Yes, in fact the code works without the macro using the Tuple{...}
notation. The entire point of the macro is to make {...} = Tuple{...},
since in this use case it looks and feels very elegant.
I felt that new Julia users would not immediately "get" the difference
between Tuple types and tuples (instances), and I was hoping the package
would be ammenable to newcomers. Currently I have a macro that converts [[
... ]] into what I want (it's a "indexing via metaprogramming" technique
for tensor contractions/einsum... e.g. `A[Tuple{1,2}] = B[Tuple{2.1}]`
gives matrix transpose, written like `@tensor A[[1,2]] = B[[2,1]]` in my
current macro, and potentially as short as `A[{1,2}] = B[{2,1}]` in Julia
0.5 notation where {} might equal Tuple{}. It's true that it is longer than
`A = B.'`, but this generalizes nicely for arbitrarily complex tensor
contractions and permutations).
Andy
On Thursday, August 27, 2015 at 7:59:51 PM UTC+2, Yichao Yu wrote:
>
> On Thu, Aug 27, 2015 at 6:44 AM, Andy Ferris <[email protected]
> <javascript:>> wrote:
> > I'm writing a macro (for Julia 0.4) that is trying to take over the {}
> > syntax and do something special (related to tensor contraction) but I
> notice
> > the code to the macro go's through parse() (or similar) before being
> parsed
> > to my macro function, which means I get deprecation warning for {} as
> Julia
> > 0.3 meaning a constructor to Array{Any} but no longer for later
> versions.
> >
> > Is there a way to deactivate the deprecation warning, in general?
> >
> > Is there a way to deactivate the deprecation warning, just for a
> specific
> > macro? Is there any control macros have over parsing behaviour in
> general,
> > or could receive the raw text (without the end user needing to quote in
> a
> > string)?
> >
> > As a side note, if Jeff Bezanson's suggestion to use {} as Tuple{} in
> Julia
> > 0.5 is taken up, then I won't even need the macro, which is my
> motivation
> > for attempting to use this syntax now...
>
> I this case, why not just use `Tuple{}`. sth like `@braces_as_tuple
> {}` won't be so much easier to write then `Tuple{}` (even if your
> macro has a more elegant/shorter name), especially since I don't think
> `{}` get's it's own expression type in `0.4`.
>
