+    for example ``3.9``.  The default is the version of the Python
     interpreter used to run mypy.
 
     This option may only be set in the global section (``[mypy]``).
@@ -1196,7 +1196,7 @@ of your repo (or append it to the end of an existing ``pyproject.toml`` file) an
     # mypy global options:
 
     [tool.mypy]
-    python_version = "2.7"
+    python_version = "3.9"
     warn_return_any = true
     warn_unused_configs = true
     exclude = [

docs/source/runtime_troubles.rst

@@ -8,10 +8,9 @@ version of Python considers legal code. This section describes these scenarios
 and explains how to get your code running again. Generally speaking, we have
 three tools at our disposal:
 
-* Use of ``from __future__ import annotations`` (:pep:`563`)
-  (this behaviour may eventually be made the default in a future Python version)
 * Use of string literal types or type comments
 * Use of ``typing.TYPE_CHECKING``
+* Use of ``from __future__ import annotations`` (:pep:`563`)
 
 We provide a description of these before moving onto discussion of specific
 problems you may encounter.

docs/source/type_narrowing.rst

@@ -8,6 +8,15 @@ techniques which are supported by mypy.
 
 Type narrowing is when you convince a type checker that a broader type is actually more specific, for instance, that an object of type ``Shape`` is actually of the narrower type ``Square``.
 
+The following type narrowing techniques are available:
+
+- :ref:`type-narrowing-expressions`
+- :ref:`casts`
+- :ref:`type-guards`
+- :ref:`typeis`
+
+
+.. _type-narrowing-expressions:
 
 Type narrowing expressions
 --------------------------
@@ -356,40 +365,6 @@ What happens here?
 
   The same will work with ``isinstance(x := a, float)`` as well.
 
-Limitations
------------
-
-Mypy's analysis is limited to individual symbols and it will not track
-relationships between symbols. For example, in the following code
-it's easy to deduce that if :code:`a` is None then :code:`b` must not be,
-therefore :code:`a or b` will always be an instance of :code:`C`,
-but Mypy will not be able to tell that:
-
-.. code-block:: python
-
-    class C:
-        pass
-
-    def f(a: C | None, b: C | None) -> C:
-        if a is not None or b is not None:
-            return a or b  # Incompatible return value type (got "C | None", expected "C")
-        return C()
-
-Tracking these sort of cross-variable conditions in a type checker would add significant complexity
-and performance overhead.
-
-You can use an ``assert`` to convince the type checker, override it with a :ref:`cast <casts>`
-or rewrite the function to be slightly more verbose:
-
-.. code-block:: python
-
-    def f(a: C | None, b: C | None) -> C:
-        if a is not None:
-            return a
-        elif b is not None:
-            return b
-        return C()
-
 
 .. _typeis:
 
@@ -555,3 +530,38 @@ You can use the assignment expression operator ``:=`` with ``TypeIs`` to create
             reveal_type(x)  # Revealed type is 'float'
             # x is narrowed to float in this block
             print(x + 1.0)
+
+
+Limitations
+-----------
+
+Mypy's analysis is limited to individual symbols and it will not track
+relationships between symbols. For example, in the following code
+it's easy to deduce that if :code:`a` is None then :code:`b` must not be,
+therefore :code:`a or b` will always be an instance of :code:`C`,
+but Mypy will not be able to tell that:
+
+.. code-block:: python
+
+    class C:
+        pass
+
+    def f(a: C | None, b: C | None) -> C:
+        if a is not None or b is not None:
+            return a or b  # Incompatible return value type (got "C | None", expected "C")
+        return C()
+
+Tracking these sort of cross-variable conditions in a type checker would add significant complexity
+and performance overhead.
+
+You can use an ``assert`` to convince the type checker, override it with a :ref:`cast <casts>`
+or rewrite the function to be slightly more verbose:
+
+.. code-block:: python
+
+    def f(a: C | None, b: C | None) -> C:
+        if a is not None:
+            return a
+        elif b is not None:
+            return b
+        return C()

misc/perf_compare.py

@@ -9,7 +9,7 @@
  * Create a temp clone of the mypy repo for each target commit to measure
  * Checkout a target commit in each of the clones
  * Compile mypyc in each of the clones *in parallel*
- * Create another temp clone of the mypy repo as the code to check
+ * Create another temp clone of the first provided revision (or, with -r, a foreign repo) as the code to check
  * Self check with each of the compiled mypys N times
  * Report the average runtimes and relative performance
  * Remove the temp clones
@@ -44,13 +44,15 @@ def build_mypy(target_dir: str) -> None:
     subprocess.run(cmd, env=env, check=True, cwd=target_dir)
 
