Optimising code for matching two strings modulo scrambling - ruby

Question

I am trying to write a function scramble(str1, str2) that returns true if a portion of str1 characters can be rearranged to match str2, otherwise returns false. Only lower case letters (a-z) will be used. No punctuation or digits will be included. For example:
str1 = 'rkqodlw'; str2 = 'world' should return true.
str1 = 'cedewaraaossoqqyt'; str2 = 'codewars' should return true.
str1 = 'katas'; str2 = 'steak' should return false.
This is my code:
def scramble(s1, s2)
#sorts strings into arrays
first = s1.split("").sort
second = s2.split("").sort
correctLetters = 0
for i in 0...first.length
#check for occurrences of first letter
occurrencesFirst = first.count(s1[i])
for j in 0...second.length
#scan through second string
occurrencesSecond = second.count(s2[j])
#if letter to be tested is correct and occurrences of first less than occurrences of second
#meaning word cannot be formed
if (s2[j] == s1[i]) && occurrencesFirst < occurrencesSecond
return false
elsif s2[j] == s1[i]
correctLetters += 1
elsif first.count(s1[s2[j]]) == 0
return false
end
end
end
if correctLetters == 0
return false
end
return true
end
I need help optimising this code. Please give me suggestions.

Answer 1

Here is one efficient and Ruby-like way of doing that.
Code
def scramble(str1, str2)
h1 = char_counts(str1)
h2 = char_counts(str2)
h2.all? { |ch, nbr| nbr <= h1[ch] }
end
def char_counts(str)
str.each_char.with_object(Hash.new(0)) { |ch, h| h[ch] += 1 }
end
Examples
scramble('abecacdeba', 'abceae')
#=> true
scramble('abecacdeba', 'abweae')
#=> false
Explanation
The three steps are as follows.
str1 = 'abecacdeba'
str2 = 'abceae'
h1 = char_counts(str1)
#=> {"a"=>3, "b"=>2, "e"=>2, "c"=>2, "d"=>1}
h2 = char_counts(str2)
#=> {"a"=>2, "b"=>1, "c"=>1, "e"=>2}
h2.all? { |ch, nbr| nbr <= h1[ch] }
#=> true
The last statement is equivalent to
2 <= 3 && 1 <= 2 && 1 <= 2 && 2 <=2
The method char_counts constructs what is sometimes called a "counting hash". To understand how char_counts works, see Hash::new, especially the explanation of the effect of providing a default value as an argument of new. In brief, if a hash is defined h = Hash.new(0), then if h does not have a key k, h[k] returns the default value, here 0 (and the hash is not changed).
Suppose, for different data,
h1 = { "a"=>2 }
h2 = { "a"=>1, "b"=>2 }
Then we would find that 1 <= 2 #=> true but 2 <= 0 #=> false, so the method would return false. The second comparison is 2 <= h1["b"]. As h1 does not have a key "b", h1["b"] returns the default value, 0.
The method char_counts is effectively a short way of writing the method expressed as follows.
def char_counts(str)
h = {}
str.each_char do |ch|
h[ch] = 0 unless h.key?(ch) # instead of Hash.new(0)
h[ch] = h[c] + 1 # instead of h[c][ += 1
end
h # no need for this if use `each_with_object`
end
See Enumerable#each_with_object, String#each_char (preferable to String.chars, as the latter produces an unneeded temporary array whereas the former returns an enumerator) and Hash#key? (or Hash#has_key?, Hash#include? or Hash#member?).
An Alternative
def scramble(str1, str2)
str2.chars.difference(str1.chars).empty?
end
class Array
def difference(other)
h = other.each_with_object(Hash.new(0)) { |e,h| h[e] += 1 }
reject { |e| h[e] > 0 && h[e] -= 1 }
end
end
I have found the method Array#difference to be so useful I proposed it be added to the Ruby Core (here). The response has been, er, underwhelming.

Answer 2

One way:
def scramble(s1,s2)
s2.chars.uniq.all? { |c| s1.count(c) >= s2.count(c) }
end
Another way:
def scramble(s1,s2)
pool = s1.chars.group_by(&:itself)
s2.chars.all? { |c| pool[c]&.pop }
end
Yet another:
def scramble(s1,s2)
('a'..'z').all? { |c| s1.count(c) >= s2.count(c) }
end
Since this appears to be from codewars, I submitted my first two there. Both got accepted and the first one was a bit faster. Then I was shown solutions of others and saw someone using ('a'..'z') and it's fast, so I include that here.
The codewars "performance tests" aren't shown explicitly but they're all up to about 45000 letters long. So I benchmarked these solutions as well as Cary's (yours was too slow to be included) on shuffles of the alphabet repeated to be about that long (and doing it 100 times):
user system total real
Stefan 1 0.812000 0.000000 0.812000 ( 0.811765)
Stefan 2 2.141000 0.000000 2.141000 ( 2.127585)
Other 0.125000 0.000000 0.125000 ( 0.122248)
Cary 1 2.562000 0.000000 2.562000 ( 2.575366)
Cary 2 3.094000 0.000000 3.094000 ( 3.106834)
Moral of the story? String#count is fast here. Like, ridiculously fast. Almost unbelievably fast (I actually had to run extra tests to believe it). It counts through about 1.9 billion letters per second (100 times 26 letters times 2 strings of ~45000 letters, all in 0.12 seconds). Note that the difference to my own first solution is just that I do s2.chars.uniq, and that increases the time from 0.12 seconds to 0.81 seconds. Meaning this double pass through one string takes about six times as long as the 52 passes for counting. The counting is about 150 times faster. I did expect it to be very fast, because it presumably just searches a byte in an array of bytes using C code (edit: looks like it does), but this speed still surprised me.
Code:
require 'benchmark'
def scramble_stefan1(s1,s2)
s2.chars.uniq.all? { |c| s1.count(c) >= s2.count(c) }
end
def scramble_stefan2(s1,s2)
pool = s1.chars.group_by(&:itself)
s2.chars.all? { |c| pool[c]&.pop }
end
def scramble_other(s1,s2)
('a'..'z').all? { |c| s1.count(c) >= s2.count(c) }
end
def scramble_cary1(str1, str2)
h1 = char_counts(str1)
h2 = char_counts(str2)
h2.all? { |ch, nbr| nbr <= h1[ch] }
end
def char_counts(str)
str.each_char.with_object(Hash.new(0)) { |ch, h| h[ch] += 1 }
end
def scramble_cary2(str1, str2)
str2.chars.difference(str1.chars).empty?
end
class Array
def difference(other)
h = other.each_with_object(Hash.new(0)) { |e,h| h[e] += 1 }
reject { |e| h[e] > 0 && h[e] -= 1 }
end
end
Benchmark.bmbm do |x|
n = 100
s1 = (('a'..'z').to_a * (45000 / 26)).shuffle.join
s2 = s1.chars.shuffle.join
x.report('Stefan 1') { n.times { scramble_stefan1(s1, s2) } }
x.report('Stefan 2') { n.times { scramble_stefan2(s1, s2) } }
x.report('Other') { n.times { scramble_other(s1, s2) } }
x.report('Cary 1') { n.times { scramble_cary1(s1, s2) } }
x.report('Cary 2') { n.times { scramble_cary2(s1, s2) } }
end