add code-formatting with black for python and with clang-format for c++

[hhorii]

Summary

Adds code-formatting with black and pre-commit

Details and comments

Aer has not used code-formatting. This PR introduces black for all python files and clang-format for all cpp files.
tox -elint validates lint and tox -eblack, and tox -epre-commit format python and c++ codes respectively.

[hhorii]

I hope to apply this to 0.11 stable. There are few changes in cpp since 0.11 and this PR includes many changes in cpp.

[mtreinish]

We should not apply this to the stable branch, it's a risky change with no end user benefit the exact kind of thing we should not backport.

[mtreinish]

In general I'm definitely in favor of making a change like this to use auto-formatting for code. We should also do the same for using black on the python code. I do have some questions, mainly around the use of precommit in particular and the creation of a separate CI job to check it.

My concern with pre-commit is it's yet another tool to manage running tests and linting. I'm just wondering what in particular it's providing here. Like if we just called clang-format in our tox config would that accomplish the same thing, or is pre-commit here doing some extra setup like installing dependencies that tox won't provide us? Like what happens if you runpre-commit without clang installed? In general I think it would be good to try and bake this style correction into our already existing workflows for this (assuming it's feasible).

For the CI job, I think we should bake running the clang-format job into our existing lint job. We're already running that job to validate python code formatting so we can just add C++ there too (we do something similar in other projects with cargo fmt for rust).

The other thought I had is around stable configuration, does clang formatting change between releases? Would it make sense to define a static config file for clang format to use for this.

Also, we should update CONTRIBUTING.md to mention how to run clang format as it will be a custom lint step for aer code.

As for the git blame we can add a .git-blame-ignore-revs file to the repo and document it, like what we have in terra from when we adopted black: https://github.com/Qiskit/qiskit-terra/blob/main/.git-blame-ignore-revs and https://github.com/Qiskit/qiskit-terra/blob/main/CONTRIBUTING.md#dealing-with-the-git-blame-ignore-list

[hhorii]

Currently, Aer uses pycodestyle and pylint for format-checking of python codes. They can not cover cpp codes in my understanding. pre-commit has hooks for various languages and we will be able to use it for the all. Also I confirmed that pre-commit does not need LLVM installation.
I reused tests.yml for pre-commit instead of standalone workflow.

[jakelishman]

fwiw: the clang-format recipe for pre-commit uses the PyPI package of the same name, which basically just stuffs the clang-format executable into a Python wheel for each system. That's a pretty questionable mismatch of different package managers to me, but as long as pre-commit isolates its internal Python environments (which it does seem to do) to avoid polluting my PATH with its clang-format, that's not the end of the world for me.

If we're bikeshedding the use of pre-commit at all, then I would be ok with putting fast formatting fixers in such a file (total runtime < 1s), but against putting linters in it. I don't think commiting should be a heavy operation, and linting (even when notionally limited to changed files - linters usually need to explore more of the package statically anyway) is not sufficiently fast. The sort of bugs it catches are often fine to handle in CI anyway - I don't think it buys anything to force devs to run a not-exactly-the-same-as-CI lint locally on every commit.

I also don't entirely see the point in allowing pre-commit hook that might cause a commit to be rejected (like lint). We're going to have to run all the proper formatting/lint suites in CI anyway because we can't enforce that somebody installs the pre-commit hooks (or doesn't use git commit --no-verify), so mostly we're just making certain things harder, like storing debug progress in series of commits to be rebased, for the users who might do that sort of thing, and the users who won't are the ones most likely to not have pre-commit configured anwyay.

(That said, even if we did put the config in the repo, I probably wouldn't actually install the hooks myself - I usually just configure my editor to auto-format on write / manual command.)

[hhorii]

I also don't entirely see the point in allowing pre-commit hook that might cause a commit to be rejected (like lint).

@mtreinish If e8ee5f9 adds pre-commit with existing Tests/lint workflow and commit will be failed if format is wrong. Original 8ee44e1 uses a different workflow for pre-commit and follows @jakelishman wants.

