# Idiom I've not seen before **URL:** \ **Category:** ruby-talk\ **Created:** [7 February 2008 03:55 UTC](https://rubytalk.org/t/idiom-ive-not-seen-before/44331 "2008-02-07T03:55:07Z")\ **Posts on this page:** 12\ **Page:** 1
**Author:** ![Rob\_Saul](https://avatars.discourse-cdn.com/v4/letter/r/7993a0/32.png) [@Rob\_Saul](https://rubytalk.org/u/Rob_Saul)\ **Post date:** [7 February 2008 03:55 UTC](https://rubytalk.org/t/idiom-ive-not-seen-before/44331/1 "2008-02-07T03:55:07Z")
I came across this :   elements.sort\_by(&:position).each in something I've taken over. I thought sort\_by took a block. Obviously I'm missing something. Can anyone shed some light on this trick?   ~Rob ---
**Author:** ![Rob\_Biedenharn1](https://avatars.discourse-cdn.com/v4/letter/r/57b2e6/32.png) [@Rob\_Biedenharn1](https://rubytalk.org/u/Rob_Biedenharn1)\ **Post date:** [7 February 2008 04:04 UTC](https://rubytalk.org/t/idiom-ive-not-seen-before/44331/2 "2008-02-07T04:04:06Z")
You've found Symbol#to\_proc That line is equivalent to: elements.sort\_by {|e| e.send(:position) }.each The "something" is a Rails project, isn't it? == vendor/rails/activesupport/lib/active\_support/core\_ext/symbol.rb == class Symbol    # Turns the symbol into a simple proc, which is especially useful for enumerations. Examples: > **···** > > On Feb 6, 2008, at 10:55 PM, Rob Saul wrote: > > > I came across this : > >   elements.sort\_by(&:position).each > > in something I've taken over. I thought sort\_by > > took a block. Obviously I'm missing something. > > Can anyone shed some light on this trick? > > > >   ~Rob > >    # >    # # The same as people.collect { |p| p.name } >    # people.collect(&:name) >    # >    # # The same as people.select { |p| p.manager? }.collect { |p| p.salary } >    # people.select(&:manager?).collect(&:salary) >    def to\_proc >      Proc.new { |\*args| args.shift.\_\_send\_\_(self, \*args) } >    end > end > > I don't know how you can send additional arguments however. > > -Rob > > Rob Biedenharn [http://agileconsultingllc.com](http://agileconsultingllc.com) > Rob@AgileConsultingLLC.com ---
**Author:** ![Jeremy\_Kemper1](https://avatars.discourse-cdn.com/v4/letter/j/c6cbf5/32.png) [@Jeremy\_Kemper1](https://rubytalk.org/u/Jeremy_Kemper1)\ **Post date:** [7 February 2008 04:06 UTC](https://rubytalk.org/t/idiom-ive-not-seen-before/44331/3 "2008-02-07T04:06:50Z")
Sure. &:position == :position.to\_proc == proc { |element, \*args| element.position(\*args) } So elements.sort\_by(&:position) is shorthand for sorting the elements by position. jeremy > **···** > > On 2/6/08, Rob Saul \ wrote: > > > I came across this : > >         elements.sort\_by(&:position).each > > in something I've taken over. I thought sort\_by > > took a block. Obviously I'm missing something. > > Can anyone shed some light on this trick? ---
**Author:** ![Gavin\_Kistner3](https://avatars.discourse-cdn.com/v4/letter/g/dfb087/32.png) [@Gavin\_Kistner3](https://rubytalk.org/u/Gavin_Kistner3)\ **Post date:** [7 February 2008 17:59 UTC](https://rubytalk.org/t/idiom-ive-not-seen-before/44331/4 "2008-02-07T17:59:58Z")
Loosely-related aside: I've been learning Io[1] recently. Io (like Lisp, I gather) allows lazy evaluation of method arguments. In Ruby terms, this would mean that I can write something like this:   # Ruby-esque pseudo-code; neither Io nor Ruby   class Array     def select       result =       self.each{ |el| result \<\< el if el.sendArg(0) }       result     end   end   my\_array.select( isCool? )   my\_array.select( \> 5 )   my\_array.select( roughlyEquals( jim.newSize ) ) In Io, the "isCool?" method/message