Rework clang-format include groups.

.clang-format

@@ -66,15 +66,52 @@ ForEachMacros:
   - Q_FOREACH
   - BOOST_FOREACH
 IncludeBlocks:   Regroup
+# Include Groups (by priority):
+#   1. C system headers:     `#include <*.h>`
+#
+#   2. C++ system headers:   `#include <*>`
+#
+#   3. PyTorch headers:
+#     a) API headers:        `#include <torch/*.h>`
+#     b) ATen headers:       `#include <ATen/*.h>`
+#     c) c10 headers:        `#include <c10/*.h>`
+#
+#     Note that PyTorch headers should be included with angle-brackets so as to
+#     keep consistency with PyTorch source code. This should also prevent
+#     headers to be included twice because of different paths.
+#
+#     Google C++ Style Guide dictates (3) and (4) should belong to the same
+#     group. However, this breaks PyTorch/XLA build. The problem is that both
+#     PyTorch and abseil-cpp (included by OpenXLA) define (among many) the
+#     `LOG()` macro. This leads to problems, such as OpenXLA functions calling
+#     PyTorch `LOG()` macro due to include order.
+#
+#     In order to avoid such a problem, we make PyTorch into its own group,
+#     just after the system libraries. This works because including OpenXLA
+#     headers will redefine PyTorch `LOG()` macro, solving the problem above.
+#     This, however might fail one day, if PyTorch starts calling `LOG()` macro
+#     in one of the headers we include in PyTorch/XLA.
+#
+#   4. External headers:
+#     a) abseil-cpp headers: `#include "absl/*.h"
+#     b) OpenXLA headers:    `#include "xla/*.h"
+#
+#   5. Project headers:
+#     a) _XLAC headers:      `#include "torch_xla/csrc/*.h"`
+#     b) Test headers:       `#include "test/*.h"`
 IncludeCategories:
-  - Regex:           '^<ext/.*\.h>'
-    Priority:        2
+  - Regex:           '^<torch/.*>'
+    Priority:        3
   - Regex:           '^<.*\.h>'
     Priority:        1
   - Regex:           '^<.*'
     Priority:        2
+  - Regex:           '"torch_xla/.*"'
+    Priority:        5
+  - Regex:           '"test/.*"'
+    Priority:        5
   - Regex:           '.*'
-    Priority:        3
+    Priority:        4
 IncludeIsMainRegex: '([-_](test|unittest))?$'
 IndentCaseLabels: true
 IndentPPDirectives: None

test/cpp/cpp_test_util.h

@@ -2,14 +2,16 @@
 #define XLA_TEST_CPP_CPP_TEST_UTIL_H_
 
 #include <gtest/gtest.h>
-#include <torch/torch.h>
 
 #include <cmath>
 #include <functional>
 #include <string>
 #include <unordered_set>
 
+#include <torch/torch.h>
+
 #include "absl/types/span.h"
+
 #include "torch_xla/csrc/debug_util.h"
 #include "torch_xla/csrc/device.h"
 #include "torch_xla/csrc/dtype.h"

test/cpp/metrics_snapshot.cpp

@@ -2,7 +2,8 @@
 
 #include <regex>
 
-#include "torch/csrc/lazy/core/metrics.h"
+#include <torch/csrc/lazy/core/metrics.h>
+
 #include "torch_xla/csrc/runtime/tf_logging.h"
 #include "torch_xla/csrc/runtime/util.h"
 

test/cpp/metrics_snapshot.h

@@ -7,9 +7,10 @@
 #include <unordered_set>
 #include <vector>
 
-#include "torch_xla/csrc/runtime/metrics.h"
 #include "xla/types.h"
 
+#include "torch_xla/csrc/runtime/metrics.h"
+
 namespace torch_xla {
 namespace cpp_test {
 

test/cpp/test_aten_xla_tensor_1.cpp

@@ -1,8 +1,12 @@
 #include <gtest/gtest.h>
-#include <torch/torch.h>
 
 #include <iostream>
 
+#include <torch/torch.h>
+
+#include "xla/permutation_util.h"
+#include "xla/util.h"
+
 #include "test/cpp/cpp_test_util.h"
 #include "test/cpp/torch_xla_test.h"
 #include "torch_xla/csrc/aten_xla_bridge.h"
@@ -12,8 +16,6 @@
 #include "torch_xla/csrc/ops/ops.h"
 #include "torch_xla/csrc/runtime/metrics.h"
 #include "torch_xla/csrc/torch_util.h"
-#include "xla/permutation_util.h"
-#include "xla/util.h"
 
