Em Chu

I took a look and haven't found a great solution to this. Options: - Make `@doc docstr f(x) = x` expand to something like ``` function f end m =...

Given that `@doc` can work with undefined things in a few special cases (see #53533: call expressions, random atoms, thismodule.x), we can't reliably return the expression value if we do.

Thanks for the help! > * Tests added to `parsestmt_with_kind_tests` in test/parser.jl to assert the correct kind for the begin/end leaf nodes in the tree ~~Am I correct in thinking...

> This distinction can be important as soon as we start work on unboxing captures I'll elaborate on my claim "we need something more complex anyway". Almost every lambda-vinfo-flags-based unboxing...

``` julia> @time @eval Base.cmp(1:30_000_000_000, 1:20_000_000_000) 7.477206 seconds (92 allocations: 3.609 KiB) ``` I would imagine there are ways this could be faster :)

Overlap is fine, but I still think it would be bothersome to get or set the numeric flags in macros accepting or producing those forms

Fair, it makes sense to wait until that's sorted out. Some examples of the current state of new macros are [here](https://github.com/c42f/JuliaLowering.jl/blob/main/test/macros.jl)