[jakelishman]

I mean, I'm not really in favour of another tool either. In this case, we could just skip the middle man and add the clang-format PyPI package to tox and just continue to use that.

I don't feel super strongly about it, and I wouldn't hold anything up though, especially since I don't do much development on Aer.

[hhorii]

We will introduce new code-formatters as the last commit for 0.12.0 release.

[hhorii]

Compilation for standalone in windows-2019 fails after applying clang-format. Now I'm narrowing scope of formatting to identify root causes with CI runs.

[mtreinish]

Thanks for doing this, I'm looking forward to having consistent code formatting everywhere.

I'm going to run this locally on a fresh clone of the repo to try and verify the giant diff is just the application of clang-format and black to all the source code prior to approving. But before I start working on that I left a few inline comments.

[mtreinish]

There are only 2 tools mentioned here, you forgot to mention clang format here.

[hhorii]

It was written in an old commit, but it was wrongly overwritten while resolving conflicts...

[mtreinish]

While we're doing this for the first time to do you want to use ~=23.0 for the latest version? black does a compatibility for a year's worth of releases and then integrates behavior/format changes with a new major version for each year. Since we're in 2023 now we can use the newer version here. Not that it particularly matters, but I figured it would just save us from having to do an upgrade and big format change again at some point in the future.

[mtreinish]

The --line-length argument here is causing this check to diverge with what you have committed to your branch. Can you add the configuration to the pyproject.toml file like: https://github.com/Qiskit/qiskit-terra/blob/main/pyproject.toml#L5-L7 and remove the argument here? That way we'll have the same line length consistently between all uses of the black command.

[mtreinish]

Is this used by anything?

[hhorii]

no longer :-)

[mtreinish]

Why is this ignored?

[hhorii]

clang-format wrongly formats pragma.

   DISABLE_WARNING(-Wall)                                                       \
   DISABLE_WARNING(-Wextra)                                                     \
   DISABLE_WARNING(-Wshadow)                                                    \
-  DISABLE_WARNING(-Wfloat-equal)                                               \
+  DISABLE_WARNING(-Wfloat - equal)                                             \
   DISABLE_WARNING(-Wundef)                                                     \
   DISABLE_WARNING(-Wpedantic)                                                  \
-  DISABLE_WARNING(-Wredundant-decls)                                           \
+  DISABLE_WARNING(-Wredundant - decls)                                         \
   DISABLE_WARNING(-Wsign-compare)

I tried to use WhitespaceSensitiveMacros but not yet find a good configuration that disable formatting of white spaces in macro.

[mtreinish]

What about the c++ files in qiskit_aer and test?

[hhorii]

good comment. let me add them into the list. maybe contrib should be included, too.

[mtreinish]

I caught one small oversight in the CI config. Also the black config isn't present in the pyproject.toml so it has the default line set of 88 characters instead of 100 (which I assume you were going for).

Besides that I'm having trouble reproducing the same output on a local clone (I'm getting subtle differences between my test repo and your branch). I'm not sure what's going on for the most part the diff is identical except for:

 controllers/aer_controller.hpp                                  |    2 +-
 framework/config.hpp                                            |    1 +
 framework/pybind_casts.hpp                                      |    1 +
 open_pulse/eval_hamiltonian.hpp                                 |   10 ++++++----
 simulators/matrix_product_state/matrix_product_state.hpp        |   10 +++++-----
 simulators/matrix_product_state/matrix_product_state_tensor.hpp |   21 +++++++++++----------
 simulators/statevector/chunk/cuStateVec_chunk_container.hpp     |    6 +++---
 simulators/statevector/statevector_state.hpp                    |   10 +++++-----
 simulators/tensor_network/tensor_net.hpp                        |    2 +-
 simulators/tensor_network/tensor_net_state.hpp                  |   10 +++++-----
 10 files changed, 39 insertions(+), 34 deletions(-)

I'm wondering if your local branch wasn't clean when you committed things. Like there are a few places the diff is more than formatting like extra #include statements which is more than I'd expect clang format to do.

