@edit a[1] no longer works

[KristofferC]

@edit a[1] = 1 still works. Seems to be because of:

julia> Meta.lower(@__MODULE__, :(a[1] = 1))
:(begin
      setindex!(a, 1, 1)
      return 1
  end)

julia> Meta.lower(@__MODULE__, :(a[1])) # Wat
:($(Expr(:thunk, CodeInfo(:(begin
      SSAValue(0) = getindex(a, 1)
      return SSAValue(0)
  end)))))

[JeffBezanson]

Calling lower is probably no longer a good idea here. Note that the setindex case works because the code for @edit painstakingly looks for that exact result from lower. I think it will be easier to manually write cases for syntax forms like these that have a (theoretically) simple correspondence with function calls.

[KristofferC]

FWIW, I think it has been there since the feature originally got introduced, 5 years ago:85fdc92#diff-c20eb2eead14027fab4293e10bacf73bR110

[vtjnash]

I think we could still call Meta.lower, but then scan the results looking for 1. the call argument to return (following SSAValues) 2. that all other operations are in some whitelist (conditionals, variables, calls, constants, labels, metadata, etc.)

[tkoolen]

In similar vein, @edit [3], @edit [3 4], and @edit [3; 4] are broken.

Is there a way to be sure that all of such cases are handled? #28122 is great, but I'm afraid this is a game of whack-a-mole.

Also, I have a package that was relying on expand (now Meta.lower, but the output is vastly different) to avoid having special cases for all of these. Could this be reinstated in some form or another?

[JeffBezanson]

Some of these are fixed but setindex! is still broken.