isn't (necessarily) evaluated when you call the select method. Instead, you can perform some introspection on the parsed message tree for each argument and choose to ignore it, change it to a string, or send it as a message to any object you want. To be clear, in the above, the "isCool?" message/method would be send/ invoked on each array element. Or each element would be sent a "\>( 5 )" message. The Symbol#to\_proc technique is clever, but not quite as clean as being able to write (in Io):   elements sortBy( position )   elements map( \* 2 )   elements select( size \> 5 ) This same ability in Io allowed me to add a debugging method "p" (in homage of Ruby) that labels a value with the exact call you made. For example:   # In Io with my custom method (not Ruby)   p( gk )   #=\> gk is Person\_0x4de0d8:   #=\> name = "Gavin Kistner"   #=\> nick = "Phrogz""   p( gk name )   #=\> gk name is Gavin Kistner   p( gk nick size )   #=\> gk nick size is 6 [1] [http://www.iolanguage.com/](http://www.iolanguage.com/) > **···** > > On Feb 6, 8:55 pm, Rob Saul \ wrote: > > > I came across this : > >         elements.sort\_by(&:position).each ---
**Author:** ![Rob\_Saul](https://avatars.discourse-cdn.com/v4/letter/r/7993a0/32.png) [@Rob\_Saul](https://rubytalk.org/u/Rob_Saul)\ **Post date:** [7 February 2008 04:14 UTC](https://rubytalk.org/t/idiom-ive-not-seen-before/44331/5 "2008-02-07T04:14:24Z")
Jeremy Kemper wrote: > Sure. &:position == :position.to\_proc == proc { |element, \*args| > element.position(\*args) } > > So elements.sort\_by(&:position) is shorthand for sorting the elements > by position. > > jeremy thanks to both Jeremy and Rob for the enlightenment. And yes, it is a Rails project. ---
**Author:** ![Ryan\_Davis1](https://yyz1.discourse-cdn.com/flex029/user_avatar/rubytalk.org/ryan_davis1/32/1848_2.png) [@Ryan\_Davis1](https://rubytalk.org/u/Ryan_Davis1)\ **Post date:** [7 February 2008 05:16 UTC](https://rubytalk.org/t/idiom-ive-not-seen-before/44331/6 "2008-02-07T05:16:14Z")
Buyer beware: #!/usr/local/bin/ruby -w # # of iterations = 10000 # user system total real # null\_time 0.000000 0.000000 0.000000 ( 0.001561) # map 0.880000 0.000000 0.880000 ( 0.878500) # to\_proc 2.790000 0.000000 2.790000 ( 2.799291) # # of iterations = 100000 # user system total real # null\_time 0.020000 0.000000 0.020000 ( 0.013951) # map 8.730000 0.010000 8.740000 ( 8.766693) # to\_proc 27.960000 0.030000 27.990000 ( 28.001288) require 'benchmark' class Symbol    def to\_proc      Proc.new { |\*args| args.shift.\_\_send\_\_(self, \*args) }    end end a = (1..100).to\_a raise "stupid" unless a.map { |n| n.to\_s } == a.map(&:to\_s) max = (ARGV.shift || 1\_000\_000).to\_i puts "# of iterations = #{max}" Benchmark::bm(20) do |x|    x.report("null\_time") do      for i in 0..max do        # do nothing      end    end    x.report("map") do      for i in 0..max do        a.map { |n| n.to\_s }      end    end    x.report("to\_proc") do      for i in 0..max do        a.map(&:to\_s)      end    end end > **···** > > On Feb 6, 2008, at 20:04 , Rob Biedenharn wrote: > > > You've found Symbol#to\_proc > > > > That line is equivalent to: > > > > elements.sort\_by {|e| e.send(:position) }.each ---
