Enable nested namespace check in clang-tidy (pytorch#118506)

It is time to enable nested namespaces in the code.

Pull Request resolved: pytorch#118506
Approved by: https://github.com/albanD

Author: cyyever

.clang-tidy

@@ -42,7 +42,6 @@ misc-*,
 -misc-non-private-member-variables-in-classes,
 -misc-confusable-identifiers,
 modernize-*,
--modernize-concat-nested-namespaces,
 -modernize-macro-to-enum,
 -modernize-return-braced-init-list,
 -modernize-use-auto,

c10/core/WrapDimMinimal.cpp

@@ -1,7 +1,6 @@
 #include <c10/core/WrapDimMinimal.h>
 
-namespace c10 {
-namespace detail {
+namespace c10::detail {
 
 template <typename T>
 T maybe_wrap_dim_slow(T dim, T dim_post_expr, bool wrap_scalar) {
@@ -40,5 +39,4 @@ maybe_wrap_dim_slow(int64_t dim, int64_t dim_post_expr, bool wrap_scalar);
 template C10_API SymInt
 maybe_wrap_dim_slow(SymInt dim, SymInt dim_post_expr, bool wrap_scalar);
 
-} // namespace detail
-} // namespace c10
+} // namespace c10::detail

c10/core/impl/DeviceGuardImplInterface.cpp

@@ -1,7 +1,6 @@
 #include <c10/core/impl/DeviceGuardImplInterface.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 // NOLINTNEXTLINE(modernize-avoid-c-arrays,cppcoreguidelines-avoid-c-arrays)
 std::atomic<const DeviceGuardImplInterface*>
@@ -14,5 +13,4 @@ DeviceGuardImplRegistrar::DeviceGuardImplRegistrar(
   device_guard_impl_registry[static_cast<size_t>(type)].store(impl);
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl

c10/core/impl/FakeGuardImpl.h

@@ -4,8 +4,7 @@
 
 #include <array>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 // FakeGuardImpl is hardcoded to have eight devices.  Not for
 // any good reason, just to simplify code.
@@ -100,5 +99,4 @@ template <DeviceType T>
 thread_local std::array<StreamId, kFakeGuardImplMaxDevices>
     FakeGuardImpl<T>::current_streams_ = {0, 0, 0, 0, 0, 0, 0, 0};
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl

c10/core/impl/GPUTrace.cpp

@@ -1,8 +1,7 @@
 #include <c10/core/impl/GPUTrace.h>
 #include <c10/util/CallOnce.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 std::atomic<const PyInterpreter*> GPUTrace::gpuTraceState{nullptr};
 
@@ -16,5 +15,4 @@ void GPUTrace::set_trace(const PyInterpreter* trace) {
   });
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl

c10/core/impl/GPUTrace.h

@@ -2,8 +2,7 @@
 
 #include <c10/core/impl/PyInterpreter.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 struct C10_API GPUTrace {
   // On the x86 architecture the atomic operations are lock-less.
@@ -26,5 +25,4 @@ struct C10_API GPUTrace {
   }
 };
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl

c10/core/impl/HermeticPyObjectTLS.cpp

@@ -1,7 +1,6 @@
 #include <c10/core/impl/HermeticPyObjectTLS.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 thread_local std::atomic<bool> hermeticPyObjectState{false};
 
@@ -19,5 +18,4 @@ void HermeticPyObjectTLS::init_state() {
   haveState_ = true;
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl

c10/core/impl/LocalDispatchKeySet.cpp

@@ -1,7 +1,6 @@
 #include <c10/core/impl/LocalDispatchKeySet.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 // NB: POD, must be zero initialized!
 // Note [TLS Initialization]
@@ -115,5 +114,4 @@ bool tls_is_dispatch_keyset_excluded(DispatchKeySet ks) {
 bool tls_is_dispatch_keyset_included(DispatchKeySet ks) {
   return raw_local_dispatch_key_set.included().isSupersetOf(ks);
 }
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl

c10/core/impl/PyInterpreter.cpp

@@ -2,8 +2,7 @@
 #include <c10/core/TensorImpl.h>
 #include <c10/core/impl/PyInterpreter.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 struct NoopPyInterpreterVTable final : public PyInterpreterVTable {
   std::string name() const override {
@@ -125,5 +124,4 @@ void PyInterpreter::disarm() noexcept {
   vtable_ = &noop_vtable;
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl

c10/core/impl/PyObjectSlot.cpp

@@ -1,7 +1,6 @@
 #include <c10/core/impl/PyObjectSlot.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 PyObjectSlot::PyObjectSlot() : pyobj_interpreter_(nullptr), pyobj_(nullptr) {}
 
@@ -71,5 +70,4 @@ void PyObjectSlot::set_owns_pyobj(bool b) {
       reinterpret_cast<uintptr_t>(_unchecked_untagged_pyobj()) | b);
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl

c10/core/impl/PythonDispatcherTLS.cpp

@@ -2,8 +2,7 @@
 #include <c10/core/impl/LocalDispatchKeySet.h>
 #include <c10/core/impl/PythonDispatcherTLS.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 thread_local PyInterpreter* pythonDispatcherState;
 
