Duplicated named arguments are ignored, but not rejected

[lewkok]

Hi,
Is example like:

#include <fmt/format.h>

int main() {
  using namespace fmt::literals;
  fmt::print("{abc} {1}", "abc"_a = 42, "abc"_a = 43 );
}

This prints: "42 43"

Or this example:

#include <fmt/format.h>

int main() {
  using namespace fmt::literals;
  fmt::print("{abc}", "abc"_a = 42, "abc"_a = 43 );
}

This prints: "42"

Is this semantics like this by design or did it happen by mistake?

[vitaut]

That's just an implementation artifact and I think we could add validation.

[dinomight]

Do you know what behavior you're expecting here? I started playing around with this idea a little bit and can get it to throw an exception at runtime if two named args have the same name. It doesn't seem like it can be caught at compile time without using static named args. Is that the expectation?

[vitaut]

Do you know what behavior you're expecting here?

Duplicate should give an exception.

[dinomight]

At the moment, I can verify the following:

fmt::print("{bar}", "bar"_a=42, "bar"_a=43);  // Fails to compile due to duplicate args.
fmt::print("{bar}", fmt::args("bar", 42), fmt::args("bar", 43));  // Throws exception at runtime due to duplicate args.

What might be helpful is some direction on the static compiled strings because the following still shows the existing behavior with my changes:

fmt::print(FMT_COMPILE("{bar}"), "bar"_a=42, "bar"_a=43);  // Compiles and outputs `42`.

I'm assuming if we can stop compile for the non-type template parameters, it should also work for compiled format strings?

Edit:
There's another issue tracking FMT_COMPILE: #4124 I would assume as that is resolved, perhaps it could handle this case as well.

[vitaut]

There's another issue tracking FMT_COMPILE: #4124 I would assume as that is resolved, perhaps it could handle this case as well.

Right.