Feature request: predicate on CHECK_THROWS about doctest

[cdeln]

Yes I think it is a good feature, and this issue clearly demonstrate the need for it.

It might take some time before that happens so I always try to document workarounds for other people to use before it's in the library.

[cdeln]

Can you write an example? It sounds like you have a custom exception class with some attached data that you want to verify.

[netheril96]

Yes, I have a custom exception with error_code method.

I'm thinking of either

const auto& e = CHECK_THROWS_AS(func(), const MyException&);
CHECK_EQ(e.error_code(), 3);

or

CHECK_THROWS_WITH_PREDICATE(func(), [](const auto& e) {
  const auto* p = dynamic_cast<const MyException*>(&e);
  CHECK(p != nullptr);
  CHECK_EQ(p->error_code(), 3);
});

[cdeln]

I like the idea of being able to further examine the exception.
If this is to be implemented I think the second version with CHECK_THROWS_WITH_PREDICATE is more viable since it's a macro and not a function call. Small improvement on how it would work below

CHECK_THROWS_WITH_PREDICATE(func(), MyException, [](auto& e) {
  CHECK_EQ(e.error_code(), 3);
});

Out of curiosity, is it possible for you to encode the error code using types instead? Or can the code be any number generated at runtime? If you have a finite number of error codes you can solve it with existing methods.

[netheril96]

If this is to be implemented I think the second version with CHECK_THROWS_WITH_PREDICATE is more viable since it's a macro and not a function call.

I think this can be worked around if CHECK_THROWS_AS expands as a lamdba call, e.g.

CHECK_THROWS_AS(func(), ExceptionType) => [&](){ MACRO_CONTENTS }();

Out of curiosity, is it possible for you to encode the error code using types instead? Or can the code be any number generated at runtime? If you have a finite number of error codes you can solve it with existing methods.

I have three major exceptions, PosixException with all possile errno codes, WindowsException with all possible GetLastError() codes, and NtException with all possible NTSTATUS values. It should be plain to you why your suggestion is infeasible in my case.

[cdeln]

I see, that makes this feature more important.
Here is a workaround draft

#define DOCTEST_CONFIG_IMPLEMENT_WITH_MAIN
#include <doctest/doctest.h>

struct PosixException : std::exception {
    int code;
    PosixException(int code) : code{code} {}
};

struct WindowsException : std::exception {
    int code;
    WindowsException(int code) : code{code} {}
};

#define REQUIRE_THROWS_WITH_PREDICATE(Expression, ExceptionType, Predicate) \
    try {                                                               \
        Expression ;                                                    \
    }                                                                   \
    catch (const ExceptionType& doctest_e) {                            \
        Predicate(doctest_e);                                           \
    }                                                                   \
    catch (...) {                                                       \
        FAIL("Unexpected exception, expected " #ExceptionType " but something else happened"); \
    }

void throwPosixException() {
    throw PosixException(1337);
}

TEST_CASE("846") {
    REQUIRE_THROWS_WITH_PREDICATE(throwPosixException(), PosixException, [](auto& e) {
        REQUIRE(e.code == 1337);
    });
    REQUIRE_THROWS_WITH_PREDICATE(throwPosixException(), WindowsException, [](auto &e) {
        REQUIRE(e.code == 1337);
    });
}

gives me

[doctest] doctest version is "2.4.11"
[doctest] run with "--help" for options
===============================================================================
main.cpp:29:
TEST CASE:  846

main.cpp:33: FATAL ERROR: Unexpected exception, expected WindowsException but something else happened

===============================================================================
[doctest] test cases: 1 | 0 passed | 1 failed | 0 skipped
[doctest] assertions: 2 | 1 passed | 1 failed |
[doctest] Status: FAILURE!

[netheril96]

That's very close to what I'm doing now, but it is better if it is in the library, isn't it?