@stable module A
using DispatchDoctor: @unstable
@unstable f1() = rand(Bool) ? 0 : 1.0
f2(x) = x
f3(; a=1) = a > 0 ? a : 0.0
end
where we use @unstable to mark functions that should not be wrapped.
(Tip: in the REPL, wrap this with @eval, because the REPL has special handling of the module keyword.)
This gives us:
julia> A.f1()
0
julia> A.f2(1.0)
1.0
julia> A.f3(a=2)
ERROR: TypeInstabilityError: Instability detected in function `f3`
with keyword arguments `@NamedTuple{a::Int64}`. Inferred to be
`Union{Float64, Int64}`, which is not a concrete type.
where we can see that the @stable was automatically applied
to all the functions, except for f1.
[!NOTE]
This will also automatically be applied to the code
being added by any include within the module, by defining
a custom include function at the top of the module.
For example:
where we use
@unstableto mark functions that should not be wrapped.(Tip: in the REPL, wrap this with
@eval, because the REPL has special handling of themodulekeyword.)This gives us:
where we can see that the
@stablewas automatically applied to all the functions, except forf1.