How to create a histogram from a flat Array in Ruby

Question

How do I create a histogram of an array of integers? For example:

data = [0,1,2,2,2,2,2,3,3,3,3,3,3,4,4,4,4,5,5,6,6,6,7,7,7,7,7,8,9,9,10]

I want to create a histogram based on how many entries there are for 0, 1, 2, and so on. Is there an easy way to do it in Ruby?

The output should be two arrays. The first array should contain the groups (bins), the second array should contain the number of occurrences (frequencies).

For data given above, I would expect the following output:

bins         # => [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
frequencies  # => [1, 1, 5, 6, 4, 2, 3, 5, 1, 2, 1]

Solution 1

Use "histogram".

data = [0,1,2,2,2,2,2,3,3,3,3,3,3,4,4,4,4,5,5,6,6,6,7,7,7,7,7,8,9,9,10]
(bins, freqs) = data.histogram 

This will create an array bins containing the bins of histogram and the array freqs containing the frequencies. The gem also supports different binning behaviors and weights/fractions.

Hope this helps.

Solution 2

Ruby's Array inherits group_by from Enumerable, which does this nicely:

Hash[*data.group_by{ |v| v }.flat_map{ |k, v| [k, v.size] }]

Which returns:

{
     0 => 1,
     1 => 1,
     2 => 5,
     3 => 6,
     4 => 4,
     5 => 2,
     6 => 3,
     7 => 5,
     8 => 1,
     9 => 2,
    10 => 1
}

That's just a nice 'n clean hash. If you want an array of each bin and frequency pair you can shorten it and use:

data = [0,1,2,2,3,3,3,4]
data.group_by{ |v| v }.map{ |k, v| [k, v.size] }
# => [[0, 1], [1, 1], [2, 2], [3, 3], [4, 1]]

Here's what the code and group_by is doing with the smaller dataset:

data.group_by{ |v| v }    
# => {0=>[0], 1=>[1], 2=>[2, 2], 3=>[3, 3, 3], 4=>[4]}

data.group_by{ |v| v }.flat_map{ |k, v| [k, v.size] }  
# => [0, 1, 1, 1, 2, 2, 3, 3, 4, 1]

---

As mentioned by Telmo Costa in the comments, Ruby introduced tally in v2.7.0. Running a quick benchmark shows that tally is about 3x faster:

require 'fruity'

puts "Ruby v#{RUBY_VERSION}"

data = [0,1,2,2,2,2,2,3,3,3,3,3,3,4,4,4,4,5,5,6,6,6,7,7,7,7,7,8,9,9,10]

data.group_by{ |v| v }.map{ |k, v| [k, v.size] }.to_h
# => {0=>1, 1=>1, 2=>5, 3=>6, 4=>4, 5=>2, 6=>3, 7=>5, 8=>1, 9=>2, 10=>1}
data.group_by { |v| v }.transform_values(&:size)
# => {0=>1, 1=>1, 2=>5, 3=>6, 4=>4, 5=>2, 6=>3, 7=>5, 8=>1, 9=>2, 10=>1}
data.tally 
# => {0=>1, 1=>1, 2=>5, 3=>6, 4=>4, 5=>2, 6=>3, 7=>5, 8=>1, 9=>2, 10=>1}
data.group_by{ |v| v }.keys.sort.map { |key| [key, data.group_by{ |v| v }[key].size] }.to_h
# => {0=>1, 1=>1, 2=>5, 3=>6, 4=>4, 5=>2, 6=>3, 7=>5, 8=>1, 9=>2, 10=>1}

compare do
  gb { data.group_by{ |v| v }.map{ |k, v| [k, v.size] }.to_h }
  rriemann { data.group_by { |v| v }.transform_values(&:size) }
  telmo_costa { data.tally }
  CBK {data.group_by{ |v| v }.keys.sort.map { |key| [key, data.group_by{ |v| v }[key].size] }.to_h }
end

Resulting in:

# >> Ruby v2.7.0
# >> Running each test 1024 times. Test will take about 2 seconds.
# >> telmo_costa is faster than rriemann by 2x ± 0.1
# >> rriemann is similar to gb
# >> gb is faster than CBK by 8x ± 1.0

So use tally.

Comments

[sawa]

What is the output format you want?

[PJP]

When you ask a question, asking for code, you need to show your research and any attempts you made to solve the problem, along with your explanation why they didn't work.

[knugie]

I would initialise a "counting hash" like that h = Hash.new(0) and count occurrences of each element: data.each{|v| h[v] += 1}. After that, h will look like this: => {0=>1, 1=>1, 2=>5, 3=>6, 4=>4, 5=>2, 6=>3, 7=>5, 8=>1, 9=>2, 10=>1} You can extract bins and freqs from that Hash using h.keys and h.values. I hope you find that useful.

[dug]

Ruby 2.7.0 introduces Enumerable#tally: data.tally => {0=>1, 1=>1, 2=>5, 3=>6, 4=>4, 5=>2, 6=>3, 7=>5, 8=>1, 9=>2, 10=>1}

[rriemann]

The latest ruby versions allow to use some syntax sugar for a shorter version: data = group_by { |v| v }.transform_values(&:size)

[sambecker]

This should be the accepted answer!

[Telmo Costa]

You can use itself: data.group_by(&:itself).transform_values(&:size). Or, has it has been said before, starting on Ruby 2.7.0 data.tally.

[CBK]

To sort the keys in ascending order: data.group_by{ |v| v }.keys.sort.map do |key| [key, data.group_by{ |v| v }[key].size] end

[PJP]

No, don't. It's at least 8x slower if you use the OPs sample array. See the benchmarks.