Hello! first post to clr. I'm asking about an attempt at a lazy ruby solution to computing fibonacci numbers for a project euler problem. seems to be a bug in lazy ruby

ruby-talk
1

Hi, first post to clr here.

I have a guess a haskell bias as that's the last language I learned,
and now I'm learning ruby.

I was trying to solve project euler problem 2 as an exercise. This is,
sum all the even fibonacci numbers less than 4 million.

There is a solution that works at

http://www.absorbeo.net/2008/01/10/project-euler-problem-2/

however I didn't quite like it because the test for evenness is in the
fold (haskell speak for inject).

puts a.inject(0) {|s,v|
  if v > 1_000_000 then
    break s
  elsif v % 2 == 0
      s+v
  else
    s
  end
}

the haskell solution reads nicer to me

t = ( putStrLn . show . sum . takeWhile (<=4000000) . filter even )
fibs2
fibs2 = 1 : 2 : zipWith (+) fibs2 (tail fibs2)

but it relies on laziness and who knows what other magic.

I decided to see if I could find a ruby solution that was more
haskellish. It turns I can... almost.

There is a ruby library for laziness

# http://lazylist.rubyforge.org/

and it would seem to do what I want. But it hangs for some reason.

I have poor sense of ruby culture. Should I report this as a bug? Is
it even a bug? Do people ever use laziness, or is this crazy
experimentation?

Thanks for help and advice!

Code below:

thartman@thartman-laptop:~/thomashartman-learning/ruby/katas/euler>cat
2.rb
require 'rubygems'
require 'lazylist' # http://lazylist.rubyforge.org/

# Solve Project Euler problem 2: sum fibonacci numbers <= 4_000_000,
which are even.
fibs = ( LazyList.tabulate(0) { |x| x < 2 ? 1 : fibs[x-2] +
fibs[x-1] } )

fibsUnder4M = ( fibs.partition{ |x| x <= 4_000_000 })[0]
evenFibsUnder4M = fibsUnder4M.select{ |x| x %2 == 0}

# problem seems to be with partition
# evenFibsLessThanFourMil = ( evenFibs.partition{ |x| x <= 4_000_000 })
[0]

# works
# puts evenFibsLessThanFourMil.take(1)

# hangs
puts evenFibsUnder4M.take(10) # this is ok
puts evenFibsUnder4M.take(11) # this hangs, laptop gets hot, fan turns
on...
# exit

# this would be the answer to the euler problem, if only there wasn't
the bug...
result = evenFibsLessThanFourMil.inject(0){ | accum,val | accum + val}

main = puts result
thartman@thartman-laptop:~/thomashartman-learning/ruby/katas/

ruby 2.rb

2
8
34
144
610
2584
10946
46368
196418
832040
   ..... hung...

2

t = ( putStrLn . show . sum . takeWhile (<=4000000) . filter even )
fibs2
fibs2 = 1 : 2 : zipWith (+) fibs2 (tail fibs2)

Gives me hives personally... I've never been able to handle the syntax of haskell.

but it relies on laziness and who knows what other magic.

The laziness I'm fine with... the other magic? not so much.

I decided to see if I could find a ruby solution that was more
haskellish. It turns I can... almost.

Ignoring my haskell bias for a bit... I'd recommend doing this, with any pairing of languages. Find the idiomatic response appropriate for the target language and you'll have a better time.

There is a ruby library for laziness

# http://lazylist.rubyforge.org/

I've not used this personally, I suspect using it is the problem.

and it would seem to do what I want. But it hangs for some reason.

I have poor sense of ruby culture. Should I report this as a bug? Is
it even a bug? Do people ever use laziness, or is this crazy
experimentation?

Assuming it is a bug, yes, report it. Again, not having used the library, I can't really weigh in whether this behavior is a bug or not.

require 'rubygems'
require 'lazylist' # http://lazylist.rubyforge.org/

# Solve Project Euler problem 2: sum fibonacci numbers <= 4_000_000,
which are even.
fibs = ( LazyList.tabulate(0) { |x| x < 2 ? 1 : fibs[x-2] +
fibs[x-1] } )

so, according to the doco, tabulate is a generator function that starts at 0 and yields the block on each successive succ.

(stop with all the parens)

fibsUnder4M = ( fibs.partition{ |x| x <= 4_000_000 })[0]

partition is customized to return two lazy lists, so this _should_ be fine.

evenFibsUnder4M = fibsUnder4M.select{ |x| x %2 == 0}

likewise, select is customized to return a lazy list. again, fine.

# hangs
puts evenFibsUnder4M.take(10) # this is ok
puts evenFibsUnder4M.take(11) # this hangs, laptop gets hot, fan turns
on...

presumably the take actually executes and generates elements from the lazy list, unwinding back through the select, the partition, and finally the tabulate block. This all looks kosher to me on the surface (albeit a bit more complicated than necessary because you're trying to match haskell.

So, I would think that you've discovered a bug in the LazyList impl, as your logic seems sound.