**Author:** ![Clifford\_Heath5](https://avatars.discourse-cdn.com/v4/letter/c/7ab992/32.png) [@Clifford\_Heath5](https://rubytalk.org/u/Clifford_Heath5)\ **Post date:** [7 February 2008 22:39 UTC](https://rubytalk.org/t/idiom-ive-not-seen-before/44331/7 "2008-02-07T22:39:56Z")
Phrogz wrote: > I've been learning Io[1] recently. Io (like Lisp, I gather) allows > lazy evaluation of method arguments. ... > In Io, the "isCool?" method/message isn't (necessarily) evaluated when > you call the select method. This is known as "call by name", as opposed to by reference or by value. It was, more than anything, the single thing that caused the most difficulty for the authors of Algol68 optimising compilers :-). Niklaus Wirt once joked that Europeans (who know how to pronounce his name) called him by name, whereas Americans called him by value (nickel's worth). Maybe apocryphal, but funny :-). Clifford Heath. ---
**Author:** ![Deepak\_Gole](https://avatars.discourse-cdn.com/v4/letter/d/e274bd/32.png) [@Deepak\_Gole](https://rubytalk.org/u/Deepak_Gole)\ **Post date:** [8 February 2008 11:33 UTC](https://rubytalk.org/t/idiom-ive-not-seen-before/44331/8 "2008-02-08T11:33:35Z")
Hello Rob   We can send arguments, Following is the way, But it throws an error.   def to\_proc     Proc.new { |\*args| args.shift.\_\_send\_\_(self, \*args) }   end   arr.collect(&[:concat,"Rocking"]) \* \*sample.rb:41:in `\_\_send\_\_': [:concat, "dee"] is not a symbol (TypeError)     from sample.rb:41:in `to\_proc'     from sample.rb:51:in `collect'     from sample.rb:51 But if we change the implementation of to\_proc then it works.    def to\_proc       lambda {|target| target.send \*self}    end    arr.collect(&[:concat,"Rocking"]) o/p =\> ["rubyRocking", "railsRocking", "rjaxRocking"] I don't know how. > **···** > > On Feb 6, 2008 8:04 PM, Rob Biedenharn \ wrote: > > > On Feb 6, 2008, at 10:55 PM, Rob Saul wrote: > > \> I came across this : > > \> elements.sort\_by(&:position).each > > \> in something I've taken over. I thought sort\_by > > \> took a block. Obviously I'm missing something. > > \> Can anyone shed some light on this trick? > > \> > > \> ~Rob > > > > You've found Symbol#to\_proc > > > > That line is equivalent to: > > > > elements.sort\_by {|e| e.send(:position) }.each > > > > The "something" is a Rails project, isn't it? > > > > == vendor/rails/activesupport/lib/active\_support/core\_ext/symbol.rb == > > class Symbol > >   # Turns the symbol into a simple proc, which is especially useful > > for enumerations. Examples: > >   # > >   # # The same as people.collect { |p| p.name } > >   # people.collect(&:name) > >   # > >   # # The same as people.select { |p| p.manager? }.collect { |p| > > p.salary } > >   # people.select(&:manager?).collect(&:salary) > >   def to\_proc > >     Proc.new { |\*args| args.shift.\_\_send\_\_(self, \*args) } > >   end > > end > > > > I don't know how you can send additional arguments however. > > > > -Rob > > > > Rob Biedenharn [http://agileconsultingllc.com](http://agileconsultingllc.com) > > Rob@AgileConsultingLLC.com ---
**Author:** ![ThoML](https://avatars.discourse-cdn.com/v4/letter/t/d78d45/32.png) [@ThoML](https://rubytalk.org/u/ThoML)\ **Post date:** [8 February 2008 14:14 UTC](https://rubytalk.org/t/idiom-ive-not-seen-before/44331/9 "2008-02-08T14:14:59Z")
> Io (like Lisp, I gather) allows lazy evaluation of method arguments. Are you referring to Lisp's macro facility? AFAIK macros are usually expanded at load/compile time. Since everything is a s-expression in lisp, the macro gets its arguments as list and can process them. The macro call is then replaced with the result. Or something along this line. It seems Io provides a similar facility to treat code as data, which is cool. Unfortunately I'm not able to compile it. IIRC Forth was able to do something similar too BTW (just a side-note). Since ruby doesn't have this feature (unless you use parsetree and ruby2ruby maybe), the ruby equivalent are strings, I'd say. One could thus write something