shallow() issues

[fowlmouth]

var x,y: seq[int]
y = @[]
shallow y

x = y
x.add 1

echo "x: ", x, ", y: ", y

output: x: @[1], y: @[1563515712, 0, 0, 2]

var x,y: string
y = ""
shallow y

x = y
x.add "42"

echo "x: ", x, ", y: ", y

output: x: 42, y:

The string test works fine when the code is in a block or the initial string has len > 0

[Araq]

Mostly a documentation bug.

[andreaferretti]

With latest Nim, the output is x: @[1], y: @[], consistent with the string case

[andreaferretti]

I think the current behaviour is correct, even though not very intuitive.

When y is assigned a literal, it becomes a sequence whose capacity is the same as its length. The copy x = y is actually shallow, but as soon as one adds an element to x, x is relocated and a copy is performed.

The following example, where y is created with enough capacity to start with, behaves as expected:

var x, y: seq[int]

y = newSeqOfCap[int](8)
y.add 1
y.add 2
y.add 3
shallow y

x = y
x.add 5

echo x # @[1, 2, 3, 5]
echo y # @[1, 2, 3, 5]

Similarly if one does not add to x, but replace existing elements:

var x, y: seq[int]

y = @[1, 2, 3]
shallow y

x = y
x[0] = 4

echo x # @[4, 2, 3]
echo y # @[4, 2, 3]

The inconsistency between strings and seqs does not happen anymore, as per my previous remark. I think this can be closed