Here is my ruby-idiomatic solution to the problem:

# Solve Project Euler problem 2:
# sum fibonacci numbers <= 4_000_000, which are even.

$fib = {} # simple memoization
def fib x
  $fib ||= x < 2 ? 1 : fib(x-2) + fib(x-1)
end

# fib 33 > 4m
fibs = (1..32).map { |n| fib n }.reject { |n| n % 2 == 1 }

sum = 0
fibs.each do |n|
  sum += n
end

p sum
# => 4613732

because of the memoization, it runs incredibly fast. If I wasn't allowed to predetermine the upper bound of the calculation, I'd wrap that up in a simple loop with a conditional... Nothing fancy.

···

On Aug 6, 2008, at 08:40 , tphyahoo wrote:

3

It is implementation failure, you are right.
LazyList::Enumerable has buggy select which works as follows:

So it search first true occurence and returns new object.

list = LazyList.tabulate(0) { |x| x }
=> [... ]
less_than_three = list.select { |x| x < 3 }
lest_than_three.take(0);lest_than_three.take(1);lest_than_three.take(2)
less_than_three.tail
=> [1, 2,... ]
less_than_three.tail.tail
=> [2,... ]
less_than_three.tail.tail.tail
# Hangs

So this should return lazy tail (which is empty, infinite loop occurs).

By the way, you are hunting ants with an rpg:P

Best Regards

Maciej

···

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

4

Personally, I like a Ruby memoizing solution, but seeing some other
lazy Ruby projects in a quick Google search, I whipped this up:

require 'lazy_stream'

even = lambda{|int| int[0] == 0}
under_4_million = lambda{|int| int <= 4_000_000 }

fibonacci = lambda{|array, index|
  index < 2 ? 1 : array[index - 2] + array[index - 1]
}

# create a "lazy stream" that generates the fibonacci sequence
fibs = LazyStream.new(fibonacci)

# create a filtered lazy stream of even fibonacci numbers
even_fibs = LazyStream.filter(fibs, even)

# sum the even fibonacci numbers that are less than 4,000,000
sum = 0
even_fibs.each_while(under_4_million){|f| sum += f}
puts sum
# => 4613732

And the evil class that makes it "work":

# A *very* simple lazy generator based on Array
class LazyStream < Array

  # When creating a LazyStream, supply a generator lambda.
  # The generator should take as parameters the stream and the index.
  def initialize generator
    @generator = generator
  end

  # Override the basic array to generate as necessary.
  def (index)
    if index.is_a? Range
      index.map{|i| self[i]}
    else
      (self.fetch(index) rescue nil) ||
        self[index] = @generator.call(self, index)
    end
  end

  # Loop through the stream until the value evaluates the block to false.
  def each_while filter
    i = 0
    loop do
      break unless filter.call(self[i])
      yield self[i]
      i += 1
    end
  end

  # Create a new stream that applies a filter to an existing stream.
  def self.filter other_stream, filter
    new_stream = new lambda{|s, index|
      v = nil; @other_index ||= 0

      index.times{|n| new_stream[n]} if index > new_stream.size

      loop{
        v = other_stream[@other_index]
        @other_index += 1
        break if filter.call(v)
      }
      v
    }
  end

private

  # Move the basic array = to private so people don't muck around with it :slight_smile:
  def =(index, value)
    super(index, value)
  end
end

···

On Wed, Aug 6, 2008 at 3:51 PM, Ryan Davis <ryand-ruby@zenspider.com> wrote:

On Aug 6, 2008, at 08:40 , tphyahoo wrote:

t = ( putStrLn . show . sum . takeWhile (<=4000000) . filter even )
fibs2
fibs2 = 1 : 2 : zipWith (+) fibs2 (tail fibs2)

Gives me hives personally... I've never been able to handle the syntax of
haskell.

but it relies on laziness and who knows what other magic.

The laziness I'm fine with... the other magic? not so much.

I decided to see if I could find a ruby solution that was more
haskellish. It turns I can... almost.

Here is my ruby-idiomatic solution to the problem:

# Solve Project Euler problem 2:
# sum fibonacci numbers <= 4_000_000, which are even.

$fib = {} # simple memoization
def fib x
$fib ||= x < 2 ? 1 : fib(x-2) + fib(x-1)
end

# fib 33 > 4m
fibs = (1..32).map { |n| fib n }.reject { |n| n % 2 == 1 }

sum = 0
fibs.each do |n|
sum += n
end

p sum
# => 4613732

because of the memoization, it runs incredibly fast. If I wasn't allowed to
predetermine the upper bound of the calculation, I'd wrap that up in a
simple loop with a conditional... Nothing fancy.

5

Maciej Tomaka wrote:

By the way, you are hunting ants with an rpg:P

He is hunting ants with a role playing game? :wink:

···

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

6

There is an archive kept of threads and you can find some of this
discussed already by searching through it. As this is your first, post,
allow me to give it to as a freebie.

http://www.ruby-forum.com/forum/4?filter=fibonacci

···

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

7

Lloyd Linklater wrote:

Maciej Tomaka wrote:

By the way, you are hunting ants with an rpg:P

He is hunting ants with a role playing game? :wink:

:slight_smile: Have you ever seen Duke Nukem? RPG was a rocket launcher :wink:

···

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