Rand(-10..10)

Hi all,

wouldn't it make sense to allow

rand -10..10

be aquivalent to

-10 + rand(20)

the first is by far more obvious.
Of course rand -10...10 should work too.
What do you think?

···

--
Daniel

one solution could be

irb(main):210:0* class Range
irb(main):211:1> def rand
irb(main):212:2> return self.begin + Kernel.rand(self.end-self.begin)
irb(main):213:2> end
irb(main):214:1> end
=> nil
irb(main):215:0> (-10..10).rand
=> 2
irb(main):216:0> (-10..10).rand
=> 6
irb(main):217:0> (-10..10).rand
=> 4
irb(main):218:0> (-10..10).rand
=> -1

it's a matter of style to write

rand(-10..10) or (-10..10).rand

but since 10.sin doesn't work, as opposite tosin(10)
non object oriented rand(-10..10) is also a nice thing to have.

I had similar thoughts a while back. I decided that extending Kernel and mimicing the behavior of rand was the way to go. My code follows:

Regards, Morton

------ code starts
#!/usr/bin/ruby

# Extend Kernel with an enhanced uniform random number generator.
# Has the same behavior as rand for a single argument.
module Kernel
    URAND_ARG_ERR = "Arguments not valid for urand"
    # Return a pseudo-random number in the range 0.0...1.0, 0...m, or m..n.
    def urand(m=0, n=nil)
       case m
       when Range
          m, n = m.begin, m.end
       when Integer
          return rand(m) if n.nil?
       else
          raise ArgumentError, URAND_ARG_ERR
       end
       raise ArgumentError, URAND_ARG_ERR if n < m
       m + rand(n - m + 1)
    end
end

# Now for a little testing ...
def test(times, m=0, n=nil)
    r = Array.new(times)
    if n.nil? then
       if m == 0 then r.collect! {urand}
       else r.collect! {urand(m)}
       end
    else r.collect! {urand(m, n)}
    end
    p r
end
test(3) # => [0.0403243284672499, 0.875487065408379, 0.142408860381693]
test(3, 11, 11) # => [11, 11, 11]
test(10, 3) # => [0, 1, 2, 0, 1, 0, 1, 2, 0, 0]
test(10, -1..1) # => [0, 0, 1, -1, -1, 1, -1, 1, -1, 0]
test(10, 100, 200) # => [126, 148, 183, 140, 188, 175, 115, 157, 136, 179]
begin
    test(1, 3, -1)
rescue ArgumentError => error
    puts error.message
end # => Arguments not valid for urand

------ code ends

···

On Jul 26, 2006, at 8:50 PM, Schüle Daniel wrote:

one solution could be

irb(main):210:0* class Range
irb(main):211:1> def rand
irb(main):212:2> return self.begin + Kernel.rand(self.end-self.begin)
irb(main):213:2> end
irb(main):214:1> end
=> nil
irb(main):215:0> (-10..10).rand
=> 2
irb(main):216:0> (-10..10).rand
=> 6
irb(main):217:0> (-10..10).rand
=> 4
irb(main):218:0> (-10..10).rand
=> -1

it's a matter of style to write

rand(-10..10) or (-10..10).rand

but since 10.sin doesn't work, as opposite tosin(10)
non object oriented rand(-10..10) is also a nice thing to have.

irb(main):215:0> (-10..10).rand
=> 2
irb(main):216:0> (-10..10).rand
=> 6
irb(main):217:0> (-10..10).rand
=> 4
irb(main):218:0> (-10..10).rand
=> -1

*Sniff* God I love Ruby.

···

--
Posted via http://www.ruby-forum.com/\.

"the first is by far more obvious."

I think it looks ugly. :slight_smile:

(-10..10).rand
looks much nicer IMHO.

···

--
Posted via http://www.ruby-forum.com/.

Marc Heiler wrote:

"the first is by far more obvious."

I think it looks ugly. :slight_smile:

(-10..10).rand
looks much nicer IMHO.

This should work with all enumerables:

   module Enumerable
     def rand
       entries = entries
       entries.at(Kernel.rand(entries.length))
     end
   end

I can't seem to get around the `entries = entries' part -- I don't want to convert the enumerable to an array more than once. Any ideas?

Cheers,
Daniel

Daniel Schierbeck wrote:
...

This should work with all enumerables:

   module Enumerable
     def rand
       entries = entries
       entries.at(Kernel.rand(entries.length))
     end
   end

I can't seem to get around the `entries = entries' part -- I don't want
to convert the enumerable to an array more than once. Any ideas?

Enumerable supports sort_by, so could just use:

module Enumerable
  def rand
    self.sort_by{Kernel.rand}.last
  end
end

Cheers
Chris

ChrisH wrote:

Daniel Schierbeck wrote:
...

This should work with all enumerables:

   module Enumerable
     def rand
       entries = entries
       entries.at(Kernel.rand(entries.length))
     end
   end

I can't seem to get around the `entries = entries' part -- I don't want
to convert the enumerable to an array more than once. Any ideas?

Enumerable supports sort_by, so could just use:

module Enumerable
  def rand
    self.sort_by{Kernel.rand}.last
  end
end

I think sorting the entire enumerable may be slight overkill, although it of course is prettier :slight_smile:

Daniel

Daniel Schierbeck <daniel.schierbeck@gmail.com> writes:

This should work with all enumerables:

   module Enumerable
     def rand
       entries = entries
       entries.at(Kernel.rand(entries.length))
     end
   end

I can't seem to get around the `entries = entries' part -- I don't want
to convert the enumerable to an array more than once. Any ideas?

There's also this, which is possibly more memory-efficient (if the
Enumerable in question chews large amounts of memory when being
converted to an array, but not with each), but is almost certainly
much slower, since it calls rand once per element:

module Enumerable
  def rand
    ret, i = nil, 0
    each {|v| ret = v if 0 == Kernel.rand(i += 1)}
    ret
  end
end

Test code for irb:

h=Hash.new(0); 10000.times {h[%w{a b c d e}.rand] += 1}; h

Daniel Martin wrote:

module Enumerable
  def rand
    ret, i = nil, 0
    each {|v| ret = v if 0 == Kernel.rand(i += 1)}
    ret
  end
end

Test code for irb:

h=Hash.new(0); 10000.times {h[%w{a b c d e}.rand] += 1}; h
  

I think you shouldn't name this method rand. It shadows Kernel.rand in all classes, that already call rand and include Enumerable. The same is true for Range#rand, if people inherit from Range and already call rand. Ruby makes those things possible, but it's also very easy to shoot yourself or others into the foot, if you aren't careful.

The Kernel#rand(-5..15) version wouldn't suffer from this problem. It would be a good idea to add this to ruby core, because this functionality is something, that is needed very often.

···

--
Florian Frank

[...]

The Kernel#rand(-5..15) version wouldn't suffer from this problem. It would be a good idea to add this to ruby core, because this functionality is something, that is needed very often.

yes, this was exactly my point
if we already had Kernel#rand with Range object, people wouldn't implement their own methods, making scripts shoter and slightly more portable (regarding copy&paste other code)

Regards, Daniel