[swift-users] Rational behind having 2 different behaviors for flatMap?

[Maxim Veksler]

Hi everyone,

I've discovered today that Swift will actually choose 2 very differently
behaving types of flatMap implementation based on the input signature.

For a Sequence of options it will call a flatMap that filters out nil's.
For a Sequence of Sequence's it will call a flattening function, without
filtering.

Leading to code that (IMHO) reads very not inconsistency, and unexpected.
Sometime even looking a bit funny such as collection.flatMap.flatMap:

  5> let deep = [["1989", nil], [nil, "Red"], [nil, nil]]
deep: [[String?]] = 3 values {
  [0] = 2 values {
    [0] = "1989"
    [1] = nil
  }
  [1] = 2 values {
    [0] = nil
    [1] = "Red"
  }
  [2] = 2 values {
    [0] = nil
    [1] = nil
  }
}
  6> deep.flatMap { $0 }
$R1: [String?] = 6 values {
  [0] = "1989"
  [1] = nil
  [2] = nil
  [3] = "Red"
  [4] = nil
  [5] = nil
}
  7> deep.flatMap { $0 }.flatMap { $0 }
$R2: [String] = 2 values {
  [0] = "1989"
  [1] = "Red"
}

I wonder why it was implemented this way?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-users/attachments/20170214/2a04ffbc/attachment.html>