 
-def clone(target_dir: str, commit: str | None) -> None:
-    heading(f"Cloning mypy to {target_dir}")
-    repo_dir = os.getcwd()
+def clone(target_dir: str, commit: str | None, repo_source: str | None = None) -> None:
+    source_name = repo_source or "mypy"
+    heading(f"Cloning {source_name} to {target_dir}")
+    if repo_source is None:
+        repo_source = os.getcwd()
     if os.path.isdir(target_dir):
         print(f"{target_dir} exists: deleting")
         shutil.rmtree(target_dir)
-    subprocess.run(["git", "clone", repo_dir, target_dir], check=True)
+    subprocess.run(["git", "clone", repo_source, target_dir], check=True)
     if commit:
         subprocess.run(["git", "checkout", commit], check=True, cwd=target_dir)
 
@@ -64,7 +66,7 @@ def edit_python_file(fnam: str) -> None:
 
 
 def run_benchmark(
-    compiled_dir: str, check_dir: str, *, incremental: bool, code: str | None
+    compiled_dir: str, check_dir: str, *, incremental: bool, code: str | None, foreign: bool | None
 ) -> float:
     cache_dir = os.path.join(compiled_dir, ".mypy_cache")
     if os.path.isdir(cache_dir) and not incremental:
@@ -76,6 +78,8 @@ def run_benchmark(
     cmd = [sys.executable, "-m", "mypy"]
     if code:
         cmd += ["-c", code]
+    elif foreign:
+        pass
     else:
         cmd += ["--config-file", os.path.join(abschk, "mypy_self_check.ini")]
         cmd += glob.glob(os.path.join(abschk, "mypy/*.py"))
@@ -86,18 +90,33 @@ def run_benchmark(
             edit_python_file(os.path.join(abschk, "mypy/test/testcheck.py"))
     t0 = time.time()
     # Ignore errors, since some commits being measured may generate additional errors.
-    subprocess.run(cmd, cwd=compiled_dir, env=env)
+    if foreign:
+        subprocess.run(cmd, cwd=check_dir, env=env)
+    else:
+        subprocess.run(cmd, cwd=compiled_dir, env=env)
     return time.time() - t0
 
 
 def main() -> None:
-    parser = argparse.ArgumentParser()
+    whole_program_time_0 = time.time()
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+        description=__doc__,
+        epilog="Remember: you usually want the first argument to this command to be 'master'.",
+    )
     parser.add_argument(
         "--incremental",
         default=False,
         action="store_true",
         help="measure incremental run (fully cached)",
     )
+    parser.add_argument(
+        "--dont-setup",
+        default=False,
+        action="store_true",
+        help="don't make the clones or compile mypy, just run the performance measurement benchmark "
+        + "(this will fail unless the clones already exist, such as from a previous run that was canceled before it deleted them)",
+    )
     parser.add_argument(
         "--num-runs",
         metavar="N",
@@ -112,42 +131,65 @@ def main() -> None:
         type=int,
         help="set maximum number of parallel builds (default=8)",
     )
+    parser.add_argument(
+        "-r",
+        metavar="FOREIGN_REPOSITORY",
+        default=None,
+        type=str,
+        help="measure time to typecheck the project at FOREIGN_REPOSITORY instead of mypy self-check; "
+        + "the provided value must be the URL or path of a git repo "
+        + "(note that this script will take no special steps to *install* the foreign repo, so you will probably get a lot of missing import errors)",
+    )
     parser.add_argument(
         "-c",
         metavar="CODE",
         default=None,
         type=str,
         help="measure time to type check Python code fragment instead of mypy self-check",
     )
-    parser.add_argument("commit", nargs="+", help="git revision to measure (e.g. branch name)")
+    parser.add_argument(
+        "commit",
+        nargs="+",
+        help="git revision(s), e.g. branch name or commit id, to measure the performance of",
+    )
     args = parser.parse_args()
     incremental: bool = args.incremental
+    dont_setup: bool = args.dont_setup
     commits = args.commit
     num_runs: int = args.num_runs + 1
     max_workers: int = args.j
     code: str | None = args.c
+    foreign_repo: str | None = args.r
 