[mtreinish]

This is the full diff the a side is my fresh clone that I copied the config files into and ran tox -eclang-format and tox -eblack on. The b side is your PR branch at the current HEAD (54399d1). Some of this diff looks like things I'd expect clang format to fix like comment spacing and comment types. But others I'm confused by. I might have missed a config in my copy over or maybe there is a clang format difference on linux?

diff --git a/src/controllers/aer_controller.hpp b/home/mtreinish/git/qiskit/qiskit-aer/src/controllers/aer_controller.hpp
index bf144fc2..f0ee3acd 100644
--- a/src/controllers/aer_controller.hpp
+++ b/home/mtreinish/git/qiskit/qiskit-aer/src/controllers/aer_controller.hpp
@@ -989,7 +989,7 @@ Result Controller::execute(std::vector<Circuit> &circuits,
     const int NUM_RESULTS = result.results.size();
     // following looks very similar but we have to separate them to avoid omp
     // nested loops that causes performance degradation (DO NOT use if statement
-    //in #pragma omp)
+    // in #pragma omp)
     if (parallel_experiments_ == 1) {
       for (int j = 0; j < NUM_RESULTS; ++j) {
         set_parallelization_circuit(circuits[j], noise_model, methods[j]);
diff --git a/src/framework/config.hpp b/home/mtreinish/git/qiskit/qiskit-aer/src/framework/config.hpp
index 6ebe7865..2890747a 100644
--- a/src/framework/config.hpp
+++ b/home/mtreinish/git/qiskit/qiskit-aer/src/framework/config.hpp
@@ -15,6 +15,7 @@
 #ifndef _aer_framework_config_hpp_
 #define _aer_framework_config_hpp_
 
+#include "json.hpp"
 #include "types.hpp"
 #include <optional>
 #include <string>
diff --git a/src/framework/pybind_casts.hpp b/home/mtreinish/git/qiskit/qiskit-aer/src/framework/pybind_casts.hpp
index 1db098fb..7e18ee30 100644
--- a/src/framework/pybind_casts.hpp
+++ b/home/mtreinish/git/qiskit/qiskit-aer/src/framework/pybind_casts.hpp
@@ -16,6 +16,7 @@
 #define _aer_framework_pybind_casts_hpp_
 
 #include "../simulators/stabilizer/clifford.hpp"
+#include <pybind11/numpy.h>
 
 namespace py = pybind11;
 
diff --git a/src/open_pulse/eval_hamiltonian.hpp b/home/mtreinish/git/qiskit/qiskit-aer/src/open_pulse/eval_hamiltonian.hpp
index aea06967..dfa2197e 100644
--- a/src/open_pulse/eval_hamiltonian.hpp
+++ b/home/mtreinish/git/qiskit/qiskit-aer/src/open_pulse/eval_hamiltonian.hpp
@@ -15,6 +15,8 @@
 #ifndef _EVAL_HAMILTONIAN_HPP
 #define _EVAL_HAMILTONIAN_HPP
 
+#include "framework/types.hpp"
+#include "iterators.hpp"
 #include "misc/warnings.hpp"
 #include <complex>
 #include <string>
@@ -42,10 +44,10 @@ struct hash<ParserValues> {
 } // namespace std
 
