[lint] fixes to mypy linter

I thought I landed this already, but:
- Don't run one mypy instance per file, run one per config
- Do the same for flake8
- Properly handle stub files

Pull Request resolved: https://github.com/pytorch/pytorch/pull/68479

Approved by: https://github.com/janeyx99

.lintrunner.toml

@@ -53,6 +53,9 @@ include_patterns = [
     'test/cpp/tensorexpr/**/*.h',
     'test/cpp/tensorexpr/**/*.cpp',
 ]
+exclude_patterns = [
+    'torch/csrc/jit/serialization/mobile_bytecode_generated.h',
+]
 init_command = [
     'python3',
     'tools/linter/adapters/s3_init.py',
@@ -74,13 +77,15 @@ command = [
 code = 'MYPY'
 include_patterns = [
     'torch/**/*.py',
+    'torch/**/*.pyi',
     'caffe2/**/*.py',
+    'caffe2/**/*.pyi',
     'test/test_bundled_images.py',
     'test/test_bundled_inputs.py',
     'test/test_complex.py',
     'test/test_datapipe.py',
     'test/test_futures.py',
-    'test/test_numpy_interop.py',
+    # 'test/test_numpy_interop.py',
     'test/test_torch.py',
     'test/test_type_hints.py',
     'test/test_type_info.py',
@@ -90,6 +95,17 @@ exclude_patterns = [
     'torch/include/**',
     'torch/csrc/**',
     'torch/distributed/elastic/agent/server/api.py',
+    'torch/testing/_internal/**',
+    'torch/distributed/fsdp/fully_sharded_data_parallel.py',
+    # TODO(suo): these exclusions were added just to get lint clean on master.
+    # Follow up to do more target suppressions and remove them.
+    'torch/distributed/fsdp/flatten_params_wrapper.py',
+    'torch/ao/quantization/fx/convert.py',
+    'torch/ao/quantization/_dbr/function_fusion.py',
+    'test/test_datapipe.py',
+    'caffe2/contrib/fakelowp/test/test_batchmatmul_nnpi_fp16.py',
+    'test/test_numpy_interop.py',
+    'torch/torch_version.py',
 ]
 command = [
     'python3',
@@ -108,6 +124,7 @@ init_command = [
     'mypy==0.812',
     'junitparser==2.1.1',
     'rich==10.9.0',
+    'pyyaml==6.0',
 ]
 
 [[linter]]
@@ -319,6 +336,7 @@ include_patterns = [
 ]
 exclude_patterns = [
     'aten/src/ATen/native/quantized/cpu/qnnpack/**',
+    'torch/csrc/jit/serialization/mobile_bytecode_generated.h',
 ]
 command = [
     'python3',

tools/linter/adapters/flake8_linter.py

@@ -1,5 +1,4 @@
 import argparse
-import concurrent.futures
 import json
 import logging
 import os
@@ -244,8 +243,8 @@ def get_issue_documentation_url(code: str) -> str:
     return ""
 
 
-def check_file(
+def check_files(
-    filename: str,
+    filenames: List[str],
     binary: str,
     flake8_plugins_path: Optional[str],
     severities: Dict[str, LintSeverity],
@@ -253,7 +252,7 @@ def check_file(
 ) -> List[LintMessage]:
     try:
         proc = run_command(
-            [binary, "--exit-zero", filename],
+            [binary, "--exit-zero"] + filenames,
             extra_env={"FLAKE8_PLUGINS_PATH": flake8_plugins_path}
             if flake8_plugins_path
             else None,
@@ -262,7 +261,7 @@ def check_file(
     except (OSError, subprocess.CalledProcessError) as err:
         return [
             LintMessage(
-                path=filename,
+                path=None,
                 line=None,
                 char=None,
                 code="FLAKE8",
@@ -369,28 +368,9 @@ def main() -> None:
             assert len(parts) == 2, f"invalid severity `{severity}`"
             severities[parts[0]] = LintSeverity(parts[1])
 
-    with concurrent.futures.ThreadPoolExecutor(
+    lint_messages = check_files(args.filenames, args.binary, flake8_plugins_path, severities, args.retries)
-        max_workers=os.cpu_count(),
+    for lint_message in lint_messages:
-        thread_name_prefix="Thread",
+        print(json.dumps(lint_message._asdict()), flush=True)
-    ) as executor:
-        futures = {
-            executor.submit(
-                check_file,
-                filename,
-                args.binary,
-                flake8_plugins_path,
-                severities,
-                args.retries,
-            ): filename
-            for filename in args.filenames
-        }
-        for future in concurrent.futures.as_completed(futures):
-            try:
-                for lint_message in future.result():
-                    print(json.dumps(lint_message._asdict()), flush=True)
-            except Exception:
-                logging.critical('Failed at "%s".', futures[future])
-                raise
 
 
 if __name__ == "__main__":

tools/linter/adapters/mypy_linter.py

@@ -1,5 +1,4 @@
 import argparse
-import concurrent.futures
 import json
 import logging
 import os
@@ -8,6 +7,7 @@ import subprocess
 import sys
 import time
 from enum import Enum
+from pathlib import Path
 from typing import Any, Dict, List, NamedTuple, Optional, Pattern
 
