Add a python script to test the code quality of ABACUS codes

[mohanchen]

Add a python script to test the code quality of ABACUS codes

[AsTonyshment]

Keeping the source compatible with C++11 is a reasonable goal, especially for older HPC systems (even though the current CMake configuration uses at least C++14 when ENABLE_PEXSI, ENABLE_LIBRI, or BUILD_TESTING is enabled, when USE_CUDA is used with CUDA Toolkit < 13, or when 1.5.0 <= Torch_VERSION < 2.1.0; it uses at least C++17 when USE_CUDA is used with CUDA Toolkit >= 13, or when Torch_VERSION >= 2.1.0).

However, my coding agent tested the parser with many small, valid C++11 examples and found several reproducible false positives and missed cases. I have described these cases, together with a few suggestions of my own about the related C++ changes and scoring rules, in the inline comments. I hope they are helpful :)

[AsTonyshment]

This also matches class inside a template parameter list. Reproducer:

template <class T>
void f()
{
    int x = 0;
}

find_class_blocks() returns [(1, 5, "class", "T")], treating the function body as class T. Please distinguish class definitions from template parameters.

Furthermore, this also matches scoped enums. Reproducer:

enum class Kind
{
    first,
    second
};

find_class_blocks() returns a class block named Kind. Please exclude enum class before running class-member checks.

[AsTonyshment]

Strings are still present in stripped, so words in messages count as control flow. Reproducer:

void f()
{
    const char* text = "if if if if if if if if if if if";
}

The function is reported with complexity 11. Please remove string and character literals before applying CYCLO_KEYWORDS_RE.

[AsTonyshment]

This only accepts qualifiers followed directly by {, so valid C++11 trailing-return functions are missed. Reproducer:

auto f(int x) -> int
{
    if (x > 0)
    {
        return 1;
    }
    return 0;
}

find_function_bodies() returns an empty list. Please handle -> return_type before looking for the body.

Furthermore, constructor initializer lists are not handled. Reproducer:

A::A(int x) : x_(x)
{
    if (x > 0)
    {
        x_ = 0;
    }
}

The parser reports a function named x_ instead of constructor A::A. Please parse or skip the initializer list before finding the body.

[AsTonyshment]

Every ' starts character-literal mode here, including a digit separator. Reproducer:

int value = 1'000'000;

find_post_cpp11_features() returns no match, so the advertised C++14 digit-separator rule does not work. Please distinguish numeric separators from character literals.

[AsTonyshment]

A non-empty prefix is not enough to identify a declaration. Reproducer:

int f(int a, int b, int c, int d, int e, int f, int g, int h);

int g()
{
    return f(1, 2, 3, 4, 5, 6, 7, 8);
}

find_long_function_signatures() reports f twice. The second match is only a call in a return statement.

[AsTonyshment]

A continued type alias is accepted as a member variable. Reproducer:

struct A
{
    using value_type
        = std::vector<int>;
};

The result is public member in struct A: = std::vector<int>;. The same false positive occurs in sto_tool.h; please account for multi-line using declarations.

[AsTonyshment]

Counting braces on lines that still contain strings breaks the class depth. Reproducer:

class A
{
public:
    void f()
    {
        const char* text = "{";
    }

    int value;
};

analyze_class_blocks() does not find public member value. Please remove literals before counting braces.

[AsTonyshment]

The lookbehind excludes x.UPPER and Type::UPPER, but not ptr->UPPER. Reproducer:

value = ptr->UPPER_MEMBER;

UPPERCASE_CONST_RE matches UPPER_MEMBER, so equivalent member accesses receive different scores. Please exclude -> member access as well.

[AsTonyshment]

one and zero are only read by GEMM, so I think they do not need heap allocation. Using

const T one(1.0);
const T zero(0.0);

and then passing &one and &zero would be much simpler and would remain fully C++11-compatible.

[AsTonyshment]

I might suggest slightly changing the rule for public members in structs. ABACUS 开源项目 C++ 代码规范 says:

So I think most public data members (default behavior of struct) are intended design of these data-only structs. In my own opinion, it would be more useful to flag member functions in structs instead of their public data members.