 // TODO: Document
-complex_t evaluate_hamiltonian_expression(
+AER::complex_t evaluate_hamiltonian_expression(
     const std::string &expr_string, const std::vector<double> &vars,
     const std::vector<std::string> &vars_names,
-    const std::unordered_map<std::string, complex_t> &chan_values) {
+    const std::unordered_map<std::string, AER::complex_t> &chan_values) {
 
   static std::unordered_map<std::string, std::unique_ptr<ParserValues>>
       parser_expr;
@@ -75,7 +77,7 @@ complex_t evaluate_hamiltonian_expression(
   // std::cout << "Getting parser " << std::hex << parser << "\n";
 
   auto maybe_update_value = [parser](const std::string &var_name,
-                                     const complex_t &var_value) {
+                                     const AER::complex_t &var_value) {
     if (parser->var_values.find(var_name) == parser->var_values.end()) {
       parser->var_values.emplace(var_name,
                                  std::make_unique<mup::Value>(var_value));
@@ -93,7 +95,7 @@ complex_t evaluate_hamiltonian_expression(
 
   for (const auto &idx_var : enumerate(vars)) {
     size_t index = idx_var.first;
-    auto var_value = static_cast<complex_t>(idx_var.second);
+    auto var_value = static_cast<AER::complex_t>(idx_var.second);
     maybe_update_value(vars_names[index], var_value);
   }
 
diff --git a/src/simulators/matrix_product_state/matrix_product_state.hpp b/home/mtreinish/git/qiskit/qiskit-aer/src/simulators/matrix_product_state/matrix_product_state.hpp
index 354e6021..4105fd3b 100644
--- a/src/simulators/matrix_product_state/matrix_product_state.hpp
+++ b/home/mtreinish/git/qiskit/qiskit-aer/src/simulators/matrix_product_state/matrix_product_state.hpp
@@ -274,14 +274,14 @@ const stringmap_t<Gates>
                      {"rx", Gates::rx},     // Pauli-X rotation gate
                      {"ry", Gates::ry},     // Pauli-Y rotation gate
                      {"rz", Gates::rz},     // Pauli-Z rotation gate
-                     // Waltz Gates
+                     /* Waltz Gates */
                      {"p", Gates::u1},  // zero-X90 pulse waltz gate
                      {"u1", Gates::u1}, // zero-X90 pulse waltz gate
                      {"u2", Gates::u2}, // single-X90 pulse waltz gate
                      {"u3", Gates::u3}, // two X90 pulse waltz gate
                      {"u", Gates::u3},  // two X90 pulse waltz gate
                      {"U", Gates::u3},  // two X90 pulse waltz gate
-                     // Two-qubit gates
+                     /* Two-qubit gates */
                      {"CX", Gates::cx},   // Controlled-X gate (CNOT)
                      {"cx", Gates::cx},   // Controlled-X gate (CNOT)
                      {"cy", Gates::cy},   // Controlled-Y gate
@@ -294,10 +294,10 @@ const stringmap_t<Gates>
                      {"ryy", Gates::ryy},   // Pauli-YY rotation gate
                      {"rzz", Gates::rzz},   // Pauli-ZZ rotation gate
                      {"rzx", Gates::rzx},   // Pauli-ZX rotation gate
-                     // Three-qubit gates
+                     /* Three-qubit gates */
                      {"ccx", Gates::ccx}, // Controlled-CX gate (Toffoli)
                      {"cswap", Gates::cswap},
-                     // Pauli
+                     /* Pauli */
                      {"pauli", Gates::pauli}});
 
 //=========================================================================
@@ -813,4 +813,4 @@ std::pair<uint_t, double> State::sample_measure_with_prob(const reg_t &qubits,
 //-------------------------------------------------------------------------
 } // end namespace AER
 //-------------------------------------------------------------------------
-#endif
+#endif
\ No newline at end of file
diff --git a/src/simulators/matrix_product_state/matrix_product_state_tensor.hpp b/home/mtreinish/git/qiskit/qiskit-aer/src/simulators/matrix_product_state/matrix_product_state_tensor.hpp
index f09dbe53..01559258 100644
--- a/src/simulators/matrix_product_state/matrix_product_state_tensor.hpp
+++ b/home/mtreinish/git/qiskit/qiskit-aer/src/simulators/matrix_product_state/matrix_product_state_tensor.hpp
@@ -556,9 +556,11 @@ void MPS_Tensor::contract_2_dimensions(const MPS_Tensor &left_gamma,
                                          (omp_threads > 1))                    \
     num_threads(omp_threads)
 #endif
-  for (int_t l_row = 0; l_row < left_rows; l_row++)
-    for (int_t r_col = 0; r_col < right_columns; r_col++)
+  for (int_t l_row = 0; l_row < left_rows; l_row++) {
+    for (int_t r_col = 0; r_col < right_columns; r_col++) {
       result(l_row, r_col) = 0;
+    }
+  }
 