 namespace torch_xla {
 namespace cpp_test {

test/cpp/test_aten_xla_tensor_2.cpp

@@ -1,9 +1,13 @@
 #include <gtest/gtest.h>
-#include <torch/torch.h>
 
 #include <iostream>
 #include <tuple>
 
+#include <torch/torch.h>
+
+#include "xla/permutation_util.h"
+#include "xla/util.h"
+
 #include "test/cpp/cpp_test_util.h"
 #include "test/cpp/torch_xla_test.h"
 #include "torch_xla/csrc/aten_xla_bridge.h"
@@ -13,8 +17,6 @@
 #include "torch_xla/csrc/ops/ops.h"
 #include "torch_xla/csrc/runtime/metrics.h"
 #include "torch_xla/csrc/torch_util.h"
-#include "xla/permutation_util.h"
-#include "xla/util.h"
 
 namespace torch_xla {
 namespace cpp_test {

test/cpp/test_aten_xla_tensor_3.cpp

@@ -1,8 +1,12 @@
 #include <gtest/gtest.h>
-#include <torch/torch.h>
 
 #include <iostream>
 
+#include <torch/torch.h>
+
+#include "xla/permutation_util.h"
+#include "xla/util.h"
+
 #include "test/cpp/cpp_test_util.h"
 #include "test/cpp/torch_xla_test.h"
 #include "torch_xla/csrc/aten_xla_bridge.h"
@@ -12,8 +16,6 @@
 #include "torch_xla/csrc/ops/ops.h"
 #include "torch_xla/csrc/runtime/metrics.h"
 #include "torch_xla/csrc/torch_util.h"
-#include "xla/permutation_util.h"
-#include "xla/util.h"
 
 namespace torch_xla {
 namespace cpp_test {

test/cpp/test_aten_xla_tensor_4.cpp

@@ -1,8 +1,12 @@
 #include <gtest/gtest.h>
-#include <torch/torch.h>
 
 #include <iostream>
 
+#include <torch/torch.h>
+
+#include "xla/permutation_util.h"
+#include "xla/util.h"
+
 #include "test/cpp/cpp_test_util.h"
 #include "test/cpp/torch_xla_test.h"
 #include "torch_xla/csrc/aten_xla_bridge.h"
@@ -12,8 +16,6 @@
 #include "torch_xla/csrc/ops/ops.h"
 #include "torch_xla/csrc/runtime/metrics.h"
 #include "torch_xla/csrc/torch_util.h"
-#include "xla/permutation_util.h"
-#include "xla/util.h"
 
 namespace torch_xla {
 namespace cpp_test {

test/cpp/test_aten_xla_tensor_5.cpp

@@ -1,8 +1,12 @@
 #include <gtest/gtest.h>
-#include <torch/torch.h>
 
 #include <iostream>
 
+#include <torch/torch.h>
+
+#include "xla/permutation_util.h"
+#include "xla/util.h"
+
 #include "test/cpp/cpp_test_util.h"
 #include "test/cpp/torch_xla_test.h"
 #include "torch_xla/csrc/aten_xla_bridge.h"
@@ -12,8 +16,6 @@
 #include "torch_xla/csrc/ops/ops.h"
 #include "torch_xla/csrc/runtime/metrics.h"
 #include "torch_xla/csrc/torch_util.h"
-#include "xla/permutation_util.h"
-#include "xla/util.h"
 
 namespace torch_xla {
 namespace cpp_test {

test/cpp/test_aten_xla_tensor_6.cpp

@@ -1,8 +1,12 @@
 #include <gtest/gtest.h>
-#include <torch/torch.h>
 
 #include <iostream>
 
+#include <torch/torch.h>
+
+#include "xla/permutation_util.h"
+#include "xla/util.h"
+
 #include "test/cpp/cpp_test_util.h"
 #include "test/cpp/torch_xla_test.h"
 #include "torch_xla/csrc/aten_xla_bridge.h"
@@ -12,8 +16,6 @@
 #include "torch_xla/csrc/ops/ops.h"
 #include "torch_xla/csrc/runtime/metrics.h"
 #include "torch_xla/csrc/torch_util.h"
-#include "xla/permutation_util.h"
-#include "xla/util.h"
 
 namespace torch_xla {
 namespace cpp_test {

test/cpp/test_device.cpp

@@ -1,6 +1,7 @@
 #include <gtest/gtest.h>
 
 #include "absl/strings/str_cat.h"
+
 #include "torch_xla/csrc/device.h"
 
 static void CheckFormatError(const std::string& spec) {

test/cpp/test_lazy.cpp

@@ -1,9 +1,11 @@
 #include <gtest/gtest.h>
 
+#include <torch/csrc/lazy/core/shape.h>
+
+#include "xla/shape.h"
+
 #include "test/cpp/torch_xla_test.h"
-#include "torch/csrc/lazy/core/shape.h"
 #include "torch_xla/csrc/helpers.h"
-#include "xla/shape.h"
 
