Hi Mauro,
2016-06-22 21:13 GMT+02:00 Mauro:
>
> A REPL session of mine trying to figure out how to make that macro would
> look something like this:
>
> julia> ex = :(f(x, y, z, zz)) # this is what is passed into the macro
> :(f(x,y,z,zz))
>
> julia> xdump(ex) # xdump is nice to get an overview of nested
> datastructures
> Expr
> head: Symbol call
> args: Array(Any,(5,))
> 1: Symbol f
> 2: Symbol x
> 3: Symbol y
> 4: Symbol z
> 5: Symbol zz
> typ: Any::DataType <: Any
>
I knew and used "dump" but not "xdump". I've seen that the latter has been
deprecated in favor of the former in Julia 0.5.
> julia> target = :(f([x.re, y.re, z.re, zz.re].^2 + [x.im, y.im, z.im,
> zz.im])) # this is your target
> :(f([x.re,y.re,z.re,zz.re] .^ 2 + [x.im,y.im,z.im,zz.im]))
>
> julia> xdump(target)
> Expr
> head: Symbol call
> args: Array(Any,(2,))
> 1: Symbol f
> 2: Expr
> head: Symbol call
> args: Array(Any,(3,))
> 1: Symbol +
> 2: Expr
> head: Symbol call
> args: Array(Any,(3,))
> typ: Any::DataType <: Any
> 3: Expr
> head: Symbol vect
> args: Array(Any,(4,))
> typ: Any::DataType <: Any
> typ: Any::DataType <: Any
> typ: Any::DataType <: Any
>
>
> Now, writing the macro is really an exercise in manipulating nested data
> structures. With the added bonus that there are convenient constructors
> for those datastructures, namely expression such as e.g. :(z = 4 +7).
>
> julia> fn = ex.args[1] # extract the name of the function
> :f
>
> julia> args = ex.args[2:end] # its arguments
> 4-element Array{Any,1}:
> :x
> :y
> :z
> :zz
>
> julia> ar1 =:([]) # build up the first array
> :([])
>
> julia> ar1.head # check
> :vect
>
> julia> [push!(ar1.args, :($(args[i]).re)) for i=1:length(args)] #add the
> elements
> 4-element Array{Any,1}:
> Any[:(x.re),:(y.re),:(z.re),:(zz.re)]
> Any[:(x.re),:(y.re),:(z.re),:(zz.re)]
> Any[:(x.re),:(y.re),:(z.re),:(zz.re)]
> Any[:(x.re),:(y.re),:(z.re),:(zz.re)]
>
> julia> ar1
> :([x.re,y.re,z.re,zz.re])
>
> julia> :($ar1^2) # square
> :([x.re,y.re,z.re,zz.re] ^ 2)
>
> etc.
>
> It takes a bit of practice and patience. Important, as Kristoffer said,
> write out and test the code you expect the macro to form (at least until
> you become a macro ninja), then write the macro. For instance in one of
> my projects I hand wrote all of the code I later built macros to
> generate:
> https://github.com/mauro3/Traits.jl/blob/master/test/manual-traitdef.jl
>
Thank you so much, I've been able to change my "@uncertain" macro for
Measurements.jl <https://github.com/giordano/Measurements.jl> package
following your hints. In the end, the trick was to quote rather than to
evaluate the array of arguments. Probably this is a good rule of thumb
when writing macros.
Bye,
Mosè
