sed -i 's/\(\bfn .*) \)%\(.*{\)$/\1!\2/g' $(find . -name "*.zig")


the literal translation of `%T` to this new code is `error!T`.
however this would not take advantage of error sets. It's
recommended to generally have all your functions which return possible
errors to use error set inference, like this:

fn foo() !void {

}

then you can return void, or any error, and the error set is inferred.


you can get the compiler to tell you the possible errors for an inferred error set like this:

foo() catch |err| switch (err) {};

// TODO this is an explicit cast and should actually coerce the type
   erorr set casting
   // add a runtime safety check


test err should be comptime if error set has 0 members

comptime calling fn with inferred error set should give empty error set but still you can use try

comptime err to int of empty err set and of size 1 err set

comptime test for err


undefined in infer error 

syntax -  ?a!b  should be ?(a!b) but it's (?a)!b

syntax - (error{}!void) as the return type


passing a fn()error{}!T to a fn()error!T should be a compile error, they're not compatible