 namespace torch_xla {
 namespace cpp_test {

test/cpp/test_replication.cpp

@@ -4,6 +4,9 @@
 #include <iostream>
 
 #include "absl/synchronization/blocking_counter.h"
+#include "xla/hlo/builder/xla_builder.h"
+#include "xla/shape_util.h"
+
 #include "test/cpp/cpp_test_util.h"
 #include "test/cpp/torch_xla_test.h"
 #include "torch_xla/csrc/aten_xla_bridge.h"
@@ -14,8 +17,6 @@
 #include "torch_xla/csrc/tensor_util.h"
 #include "torch_xla/csrc/thread_pool.h"
 #include "torch_xla/csrc/torch_util.h"
-#include "xla/hlo/builder/xla_builder.h"
-#include "xla/shape_util.h"
 
 namespace torch_xla {
 namespace cpp_test {

test/cpp/test_status_common.h

@@ -29,6 +29,7 @@
 
 #include "absl/status/status.h"
 #include "absl/status/statusor.h"
+
 #include "test/cpp/cpp_test_util.h"
 #include "torch_xla/csrc/runtime/env_vars.h"
 #include "torch_xla/csrc/status.h"

test/cpp/test_tensor.cpp

@@ -4,10 +4,12 @@
 #include <limits>
 #include <vector>
 
+#include <torch/csrc/autograd/variable.h>
+
 #include "absl/base/nullability.h"
+
 #include "test/cpp/cpp_test_util.h"
 #include "test/cpp/torch_xla_test.h"
-#include "torch/csrc/autograd/variable.h"
 #include "torch_xla/csrc/aten_xla_bridge.h"
 #include "torch_xla/csrc/tensor.h"
 #include "torch_xla/csrc/tensor_methods.h"

test/cpp/test_xla_generator.cpp

@@ -1,9 +1,10 @@
 #include <gmock/gmock.h>
 #include <gtest/gtest.h>
-#include <torch/torch.h>
 
 #include <cstdlib>
 
+#include <torch/torch.h>
+
 #include "test/cpp/torch_xla_test.h"
 #include "torch_xla/csrc/xla_generator.h"
 

test/cpp/test_xla_sharding.cpp

@@ -2,10 +2,14 @@
 #include <google/protobuf/repeated_field.h>
 #include <gtest/gtest.h>
 #include <stdlib.h>
-#include <torch/csrc/lazy/core/lazy_graph_executor.h>
 
 #include <iostream>
 
+#include <torch/csrc/lazy/core/lazy_graph_executor.h>
+
+#include "xla/protobuf_util.h"
+#include "xla/xla_data.pb.h"
+
 #include "test/cpp/cpp_test_util.h"
 #include "test/cpp/torch_xla_test.h"
 #include "torch_xla/csrc/aten_xla_bridge.h"
@@ -19,8 +23,6 @@
 #include "torch_xla/csrc/tensor_methods.h"
 #include "torch_xla/csrc/tensor_util.h"
 #include "torch_xla/csrc/xla_sharding_util.h"
-#include "xla/protobuf_util.h"
-#include "xla/xla_data.pb.h"
 
 namespace torch_xla {
 namespace cpp_test {

test/cpp/torch_xla_test.cpp

@@ -3,6 +3,7 @@
 #include <ATen/ATen.h>
 
 #include "absl/memory/memory.h"
+
 #include "torch_xla/csrc/aten_xla_bridge.h"
 #include "torch_xla/csrc/device.h"
 #include "torch_xla/csrc/helpers.h"

torch_xla/csrc/aten_fallback.cpp

@@ -4,11 +4,12 @@
 #include <ATen/native/CPUFallback.h>
 #include <ATen/ops/_copy_from_and_resize.h>
 #include <ATen/ops/_to_cpu.h>
-#include <torch/csrc/utils/device_lazy_init.h>
 
 #include <unordered_map>
 #include <vector>
 
+#include <torch/csrc/utils/device_lazy_init.h>
+
 #include "torch_xla/csrc/function_call_tracker.h"
 #include "torch_xla/csrc/runtime/metrics.h"
 #include "torch_xla/csrc/runtime/tf_logging.h"

torch_xla/csrc/aten_xla_bridge.cpp

@@ -1,14 +1,16 @@
 #include "torch_xla/csrc/aten_xla_bridge.h"
 
 #include <ATen/FunctionalTensorWrapper.h>
-#include <torch/csrc/lazy/core/tensor_util.h>
 
 #include <map>
 #include <string>
 #include <vector>
 
+#include <torch/csrc/lazy/core/tensor_util.h>
+
 #include "absl/log/absl_check.h"
 #include "absl/strings/str_cat.h"
+
 #include "torch_xla/csrc/device.h"
 #include "torch_xla/csrc/runtime/debug_macros.h"
 #include "torch_xla/csrc/runtime/runtime.h"

torch_xla/csrc/aten_xla_bridge.h

@@ -10,6 +10,7 @@
 #include "absl/base/nullability.h"
 #include "absl/status/statusor.h"
 #include "absl/types/span.h"
+
 #include "torch_xla/csrc/device.h"
 #include "torch_xla/csrc/tensor.h"