comp.rb

require 'daru'
require 'benchmark'


# Vector :b is not required here. Taking it only to compare the performance
# with second dataframe which also has two columns.
df1 = Daru::DataFrame.new({
  a: [1, 2, 3]*1000,
  b: [1, 2, 3]*1000},
  index: Daru::CategoricalIndex.new([:a, :b, :c, :d, :e, :f]*500)
)

# Here we instead made the categorical index a column
df2 = Daru::DataFrame.new({
  a: [1, 2, 3]*1000, 
  idx: [:a, :b, :c, :d, :e, :f]*500
})

# Lets fetch entries with value :a
Benchmark.bm do |x|
  x.report 'with categorical index' do
    1000.times { df1[:a] }
  end
  
  x.report 'without categorical index' do
    1000.times { df2.where df2[:idx].eq(:a) }
  end
end

# Result
                             user     system      total        real
with categorical index     0.010000   0.000000   0.010000 (  0.010677)
without categorical index  4.820000   0.020000   4.840000 (  4.885409)