Variadic (or "Varargs") destructuring of tuples

[habemus-papadum]

In julia, is there some syntax similar to
(x,z...) = (1,2,3) which would result in
x=1 & z=(2,3)?

If not, are there reasons why this was not supported, or perhaps this could be a reasonable feature request (or, more constructively, prior to working on a PR, do others have thoughts about this)?

Ideally, since nested destructuring e.g. (a,(b,c)) = (1,(2,3)) works as expected, any variadic version would also work e.g.
(a,(b,c,d...),e...) = (1,(2,3,4,5),6,7) would yield

a=1
b=2
c=3
d=(4,5)
e=(6,7)

And perhaps

(a,(b,c,d...),e..., nexttolast, last) = (1,(2,3,4,5),6,7,8,9,10) would yield

a=1
b=2
c=3
d=(4,5)
e=(6,7,8)
nexttolast=9
last=10

While (a,(b,c,d...),e..., nexttolast, last) = (1,(2,3,4,5),9,10) would yield

a=1
b=2
c=3
d=(4,5)
e=()
nexttolast=9
last=10

And finally (a,(b,c,d...),e..., nexttolast, last) = (1,(2,3,4,5),10) would result in an error, just
like the current behavior for (a,b) = 1

[mschauer]

The package https://github.com/kmsquire/Match.jl allows to do this

> a, b, c = @match (1,2,(3,)) (a,b, (c,))
(1, 2, 3)

[jiahao]

This is a usage question for the julia-users mailing list. The issue tracker is for bug reports.

[timholy]

You could interpret this as a feature request for such syntax.

@lilinjn, tuplesplit(x) = x[1], tuple(x[2:end]...) should get you started. To me that looks like attractive syntax, so if you can figure out how to add it to the parser, a PR might be nice. (Just my opinion.)

[StefanKarpinski]

@jiahao, the topmost syntax doesn't work:

julia> (x,z...) = (1,2,3)
ERROR: syntax: invalid assignment location

I'm pretty sure I've proposed this before and I think we can reasonably view this issue as a feature request for this syntax.

[ivarne]

Implementing this can also prepare the way for deprecating the sometimes surprising behaviour where a, b = 1,2,3 discards the third value.

[StefanKarpinski]

I'm torn about discarding trailing values. It makes it possible to have silent bugs where you drop values that you didn't mean to, but on the other hand, it allows gracefully extending APIs by returning additional values without breaking old call sites, which is a very handy property.

[jiahao]

Sorry, misread.

[habemus-papadum]

Thanks for the feedback -- I should have been more clear that I was seeking comments prior to working on a potential PR (as well as making sure that I wasn't missing some alternate syntax).

[StefanKarpinski]

No worries – I thought it was clear enough. A PR would be excellent and most appreciated!

[JeffBezanson]

duplicate of #2626

[StefanKarpinski]

See? I knew it was a good idea.