Reduce in Paginate produces an array inside an array

[aprilmintacpineda]

Reduce(
  Lambda(
    ['accumulator', 'ref'],
    Append(
      Get(Var('ref')),
      Var('accumulator')
    )
  ),
  [],
  Paginate(Documents(Collection('notifications')), {
    size: 2
  })
)

Will result to:

{
  after: [Ref(Collection("notifications"), "296811454570430977")],
  data: [
    [
      {
        ref: Ref(Collection("notifications"), "295665276532294149"),
        ts: 1619949211300000,
        data: {
          createdAt: "2021-04-12T11:33:50.342615Z",
          updatedAt: "2021-05-02T09:53:31.133472Z"
        }
      },
      {
        ref: Ref(Collection("notifications"), "295666276580196871"),
        ts: 1619943100860000,
        data: {
          createdAt: "2021-04-12T11:49:44.073589Z",
          updatedAt: "2021-05-02T08:11:40.677455Z"
        }
      }
    ]
  ]
}

Notice how data is an array containing an array.

The expected output is:

{
  after: [Ref(Collection("notifications"), "296811454570430977")],
  data: [
      {
        ref: Ref(Collection("notifications"), "295665276532294149"),
        ts: 1619949211300000,
        data: {
          createdAt: "2021-04-12T11:33:50.342615Z",
          updatedAt: "2021-05-02T09:53:31.133472Z"
        }
      },
      {
        ref: Ref(Collection("notifications"), "295666276580196871"),
        ts: 1619943100860000,
        data: {
          createdAt: "2021-04-12T11:49:44.073589Z",
          updatedAt: "2021-05-02T08:11:40.677455Z"
        }
      }
    ]
}

The reason I’m using Reduce is that in my app I’m doing Map and then Filter, I can combine those two loops by doing reduce but the output is unexpected.

[ewan]

It looks like the accumulator gets “upgraded” to a Page, since that’s what Paginate returns. That’s why the after cursor is part of the result.

The simplest change to make would be to use Select:

Reduce(
  Lambda(
    ['accumulator', 'ref'],
    Append(
      Get(Var('ref')),
      Var('accumulator')
    )
  ),
  [],
  Select(
    ["data"],
    Paginate(Documents(Collection('notifications')), { size: 2})
  )
)

Doing that drops the after cursor, so if you need to iterate over multiple pages, your original query would be better.

[aprilmintacpineda]

Hmm, then I think Reduce does not work well with Paginate, I think an update to make it work well with Paginate will be great, and if using it with Paginate, I should not be required to provide an initialValue because the Reduce should default to using the data in Page as the initialValue.

[ptpaterson]

+1 to having a separate Fold operation from Reduce. The current FQL Reduce does work more like a Fold should, requiring the initial value. It could be useful to have an operation start accumulating on the first element of the provided Page/Array.

Note that the “upgrade to Page” is very intentional. If your into functional programming, think of a Page as a Monad – The Page “knows” how to perform Map/Filter/Reduce on the data field while preserving the pagination cursors, which is really cool IMHO.

But yea… the result you show is a bit clunky, though it is appropriate the way that Reduce is defined. The most appropriate functional-programming thing to remove the extra array layer would probably be to wrap the result in a Flatten or FlatMap type call, but neither of those exist in FQL.

If you need to the pagination cursors, it looks like the best thing is to Select(['data', 0], ...) from the whole result where you need it. If you don’t need the cursors, then Ewan’s example is a good one.

[aprilmintacpineda]

In my case, I really needed the cursor so I ended up doing Select(['data', 0], ...) which made me sad.

Fold is a pretty clever name haha I like it, especially it operates on Page, that’s pretty cool, hoping for it to be implemented soon!