# Justification for Array#join behaviour? **URL:** \ **Category:** ruby-talk\ **Created:** [14 February 2007 07:52 UTC](https://rubytalk.org/t/justification-for-array-join-behaviour/35347 "2007-02-14T07:52:00Z")\ **Posts on this page:** 9\ **Page:** 1
**Author:** ![Farrel\_Lifson](https://avatars.discourse-cdn.com/v4/letter/f/3da27b/32.png) [@Farrel\_Lifson](https://rubytalk.org/u/Farrel_Lifson)\ **Post date:** [14 February 2007 07:52 UTC](https://rubytalk.org/t/justification-for-array-join-behaviour/35347/1 "2007-02-14T07:52:00Z")
What is the justification for [1,2,[3,4]].join(',') being '1,2,3,4' instead of '1,2,34'? According to [ruby-doc.org](http://ruby-doc.org) join "Returns a string created by converting each element of the array to a string, separated by sep.". Why should an element be recursively joined if it's also an array? Farrel ---
**Author:** ![Aur\_Saraf](https://avatars.discourse-cdn.com/v4/letter/a/bcef8e/32.png) [@Aur\_Saraf](https://rubytalk.org/u/Aur_Saraf)\ **Post date:** [14 February 2007 08:07 UTC](https://rubytalk.org/t/justification-for-array-join-behaviour/35347/2 "2007-02-14T08:07:27Z")
I second that. > **···** > > On 2/14/07, Farrel Lifson \ wrote: > > > What is the justification for [1,2,[3,4]].join(',') being '1,2,3,4' > > instead of '1,2,34'? According to ruby-doc.org join "Returns a string > > created by converting each element of the array to a string, separated > > by sep.". Why should an element be recursively joined if it's also an > > array? > > > > Farrel ---
**Author:** ![Gavin\_Kistner2](https://avatars.discourse-cdn.com/v4/letter/g/b9e5f3/32.png) [@Gavin\_Kistner2](https://rubytalk.org/u/Gavin_Kistner2)\ **Post date:** [14 February 2007 15:35 UTC](https://rubytalk.org/t/justification-for-array-join-behaviour/35347/3 "2007-02-14T15:35:05Z")
What's the justification for it NOT doing that? (Just because the docs say so?) I suspect that it's more like the join method #flatten s the array before joining, instead of recursing. But my C isn't good enough to tell what rb\_ary\_join is really doing inside. I'm not sure I can think of a real use case for either scenario. The only one I can make up is where you want to override Array#to\_s, where you'd want the last line below... irb(main):001:0\> class Foo; def to\_s;"SSS";end; def inspect;"III";end; end =\> nil irb(main):002:0\> a = [1, 2, Foo.new, [ 3, Foo.new, [ 4, 5] ] ] =\> [1, 2, III, [3, III, [4, 5]]] irb(main):003:0\> a.join( ',' ) =\> "1,2,SSS,3,SSS,4,5" irb(main):005:0\> class Array; def to\_s; "\"; end; end =\> nil irb(main):006:0\> a.to\_s =\> "\" irb(main):008:0\> a.join( ',' ) =\> "1,2,SSS,3,SSS,4,5" ...to return "1,2,SSS,\" instead. Hrm...yeah, I think I'd prefer that simpler and more powerful implementation. Of course, if you really wanted it, you could change it yourself: irb(main):009:0\> class Array irb(main):010:1\> def join( sep='' ) irb(main):011:2\> output = "" irb(main):012:2\> each\_with\_index do |v,i| irb(main):013:3\* output \<\< v.to\_s irb(main):014:3\> output \<\< sep unless i == (self.length-1) irb(main):015:3\> end irb(main):016:2\> output irb(main):017:2\> end irb(main):018:1\> end =\> nil irb(main):019:0\> a.join( ',' ) =\> "1,2,SSS,\" ...but (barring someone else's good explanation) I'd rather see the implementation changed to match the documentation (and force you to use a.flatten.join(',') if that's what you want) than the reverse. > **···** > > On Feb 14, 12:52 am, "Farrel Lifson" \ wrote: > > > What is the justification for [1,2,[3,4]].join(',') being '1,2,3,4' > > instead of '1,2,34'? According to ruby-doc.org join "Returns a string > > created by converting each element of the array to a string, separated > > by sep.". Why should an element be recursively joined if it's also an > > array? ---
