compiler: eliminate Core.Compiler. qualifier
#52790
Conversation
| @@ -331,7 +331,7 @@ end | |||
| add_tfunc(Core.ifelse, 3, 3, ifelse_tfunc, 1) | |||
|
|
|||
| @nospecs function ifelse_nothrow(π::AbstractLattice, cond, x, y) | |||
| β = Core.Compiler.:β(π) | |||
| β = partialorder(π) | |||
There was a problem hiding this comment.
FWIW, the Compiler.:β is okay, just not Core to make it global. Or this very annoying syntax with let global to rebind it:
julia> using Core.Compiler: β
julia> f(π) = (β = let; global β; β(π); end)
f (generic function with 1 method)
julia> f(π)
There was a problem hiding this comment.
I think whether β = partialorder(π) is clearer than β = Compiler.:β(π). In the same vein, I'm also thinking about replacing tmerge with β = join(π). What do you think?
There was a problem hiding this comment.
That seems fine too. Though maybe a bit confusing that it is not an optimal join on the π (that would be Union), but rather on the complexity co-lattice
There was a problem hiding this comment.
Got it. Then, we might consider refactoring to switch to the 2-arg form using partialorder(π) or widening(π) (or weakjoin(π)?). This could turn into a significant refactor, so it might be best to undertake it after finishing the upcoming expected refactors.
Anyway, I would like to move this PR forward for now. @Keno please let me know if you have a preference for the current 3-arg form.
To avoid symbol name clashes, this commit defines and uses
partialorder(π::AbstractLattice) = β(π). While I have a preference for a 2-arg binary operator for lattice operations, it's currently coexisting with a 3-arg version, suggesting that a unification might be beneficial.