 #ifdef _WIN32
 #pragma omp parallel for if ((omp_limit > MATRIX_OMP_THRESHOLD) &&             \
@@ -568,24 +570,23 @@ void MPS_Tensor::contract_2_dimensions(const MPS_Tensor &left_gamma,
                                          (omp_threads > 1))                    \
     num_threads(omp_threads)
 #endif
-  for (int_t l_row = 0; l_row < left_rows; l_row++)
+  for (int_t l_row = 0; l_row < left_rows; l_row++) {
     for (int_t r_col = 0; r_col < right_columns; r_col++) {
-
       for (int_t size = 0; size < left_size; size++)
         for (int_t index = 0; index < left_columns; index++) {
           result(l_row, r_col) += left_gamma.data_[size](l_row, index) *
                                   right_gamma.data_[size](index, r_col);
         }
     }
+  }
 }
 //---------------------------------------------------------------
 // function name: Decompose
-// Description: Decompose a tensor into two Gamma tensors and the Lambda between
-// 				them. Usually used after applying a 2-qubit
-// gate. Parameters: MPS_Tensor &temp - the tensor to decompose.
-//			   MPS_Tensor &left_gamma, &right_gamma , rvector_t &lambda
-//-
-// 			   tensors for the result.
+// Description: Decompose a tensor into two Gamma tensors and the
+//        Lambda between them. Usually used after applying a 2-qubit gate.
+// Parameters: MPS_Tensor &temp - the tensor to decompose.
+//			       MPS_Tensor &left_gamma, &right_gamma
+//             rvector_t &lambda - tensors for the result.
 // Returns: none.
 //---------------------------------------------------------------
 double MPS_Tensor::Decompose(MPS_Tensor &temp, MPS_Tensor &left_gamma,
diff --git a/src/simulators/statevector/chunk/cuStateVec_chunk_container.hpp b/home/mtreinish/git/qiskit/qiskit-aer/src/simulators/statevector/chunk/cuStateVec_chunk_container.hpp
index 88909884..3fd95b94 100644
--- a/src/simulators/statevector/chunk/cuStateVec_chunk_container.hpp
+++ b/home/mtreinish/git/qiskit/qiskit-aer/src/simulators/statevector/chunk/cuStateVec_chunk_container.hpp
@@ -29,11 +29,11 @@ namespace Chunk {
 template <typename data_t>
 class cuStateVecChunkContainer : public DeviceChunkContainer<data_t> {
 protected:
-  custatevecHandle_t
-      custatevec_handle_; // cuStatevec handle for this chunk container
+  custatevecHandle_t custatevec_handle_; // cuStatevec handle
+                                         // for this chunk container
   uint_t custatevec_chunk_total_qubits_; // total qubits of statevector passed
                                          // to ApplyMatrix
-  uint_t custatevec_chunk_count_; // number of counts for all chunks
+  uint_t custatevec_chunk_count_;        // number of counts for all chunks
 