**Author:** ![Wolfgang\_Nadasi-Don1](https://avatars.discourse-cdn.com/v4/letter/w/958977/32.png) [@Wolfgang\_Nadasi-Don1](https://rubytalk.org/u/Wolfgang_Nadasi-Don1)\ **Post date:** [14 February 2007 18:05 UTC](https://rubytalk.org/t/justification-for-array-join-behaviour/35347/4 "2007-02-14T18:05:07Z")
Farrel Lifson schrieb: > What is the justification for [1,2,[3,4]].join(',') being '1,2,3,4' > instead of '1,2,34'? According to ruby-doc.org join "Returns a string > created by converting each element of the array to a string, separated > by sep.". Why should an element be recursively joined if it's also an > array? > > Farrel It is easy to come to the required result by irb(main):001:0\> [1,2,[3,4]].map{|e|e.to\_s}.join(',') =\> "1,2,34" ...but I think this should be the "natural" way of handling it by "join". Implicit recursion (or "flatten" in this special case) can be confusing in some situation. Wolfgang Nádasi-Donner ---
**Author:** ![Bertram\_Scharpf](https://avatars.discourse-cdn.com/v4/letter/b/43a26b/32.png) [@Bertram\_Scharpf](https://rubytalk.org/u/Bertram_Scharpf)\ **Post date:** [14 February 2007 18:49 UTC](https://rubytalk.org/t/justification-for-array-join-behaviour/35347/5 "2007-02-14T18:49:42Z")
Hi, > What is the justification for [1,2,[3,4]].join(',') being '1,2,3,4' > instead of '1,2,34'? According to ruby-doc.org join "Returns a string > created by converting each element of the array to a string, separated > by sep.". Why should an element be recursively joined if it's also an > array? You would have to ask for the objects type:   class Array     def deep\_join sep       map { |x|         case x           when Array then x.deep\_join sep           else x.to\_s         end       }.join sep     end   end I don't look at this as a smart programming style actually. The object-oriented approach is to redefine `to\_s'. Bertram > **···** > > Am Mittwoch, 14. Feb 2007, 16:52:00 +0900 schrieb Farrel Lifson: > > -- > Bertram Scharpf > Stuttgart, Deutschland/Germany > [http://www.bertram-scharpf.de](http://www.bertram-scharpf.de) ---
**Author:** ![Gary\_Wright](https://avatars.discourse-cdn.com/v4/letter/g/58956e/32.png) [@Gary\_Wright](https://rubytalk.org/u/Gary_Wright)\ **Post date:** [14 February 2007 20:22 UTC](https://rubytalk.org/t/justification-for-array-join-behaviour/35347/6 "2007-02-14T20:22:37Z")
Be aware that Array#to\_s has changed in Ruby 1.9 $ irb-1.9 irb(main):001:0\> a = [1,2,[3,4]] =\> [1, 2, [3, 4]] irb(main):002:0\> a.to\_s =\> "[1, 2, [3, 4]]" irb(main):003:0\> a.join =\> "1234" irb(main):004:0\> The issue is that Array#to\_s defers to Array#inspect in 1.9 instead of simply concatenating the results of #to\_s on all the elements. I didn't expect this change either and matz pointed me to Array#join Check out this thread: ruby-talk:220698 Gary Wright > **···** > > On Feb 14, 2007, at 2:52 AM, Farrel Lifson wrote: > > > What is the justification for [1,2,[3,4]].join(',') being '1,2,3,4' > > instead of '1,2,34'? According to ruby-doc.org join "Returns a string > > created by converting each element of the array to a string, separated > > by sep.". Why should an element be recursively joined if it's also an > > array? ---
