From: Yukihiro Matsumoto <matz@...>
Date: 2008-11-04T15:38:50+09:00
Subject: [ruby-core:19691] Re: Odd TypeError in inject (1.9.1 preview 1)

Hi,

In message "Re: [ruby-core:19683] Re: Odd TypeError in inject (1.9.1 preview 1)"
    on Mon, 3 Nov 2008 23:36:38 +0900, "David A. Black" <dblack@rubypal.com> writes:

|I think I was starting from the 1.8 behavior, and not grasping the
|thing about the magic symbol-test. So I expected it to return 0, and
|it was the symbol complaint that originally struck me as odd (and I'm
|still not sure it makes complete sense to me).

Indeed 1.8.6 returns 0 for [].inject(0), but it's kinda odd corner
case.  1.8.6 inject is inconsistent.

                1.8.6     1.8.7      1.9.1
 [].inject(0)     0       error      error
 [1].inject(0)  error     error	    error
 [].inject       nil      error      error
 [1].inject       1       error	    error

I think the behavior of #inject should be more consistent than 1.8.6.
The possible alternative behavior is

 [].inject(0)   enumerator
 [1].inject(0)  enumerator
 [].inject      enumerator
 [1].inject     enumerator

I am not sure how much it is useful.

|On the general subject of #inject and enumerators: what about this?
|
|>> a
|=> [1, 2, 3, 4]
|>> e = a.enum_for(:inject, Set.new)
|=> #<Enumerator:0x458f08>
|>> e.each {|s,i| s << i }
|=> #<Set: {1, 2, 3, 4}>
|
|This #each does return a value from the block, because it's really a
|front-end to #inject. Or am I misunderstanding what you're saying
|about #inject not working with enumerators?

I meant generator like behavior introduced in 1.9.  #next receives a
value from the iterator but there's no way to give a value back from.

							matz.