Feature #16155: Add an Array#intersection method

• Status: Closed
• Assignee: -
• Target version: -

[connorshea (Connor Shea)]

Array#union and Array#difference were added in Ruby 2.6 (see this bug), but an equivalent for & (intersection) was not.

I'd like to propose Array#intersection. This would essentially just be a more readable alias for Array#&, in the same way that Array#| and Array#- have Array#union and Array#difference.

I think it'd make sense for Ruby to have a more readable name for this method :)

Current syntax:

[ 1, 1, 3, 5 ] & [ 3, 2, 1 ]                 #=> [ 1, 3 ]
[ 'a', 'b', 'b', 'z' ] & [ 'a', 'b', 'c' ]   #=> [ 'a', 'b' ]

What I'd like to see added:

[ 1, 1, 3, 5 ].intersection([ 3, 2, 1 ])                #=> [ 1, 3 ]
[ 'a', 'b', 'b', 'z' ].intersection([ 'a', 'b', 'c' ])  #=> [ 'a', 'b' ]

mame asks about intersection in this comment on the union/difference bug, but as far as I can tell it was never addressed.

Set#intersection already exists and is an alias for Set#&, so there's precedent for such a method to exist.

Thanks for Ruby, I enjoy using it a lot! :)

Related links:

• PR for union
• PR for difference
• Bug for adding union and difference methods

[shevegen (Robert A. Heiler)]

I sort of agree with your reasoning. A slight add-on, though,
on that part:

This would essentially just be a more readable alias for Array#&

It is probably easier to search for it (e. g. a google-search for
.intersection) than for &. But if we compare relative merits then
we also have to remember an advantage for & being that it is
short/succinct.

It's an aside, though, because I agree with your reasoning anyway;
it would make sense to also have .intersection, IMO. :)

You could consider adding this to the upcoming developer meeting:

https://bugs.ruby-lang.org/issues/16152

And ask matz whether he would be fine with the functionality
itself; or, more importantly, the name for the method
(#intersection) since the functionality already exists.

[connorshea (Connor Shea)]

shevegen (Robert A. Heiler) wrote:

You could consider adding this to the upcoming developer meeting:

https://bugs.ruby-lang.org/issues/16152

And ask matz whether he would be fine with the functionality
itself; or, more importantly, the name for the method
(#intersection) since the functionality already exists.

Thanks, I will :)

After thinking about the proposal a bit more, one other thing I wanted to ask was about whether we should allow multiple arrays to be passed as arguments. Array#difference and Array#union both allow this, so it might make sense to have Array#intersection also take multiple arguments.

For example:

# intersection with multiple arguments.
[ 'a', 'b', 'b', 'z' ].intersection([ 'a', 'b', 'c' ], [ 'b' ])  #=> [ 'b' ]

Should we keep intersection as a simple alias for & or have it allow multiple arrays, to match the difference and union methods?

[phluid61 (Matthew Kerwin)]

connorshea (Connor Shea) wrote:

Should we keep intersection as a simple alias for & or have it allow multiple arrays, to match the difference and union methods?

The question should be: is it needed?

[matz (Yukihiro Matsumoto)]

Accepted. PR welcome.

Matz.

[prajjwal (Prajjwal Singh)]

Hi all,

I've added an initial implementation of this based on Array#& on github (https://github.com/ruby/ruby/pull/2533).

One thing that stands out to me is the naming inconsistency between rb_ary_and and rb_ary_intersection_multi, and I think either could be changed to be consistent with the other. Thoughts?

[prajjwal (Prajjwal Singh)]

Implementation is currently based on Array#&, which is elegant but might end up allocating a whole bunch of arrays holding intermediate results. If needed I can implement Array#intersection so it only allocates the result array once, but then I would like to rewrite Array#& in terms of Array#intersection to keep things DRY.

static VALUE
rb_ary_intersection_multi(int argc, VALUE *argv, VALUE ary)
{
    VALUE result = rb_ary_dup(ary);
    int i;

    for (i = 0; i < argc; i++) {
        result = rb_ary_and(result, argv[i]);
    }

    return result;
}

Let me know what you think.

[nobu (Nobuyoshi Nakada)]

• Status changed from Open to Closed