like:     class Array         def select\_by(snippet)             rv =             self.each do |e|                 rv \<\< e if e.instance\_eval '%s %s' % [e.inspect, snippet]             end             rv         end     end     a = [1,2,3,4,5]     a.select\_by '\> 3' Which is about the same as your io example. It would be nice of course to be able to have real macros that are expanded at load time so that one could write something like the following which is incorrect ruby syntax and would have to be rewritten before the parser tries to make sense out of it:     module Macros         def select\_by(code)             %{select {|e| e #{code}}}         end     end     a = [1,2,3,4,5]     a.select\_by(\> 3) This would require that macro names are unique though. But it's just a delirious day-dreaming anyway. Regards, Thomas. ---
**Author:** ![Martin\_DeMello](https://avatars.discourse-cdn.com/v4/letter/m/f17d59/32.png) [@Martin\_DeMello](https://rubytalk.org/u/Martin_DeMello)\ **Post date:** [7 February 2008 05:04 UTC](https://rubytalk.org/t/idiom-ive-not-seen-before/44331/10 "2008-02-07T05:04:37Z")
Also, I believe Symbol.to\_proc is making it into 1.9 martin > **···** > > On Feb 7, 2008 4:14 AM, Rob Saul \ wrote: > > > thanks to both Jeremy and Rob for the enlightenment. And > > yes, it is a Rails project. ---
**Author:** ![botp1](https://avatars.discourse-cdn.com/v4/letter/b/6de8d8/32.png) [@botp1](https://rubytalk.org/u/botp1)\ **Post date:** [7 February 2008 05:35 UTC](https://rubytalk.org/t/idiom-ive-not-seen-before/44331/11 "2008-02-07T05:35:18Z")
that will change in ruby1.9 since to\_proc is builtin. eg, a run in windows, C:\ruby1.9\bin\>.\ruby.exe test.rb 1\_000 # of iterations = 1000                           user system total real null\_time 0.000000 0.000000 0.000000 ( 0.000000) map 0.203000 0.000000 0.203000 ( 0.203000) to\_proc 0.156000 0.000000 0.156000 ( 0.172000) C:\ruby1.9\bin\>.\ruby.exe test.rb 10\_000 # of iterations = 10000                           user system total real null\_time 0.016000 0.000000 0.016000 ( 0.015000) map 1.937000 0.000000 1.937000 ( 2.515000) to\_proc 1.610000 0.000000 1.610000 ( 1.954000) C:\ruby1.9\bin\>.\ruby.exe test.rb 100\_000 # of iterations = 100000                           user system total real null\_time 0.328000 0.000000 0.328000 ( 0.328000) map 19.344000 0.000000 19.344000 ( 24.953000) to\_proc 17.296000 0.015000 17.311000 ( 18.016000) not bad, imho. kind regards -botp > **···** > > On Feb 7, 2008 1:16 PM, Ryan Davis \ wrote: > > > Buyer beware: > > > > # # of iterations = 10000 > > # user system total real > > # null\_time 0.000000 0.000000 0.000000 ( 0.001561) > > # map 0.880000 0.000000 0.880000 ( 0.878500) > > # to\_proc 2.790000 0.000000 2.790000 ( 2.799291) > > > > # # of iterations = 100000 > > # user system total real > > # null\_time 0.020000 0.000000 0.020000 ( 0.013951) > > # map 8.730000 0.010000 8.740000 ( 8.766693) > > # to\_proc 27.960000 0.030000 27.990000 ( 28.001288) ---
**Author:** ![Thomas\_Wieczorek](https://avatars.discourse-cdn.com/v4/letter/t/e47c2d/32.png) [@Thomas\_Wieczorek](https://rubytalk.org/u/Thomas_Wieczorek)\ **Post date:** [7 February 2008 05:13 UTC](https://rubytalk.org/t/idiom-ive-not-seen-before/44331/12 "2008-02-07T05:13:43Z")
Yes, it is in. Raganwald has a good blog entry about it, he also mentions String#to\_proc: [http://weblog.raganwald.com/2007/11/fun-with-symboltoproc.html](http://weblog.raganwald.com/2007/11/fun-with-symboltoproc.html) > **···** > > On Thu, Feb 7, 2008 at 6:04 AM, Martin DeMello \ wrote: > > > Also, I believe Symbol.to\_proc is making it into 1.9