Proposal: give functions with inferred error sets a more sane type name

[NicoElbers]

Zig Version

0.14.0-dev.1509+fb81522e0

Steps to Reproduce and Observed Behavior

foo.zig:

pub fn foo() !void {}

pub fn main() !void {
    const T = @TypeOf(foo);
    const name = @typeName(T);

    @import("std").debug.print("name: {s}\n", .{name});
}

Observe:

❯ zig run foo.zig
name: fn () @typeInfo(@typeInfo(@TypeOf(foo.foo)).@"fn".return_type.?).error_union.error_set!void

This name, albeit correct, does not give me as a programmer any information about the function. If I want to know the returned errors I would have to do more digging, and if I want to log type names for whatever reason, it causes useless unreadable clutter in the output.

Expected Behavior

I would expect either fn () !void or fn () error{}!void.

Changing foo.zig to the following:

pub fn foo() error{}!void {}

pub fn main() !void {
    const T = @TypeOf(foo);
    const name = @typeName(T);

    @import("std").debug.print("name: {s}\n", .{name});
}

Produces:

❯ zig run foo.zig
name: fn () error{}!void

Which seems much more reasonable to me. Especially since this lines up with the function signature.

[NicoElbers]

It seems the long name is hard coded here, I would personally say this name, albeit correct, is a little useless and causes for clutter. But I guess this changes this issue from a bug to a proposal.

[nektro]

#19425 has some previous discussion