 
@@ -56,7 +56,6 @@ RESULTS_RE: Pattern[str] = re.compile(
 )
 
 
-
 def run_command(
     args: List[str],
     *,
@@ -82,15 +81,16 @@ severities = {
     "note": LintSeverity.ADVICE,
 }
 
-def check_file(
+
-    filename: str,
+def check_files(
+    filenames: List[str],
     config: str,
     binary: str,
     retries: int,
 ) -> List[LintMessage]:
     try:
         proc = run_command(
-            [binary, f"--config={config}", filename],
+            [binary, f"--config={config}"] + filenames,
             extra_env={},
             retries=retries,
         )
@@ -105,9 +105,7 @@ def check_file(
                 name="command-failed",
                 original=None,
                 replacement=None,
-                description=(
+                description=(f"Failed due to {err.__class__.__name__}:\n{err}"),
-                    f"Failed due to {err.__class__.__name__}:\n{err}"
-                ),
             )
         ]
     stdout = str(proc.stdout, "utf-8").strip()
@@ -172,27 +170,26 @@ def main() -> None:
         stream=sys.stderr,
     )
 
-    with concurrent.futures.ThreadPoolExecutor(
+    # Use a dictionary here to preserve order. mypy cares about order,
-        max_workers=os.cpu_count(),
+    # tragically, e.g. https://github.com/python/mypy/issues/2015
-        thread_name_prefix="Thread",
+    filenames: Dict[str, bool] = {}
-    ) as executor:
+
-        futures = {
+    # If a stub file exists, have mypy check it instead of the original file, in
-            executor.submit(
+    # accordance with PEP-484 (see https://www.python.org/dev/peps/pep-0484/#stub-files)
-                check_file,
+    for filename in args.filenames:
-                filename,
+        if filename.endswith(".pyi"):
-                args.config,
+            filenames[filename] = True
-                args.binary,
+            continue
-                args.retries,
+
-            ): filename
+        stub_filename = filename.replace(".py", ".pyi")
-            for filename in args.filenames
+        if Path(stub_filename).exists():
-        }
+            filenames[stub_filename] = True
-        for future in concurrent.futures.as_completed(futures):
+        else:
-            try:
+            filenames[filename] = True
-                for lint_message in future.result():
+
-                    print(json.dumps(lint_message._asdict()), flush=True)
+    lint_messages = check_files(list(filenames), args.config, args.binary, args.retries)
-            except Exception:
+    for lint_message in lint_messages:
-                logging.critical('Failed at "%s".', futures[future])
+        print(json.dumps(lint_message._asdict()), flush=True)
-                raise
 
 
 if __name__ == "__main__":

torch/csrc/jit/runtime/decomposition_registry_util.cpp

@@ -7,15 +7,14 @@
  */
 #include <torch/csrc/jit/jit_log.h>
 #include <torch/csrc/jit/passes/inliner.h>
-#include <torch/csrc/jit/runtime/operator.h>
 #include <torch/csrc/jit/runtime/decomposition_registry_util.h>
+#include <torch/csrc/jit/runtime/operator.h>
 
 namespace torch {
 namespace jit {
 
-
 const std::string decomp_funcs =
-R"(def var_decomposition(input: Tensor,
+    R"(def var_decomposition(input: Tensor,
     dim: Optional[List[int]]=None,
     correction: Optional[int]=None,
     keepdim: bool=False) -> Tensor:

torch/csrc/jit/runtime/profiling_graph_executor_impl.cpp

@@ -456,7 +456,9 @@ void ProfilingGraphExecutorImpl::runNoGradOptimizations(
     for (const auto& passPair : getCustomPostPasses()) {
       passPair.first(graph);
     }
-    GRAPH_DEBUG("After customPostPasses, before RemoveTensorTypeSpecializations \n", *graph);
+    GRAPH_DEBUG(
+        "After customPostPasses, before RemoveTensorTypeSpecializations \n",
+        *graph);
     RemoveTensorTypeSpecializations(graph);
     GRAPH_DEBUG("After RemoveTensorTypeSpecializations\n", *graph);
   }