   custatevecDeviceMemHandler_t custatevec_mem_handler_;
   cudaMemPool_t memory_pool_;
diff --git a/src/simulators/statevector/statevector_state.hpp b/home/mtreinish/git/qiskit/qiskit-aer/src/simulators/statevector/statevector_state.hpp
index 8b9ff80a..405c1423 100644
--- a/src/simulators/statevector/statevector_state.hpp
+++ b/home/mtreinish/git/qiskit/qiskit-aer/src/simulators/statevector/statevector_state.hpp
@@ -347,18 +347,18 @@ const stringmap_t<Gates> State<statevec_t>::gateset_(
      {"p", Gates::mcp},       // Parameterized phase gate
      {"sx", Gates::mcsx},     // Sqrt(X) gate
      {"sxdg", Gates::mcsxdg}, // Inverse Sqrt(X) gate
-     // 1-qubit rotation Gates
+     /* 1-qubit rotation Gates */
      {"r", Gates::mcr},   // R rotation gate
      {"rx", Gates::mcrx}, // Pauli-X rotation gate
      {"ry", Gates::mcry}, // Pauli-Y rotation gate
      {"rz", Gates::mcrz}, // Pauli-Z rotation gate
-     // Waltz Gates
+     /* Waltz Gates */
      {"u1", Gates::mcp},  // zero-X90 pulse waltz gate
      {"u2", Gates::mcu2}, // single-X90 pulse waltz gate
      {"u3", Gates::mcu3}, // two X90 pulse waltz gate
      {"u", Gates::mcu3},  // two X90 pulse waltz gate
      {"U", Gates::mcu3},  // two X90 pulse waltz gate
-     // 2-qubit gates
+     /* 2-qubit gates */
      {"CX", Gates::mcx},       // Controlled-X gate (CNOT)
      {"cx", Gates::mcx},       // Controlled-X gate (CNOT)
      {"cy", Gates::mcy},       // Controlled-Y gate
@@ -377,10 +377,10 @@ const stringmap_t<Gates> State<statevec_t>::gateset_(
      {"csx", Gates::mcsx},     // Controlled-Sqrt(X) gate
      {"csxdg", Gates::mcsxdg}, // Controlled-Sqrt(X)dg gate
      {"ecr", Gates::ecr},      // ECR Gate
-     // 3-qubit gates
+     /* 3-qubit gates */
      {"ccx", Gates::mcx},      // Controlled-CX gate (Toffoli)
      {"cswap", Gates::mcswap}, // Controlled SWAP gate (Fredkin)
-     // Multi-qubit controlled gates
+     /* Multi-qubit controlled gates */
      {"mcx", Gates::mcx},       // Multi-controlled-X gate
      {"mcy", Gates::mcy},       // Multi-controlled-Y gate
      {"mcz", Gates::mcz},       // Multi-controlled-Z gate
diff --git a/src/simulators/tensor_network/tensor_net.hpp b/home/mtreinish/git/qiskit/qiskit-aer/src/simulators/tensor_network/tensor_net.hpp
index 3f2cf4e6..f4f94343 100644
--- a/src/simulators/tensor_network/tensor_net.hpp
+++ b/home/mtreinish/git/qiskit/qiskit-aer/src/simulators/tensor_network/tensor_net.hpp
@@ -62,7 +62,7 @@ protected:
   int32_t mode_index_;                                   // index of modes
   std::vector<std::shared_ptr<Tensor<data_t>>> tensors_; // list of tensors
   std::vector<std::shared_ptr<Tensor<data_t>>> qubits_;  // tail tensor for
-                                                        // qubits
+                                                         // qubits
   std::vector<std::shared_ptr<Tensor<data_t>>>
       qubits_sp_;                        // tail tensor for super qubits
   std::vector<int32_t> modes_qubits_;    // tail mode index for qubits
diff --git a/src/simulators/tensor_network/tensor_net_state.hpp b/home/mtreinish/git/qiskit/qiskit-aer/src/simulators/tensor_network/tensor_net_state.hpp
index 72bcd2ef..44a6221c 100644
--- a/src/simulators/tensor_network/tensor_net_state.hpp
+++ b/home/mtreinish/git/qiskit/qiskit-aer/src/simulators/tensor_network/tensor_net_state.hpp
@@ -330,18 +330,18 @@ const stringmap_t<Gates> State<tensor_net_t>::gateset_(
      {"p", Gates::mcp},       // Parameterized phase gate
      {"sx", Gates::mcsx},     // Sqrt(X) gate
      {"sxdg", Gates::mcsxdg}, // Inverse Sqrt(X) gate
-     // 1-qubit rotation Gates
+     /* 1-qubit rotation Gates */
      {"r", Gates::mcr},   // R rotation gate
      {"rx", Gates::mcrx}, // Pauli-X rotation gate
      {"ry", Gates::mcry}, // Pauli-Y rotation gate
      {"rz", Gates::mcrz}, // Pauli-Z rotation gate
-     // Waltz Gates
+     /* Waltz Gates */
      {"u1", Gates::mcp},  // zero-X90 pulse waltz gate
      {"u2", Gates::mcu2}, // single-X90 pulse waltz gate
      {"u3", Gates::mcu3}, // two X90 pulse waltz gate
      {"u", Gates::mcu3},  // two X90 pulse waltz gate
      {"U", Gates::mcu3},  // two X90 pulse waltz gate
-     // 2-qubit gates
+     /* 2-qubit gates */
      {"CX", Gates::mcx},       // Controlled-X gate (CNOT)
      {"cx", Gates::mcx},       // Controlled-X gate (CNOT)
      {"cy", Gates::mcy},       // Controlled-Y gate
@@ -360,10 +360,10 @@ const stringmap_t<Gates> State<tensor_net_t>::gateset_(
      {"csx", Gates::mcsx},     // Controlled-Sqrt(X) gate
      {"csxdg", Gates::mcsxdg}, // Controlled-Sqrt(X)dg gate
      {"ecr", Gates::ecr},      // ECR Gate
-     // 3-qubit gates
+     /* 3-qubit gates */
      {"ccx", Gates::mcx},      // Controlled-CX gate (Toffoli)
      {"cswap", Gates::mcswap}, // Controlled SWAP gate (Fredkin)
-     // Multi-qubit controlled gates
+     /* Multi-qubit controlled gates */
      {"mcx", Gates::mcx},       // Multi-controlled-X gate
      {"mcy", Gates::mcy},       // Multi-controlled-Y gate
      {"mcz", Gates::mcz},       // Multi-controlled-Z gate