**Author:** ![Farrel\_Lifson](https://avatars.discourse-cdn.com/v4/letter/f/3da27b/32.png) [@Farrel\_Lifson](https://rubytalk.org/u/Farrel_Lifson)\ **Post date:** [14 February 2007 19:26 UTC](https://rubytalk.org/t/justification-for-array-join-behaviour/35347/7 "2007-02-14T19:26:48Z")
I did some experimentation and it seems that the decision might not be that half bad. Take for instance a deeply nested array [1,2,[3,4[5,6]]]. Currently calling join produces "1,2,3,4,5,6". Whereas if we used the non-recursive join we'd get "1,2,3456" which actually looks a bit worse. I think it would be better if the documentation was amended to reflect that nested arrays will be flattened. Farrel > **···** > > On 14/02/07, Bertram Scharpf \ wrote: > > > Hi, > > > > Am Mittwoch, 14. Feb 2007, 16:52:00 +0900 schrieb Farrel Lifson: > > \> What is the justification for [1,2,[3,4]].join(',') being '1,2,3,4' > > \> instead of '1,2,34'? According to ruby-doc.org join "Returns a string > > \> created by converting each element of the array to a string, separated > > \> by sep.". Why should an element be recursively joined if it's also an > > \> array? > > > > You would have to ask for the objects type: > > > >   class Array > >     def deep\_join sep > >       map { |x| > >         case x > >           when Array then x.deep\_join sep > >           else x.to\_s > >         end > >       }.join sep > >     end > >   end > > > > I don't look at this as a smart programming style actually. > > > > The object-oriented approach is to redefine `to\_s'. > > > > Bertram > > > > -- > > Bertram Scharpf > > Stuttgart, Deutschland/Germany > > [http://www.bertram-scharpf.de](http://www.bertram-scharpf.de) ---
**Author:** ![Martin\_DeMello](https://avatars.discourse-cdn.com/v4/letter/m/f17d59/32.png) [@Martin\_DeMello](https://rubytalk.org/u/Martin_DeMello)\ **Post date:** [15 February 2007 07:14 UTC](https://rubytalk.org/t/justification-for-array-join-behaviour/35347/8 "2007-02-15T07:14:17Z")
Or class Object   def join     to\_s   end end module Enumerable   def join     "[" + inject {|a, e| a.join + ", " + e.join} + "]"   end end class Array   def join     super   end end class String   def join     self   end end puts [1, 2, 3, [4, 5, [6, 7], 8]].join martin > **···** > > On 2/15/07, Bertram Scharpf \ wrote: > > >   class Array > >     def deep\_join sep > >       map { |x| > >         case x > >           when Array then x.deep\_join sep > >           else x.to\_s > >         end > >       }.join sep > >     end > >   end > > > > I don't look at this as a smart programming style actually. > > > > The object-oriented approach is to redefine `to\_s'. ---
**Author:** ![Bertram\_Scharpf](https://avatars.discourse-cdn.com/v4/letter/b/43a26b/32.png) [@Bertram\_Scharpf](https://rubytalk.org/u/Bertram_Scharpf)\ **Post date:** [15 February 2007 11:37 UTC](https://rubytalk.org/t/justification-for-array-join-behaviour/35347/9 "2007-02-15T11:37:48Z")
Hi, > **···** > > Am Donnerstag, 15. Feb 2007, 16:14:17 +0900 schrieb Martin DeMello: > > > On 2/15/07, Bertram Scharpf \ wrote: > > \>The object-oriented approach is to redefine `to\_s'. > > > > Or > > > > [...] > > puts [1, 2, 3, [4, 5, [6, 7], 8]].join > > [1, 2, 3, [4, 5, [6, 7], 8]] > > Ah, yes. It's the open and close delimiters () what is > getting lost by the OP's proposal. I should have seen that > right away. > > Bertram > > -- > Bertram Scharpf > Stuttgart, Deutschland/Germany > [http://www.bertram-scharpf.de](http://www.bertram-scharpf.de)