@@ -27,5 +26,4 @@ void PythonDispatcherTLS::reset_state() {
       DispatchKey::PythonDispatcher, false);
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl

c10/core/impl/PythonDispatcherTLS.h

@@ -3,8 +3,7 @@
 #include <c10/core/impl/PyInterpreter.h>
 #include <c10/macros/Export.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 struct C10_API PythonDispatcherTLS {
   static void set_state(PyInterpreter* state);
@@ -22,5 +21,4 @@ struct C10_API DisablePythonDispatcher {
   PyInterpreter* old_;
 };
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl

c10/core/impl/SizesAndStrides.cpp

@@ -1,7 +1,6 @@
 #include <c10/core/impl/SizesAndStrides.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 void SizesAndStrides::resizeSlowPath(
     const size_t newSize,
@@ -77,5 +76,4 @@ void SizesAndStrides::resizeSlowPath(
   size_ = newSize;
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl

c10/core/impl/TorchDispatchModeTLS.cpp

@@ -6,8 +6,7 @@
 
 #include <utility>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 thread_local TorchDispatchModeTLS torchDispatchModeState;
 
@@ -192,5 +191,4 @@ std::string to_string(TorchDispatchModeKey mode_key) {
   }
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl

c10/core/impl/TorchDispatchModeTLS.h

@@ -3,8 +3,7 @@
 #include <c10/core/SafePyObject.h>
 #include <c10/macros/Export.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 enum class TorchDispatchModeKey : int8_t {
   FAKE,
@@ -61,5 +60,4 @@ C10_API bool dispatch_mode_enabled();
 
 C10_API std::string to_string(TorchDispatchModeKey mode_key);
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl

c10/cuda/CUDAAllocatorConfig.cpp

@@ -6,9 +6,7 @@
 #include <c10/cuda/driver_api.h>
 #endif
 
-namespace c10 {
-namespace cuda {
-namespace CUDACachingAllocator {
+namespace c10::cuda::CUDACachingAllocator {
 
 constexpr size_t kRoundUpPowerOfTwoIntervals = 16;
 
@@ -354,6 +352,4 @@ void setAllocatorSettings(const std::string& env) {
   CUDACachingAllocator::CUDAAllocatorConfig::instance().parseArgs(env.c_str());
 }
 
-} // namespace CUDACachingAllocator
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda::CUDACachingAllocator

c10/cuda/CUDACachingAllocator.cpp

@@ -40,8 +40,7 @@ namespace c10 {
 
 C10_DEFINE_REGISTRY(FreeCudaMemoryCallbacksRegistry, FreeMemoryCallback);
 
-namespace cuda {
-namespace CUDACachingAllocator {
+namespace cuda::CUDACachingAllocator {
 
 // Included here as this is externally used in CUDAAllocatorConfig
 const size_t kLargeBuffer =
@@ -3347,6 +3346,6 @@ struct BackendStaticInitializer {
 std::atomic<CUDAAllocator*> allocator;
 BackendStaticInitializer backend_static_initializer;
 
-} // namespace CUDACachingAllocator
-} // namespace cuda
+} // namespace cuda::CUDACachingAllocator
+
 } // namespace c10

c10/cuda/CUDADeviceAssertionHost.cpp

@@ -26,8 +26,7 @@
         false);                                                     \
   } while (0)
 
-namespace c10 {
-namespace cuda {
+namespace c10::cuda {
 
 namespace {
 
@@ -343,5 +342,4 @@ bool CUDAKernelLaunchRegistry::has_failed() const {
   return false;
 }
 
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda

c10/cuda/CUDAException.cpp

@@ -6,8 +6,7 @@
 
 #include <string>
 
-namespace c10 {
-namespace cuda {
+namespace c10::cuda {
 
 void c10_cuda_check_implementation(
     const int32_t err,
@@ -44,5 +43,4 @@ void c10_cuda_check_implementation(
   TORCH_CHECK(false, check_message);
 }
 
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda

c10/cuda/CUDAMallocAsyncAllocator.cpp

@@ -9,10 +9,7 @@
 #include <unordered_set>
 #include <vector>
 
-namespace c10 {
-namespace cuda {
-namespace CUDACachingAllocator {
-namespace CudaMallocAsync {
+namespace c10::cuda::CUDACachingAllocator::CudaMallocAsync {
 
 #if CUDA_VERSION >= 11040
 // CUDA device allocator that uses cudaMallocAsync to implement
@@ -898,7 +895,4 @@ CUDAAllocator* allocator() {
 
 #endif
 
-} // namespace CudaMallocAsync
-} // namespace CUDACachingAllocator
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda::CUDACachingAllocator::CudaMallocAsync

c10/cuda/CUDAMiscFunctions.cpp

@@ -1,8 +1,7 @@
 #include <c10/cuda/CUDAMiscFunctions.h>
 #include <stdlib.h>
 
-namespace c10 {
-namespace cuda {
+namespace c10::cuda {
 
 const char* get_cuda_check_suffix() noexcept {
   static char* device_blocking_flag = getenv("CUDA_LAUNCH_BLOCKING");
@@ -21,5 +20,4 @@ std::mutex* getFreeMutex() {
   return &cuda_free_mutex;
 }
 
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda

c10/cuda/CUDAStream.cpp

@@ -11,8 +11,7 @@
 #include <mutex>
 #include <vector>
 
-namespace c10 {
-namespace cuda {
+namespace c10::cuda {
 
 namespace {
 
@@ -339,5 +338,4 @@ std::ostream& operator<<(std::ostream& stream, const CUDAStream& s) {
   return stream << s.unwrap();
 }
 
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda

c10/cuda/driver_api.cpp

@@ -4,8 +4,8 @@
 #include <c10/util/Exception.h>
 #include <dlfcn.h>
 #include <iostream>
-namespace c10 {
-namespace cuda {
+
+namespace c10::cuda {
 
 namespace {
 
@@ -42,7 +42,6 @@ C10_EXPORT DriverAPI* DriverAPI::get() {
   return &singleton;
 }
 
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda
 
 #endif