[hhorii]

@mtreinish thank you for your checking.

clang-format -i sometimes generates codes that will not be passed with clang-format -Werror -n. For example,

      {"cswap", Gates::mcswap}, // Controlled SWAP gate (Fredkin)
      // Multi-qubit controlled gates
      {"mcx", Gates::mcx},       // Multi-controlled-X gate

will warns as

src/simulators/tensor_network/tensor_net_state.hpp:365:65: error: code should be clang-formatted [-Wclang-format-violations]
     {"cswap", Gates::mcswap}, // Controlled SWAP gate (Fredkin)

. Therefore, I manually modified the codes after applying clang-format as

      {"cswap", Gates::mcswap}, // Controlled SWAP gate (Fredkin)
      /* Multi-qubit controlled gates */
      {"mcx", Gates::mcx},       // Multi-controlled-X gate

Another disruptive change is order of include files. clang-format sorts include statements and sometimes Aer headers are not self-contained (unfortunately). Then I added some include files or some namespace (such as complex_t -> AER::complex_t).

This diff accidentally comes from my old formatting trials (no RemoveBracesLLVM).

-  for (int_t l_row = 0; l_row < left_rows; l_row++)
-    for (int_t r_col = 0; r_col < right_columns; r_col++)
+  for (int_t l_row = 0; l_row < left_rows; l_row++) {
+    for (int_t r_col = 0; r_col < right_columns; r_col++) {
       result(l_row, r_col) = 0;
+    }
+  }

[mtreinish]

Ok, then given those caveats as things you manually edited I think we can go ahead and merge this. I'm mildly concerned about the fragility of clang format if it's auto-formatting isn't always compatible with it's warning check. But we can update our usage as things progress if we encounter issues.

[hhorii]

@mtreinish can we set this PR stable-backport-potential?