     if not (os.path.isdir(".git") and os.path.isdir("mypyc")):
-        sys.exit("error: Run this the mypy repo root")
+        sys.exit("error: You must run this script from the mypy repo root")
 
     target_dirs = []
     for i, commit in enumerate(commits):
         target_dir = f"mypy.{i}.tmpdir"
         target_dirs.append(target_dir)
-        clone(target_dir, commit)
+        if not dont_setup:
+            clone(target_dir, commit)
 
-    self_check_dir = "mypy.self.tmpdir"
-    clone(self_check_dir, commits[0])
+    if foreign_repo:
+        check_dir = "mypy.foreign.tmpdir"
+        if not dont_setup:
+            clone(check_dir, None, foreign_repo)
+    else:
+        check_dir = "mypy.self.tmpdir"
+        if not dont_setup:
+            clone(check_dir, commits[0])
 
-    heading("Compiling mypy")
-    print("(This will take a while...)")
+    if not dont_setup:
+        heading("Compiling mypy")
+        print("(This will take a while...)")
 
-    with ThreadPoolExecutor(max_workers=max_workers) as executor:
-        futures = [executor.submit(build_mypy, target_dir) for target_dir in target_dirs]
-        for future in as_completed(futures):
-            future.result()
+        with ThreadPoolExecutor(max_workers=max_workers) as executor:
+            futures = [executor.submit(build_mypy, target_dir) for target_dir in target_dirs]
+            for future in as_completed(futures):
+                future.result()
 
-    print(f"Finished compiling mypy ({len(commits)} builds)")
+        print(f"Finished compiling mypy ({len(commits)} builds)")
 
     heading("Performing measurements")
 
@@ -160,7 +202,13 @@ def main() -> None:
         items = list(enumerate(commits))
         random.shuffle(items)
         for i, commit in items:
-            tt = run_benchmark(target_dirs[i], self_check_dir, incremental=incremental, code=code)
+            tt = run_benchmark(
+                target_dirs[i],
+                check_dir,
+                incremental=incremental,
+                code=code,
+                foreign=bool(foreign_repo),
+            )
             # Don't record the first warm-up run
             if n > 0:
                 print(f"{commit}: t={tt:.3f}s")
@@ -171,15 +219,28 @@ def main() -> None:
     first = -1.0
     for commit in commits:
         tt = statistics.mean(results[commit])
+        # pstdev (instead of stdev) is used here primarily to accommodate the case where num_runs=1
+        s = statistics.pstdev(results[commit]) if len(results[commit]) > 1 else 0
         if first < 0:
             delta = "0.0%"
             first = tt
         else:
             d = (tt / first) - 1
             delta = f"{d:+.1%}"
-        print(f"{commit:<25} {tt:.3f}s ({delta})")
+        print(f"{commit:<25} {tt:.3f}s ({delta}) | stdev {s:.3f}s ")
+
+    t = int(time.time() - whole_program_time_0)
+    total_time_taken_formatted = ", ".join(
+        f"{v} {n if v==1 else n+'s'}"
+        for v, n in ((t // 3600, "hour"), (t // 60 % 60, "minute"), (t % 60, "second"))
+        if v
+    )
+    print(
+        "Total time taken by the whole benchmarking program (including any setup):",
+        total_time_taken_formatted,
+    )
 
-    shutil.rmtree(self_check_dir)
+    shutil.rmtree(check_dir)
     for target_dir in target_dirs:
         shutil.rmtree(target_dir)