proposal: spec: "x..." in slice literal (syntactic sugar)

[dolmen]

We already have a syntax to expand a slice to give as an argument to a variadic function: slice.... Using the same syntax in slice literals would be a consistent syntactic sugar.

v := []string{"c", "d"}
w := []string{"a", "b", v...}

Would expand to:

v := []string{"c", "d"}
w := make([]string, 2+len(v))
w[0] = "a"
w[1] = "b"
copy(w[2:], v)

This slice expansion should work at any place (not just the tail like for slice expansion in calls to variadic functions in Go 1.0) in the literal. Related: #18605.

Use case: building a new slice by prepending elements to a given slice is quite common when wrapping variadic functions such as fmt.Sprintf.

[yaxinlx]

v := []string{"c", "d"}
w := append([]string{"a", "b"}, v...)

[mvdan]

Why isn't #18605 enough? This would complicate the language and would be a much bigger change to it.

[griesemer]

This is what append is for (see @yaxinlx comment). Or put in another way: If we had v... for composite literals, we wouldn't need append: You could always (*) write

[]T{list..., x}      // same as append(list, x)
[]T{list..., x...}   // same as append(list, x...)

(*) assuming we know the type T.

The above requires ... to work at any place in a composite literal, not just the end. Having it only at the end smells even more like append.

In other words, we already have the functionality you wish for, perhaps a bit less elegantly, but more explicitly, with append.

I am against this proposal.

[dolmen]

I don't want this feature anymore. @yaxinlx and @griesemer convinced me.