Syntax error when extending glz::json_t with glm types

[exdal]

Hey, I've been trying to customize glz::json_t for my needs, I implemented exact copy of glz::json_t but with custom glz::json_t::val_t and here is how it looks:

using val_t = std::variant<null_t, std::string, bool, array_t, object_t, double, glm::vec2>;

// ...
// Meta impl. for my custom types, outside of 'MyCustomJson' struct
template<>
struct glz::meta<glm::vec2> {
    static constexpr std::string_view name = "glm::vec2";
    using T = glm::vec2;
    static constexpr auto value = object(&T::x, &T::y);
};

For some reason, variant_is_auto_deducible always evaluates to false thus causing a weird error like ec: syntax_error, location: 0. I need to use glz::json_t because the data I am trying to read is not predictable (it's a scene deserialization with ECS stuff). Are there any ways to make glm::vec2 differ from, in my guess, array_t or object_t?

[exdal]

Exact location of where syntax error is set at:

glaze/include/glaze/json/read.hpp

Line 447 in 6d249bf

match<"null", Opts>(ctx, it, end);

it argument is a string at position 0 of my json string.

[stephenberry]

The parsing is failing there because the variant isn't auto-deducible and thus defaults to null_t and tries to parse into this null_t, which is why it is erroring in the specialization for always_null_t.

You're correct that your meta for glm::vec2 is conflicting with object_t, since both are objects (meta<glm::vec2>: object(&T::x, &T::y);).

The easiest solution is probably to just use glz::json_t and then you can call if (value.contains("x") && value.contains("y")) this should indicate that you have a glm::vec2. There is an active issue open to support reading a glz::json_t value into a concrete C++ struct here. This would be useful for you as well. Currently you would have to write the value to JSON and then read it into a concrete glm::vec2.

I also think I should add a dump function to glz::json_t so that once you find your target value you can easily dump it to a string.

[exdal]

Thank you, I will be following that issue.