Macro to add coalesce(x, false) to all if and && and || in a block

[pdeffebach]

Should this live in Missings.jl? Potentially cleaner and easier to maintain than a skipmissing keyword argument in more functions.

Note that it uses postwalk so @mfalse can act recursively on nested expressions.

julia> using MacroTools

julia> function mfalse_helper(ex::Expr)
           MacroTools.postwalk(ex) do x
               if x isa Expr && x.head == :if 
                   x.args[1] = wrap_coalesce(x.args[1])
               elseif x isa Expr && x.head == :&&
                   x.args = wrap_coalesce.(x.args)
               elseif x isa Expr && x.head == :|| 
                   x.args = wrap_coalesce.(x.args)
               end
               x
           end  
       end;

julia> wrap_coalesce(x::Expr) = :(coalesce($x, false));

julia> wrap_coalesce(x::Symbol) = :(coalesce($x, false));

julia> wrap_coalesce(x) = x;

julia> macro mfalse(x)
           esc(mfalse_helper(x))
       end;

julia> x = missing;

julia> x == 1 ? 2 : 3
ERROR: TypeError: non-boolean (Missing) used in boolean context
Stacktrace:
 [1] top-level scope
   @ REPL[21]:1

julia> @mfalse x == 1 ? 2 : 3
3

Now that .|| and .&& are in Base, we can also account for these in the macro.

[pdeffebach]

This was inspired by JuliaLang/julia#40729, which introduces a @coalesce macro. So maybe this can be added to Base?

It's also reminiscent of some of the three-valued logic in Volcanito.jl.

[nalimilan]

Interesting. Do you plan to use this in DataFramesMeta? I think it would be useful to consider practical use cases first.

[pdeffebach]

Yes, I'm thinking of something like

@subset df @byrow :x == 1 || :x == 2

which would fail if :x is missing, even if we did subset(...; skipmissing = true).

@subset df @byrow @mfalse :x == 1 || :x == 2

would work.