CODEWITHSUNDEEP
rubyselect
sunny-mittal
sunny-mittal

Reputation: 509

Strange Ruby 2+ Behavior with "select!"

I'm having an issue that I can't seem to find documented or explained anywhere so I'm hoping someone here can help me out. I've verified the unexpected behavior on three versions of Ruby, all 2.1+, and verified that it doesn't happen on an earlier version (though it's through tryruby.org and I don't know which version they're using). Anyway, for the question I'll just post some code with results and hopefully someone can help me debug it.

arr = %w( r a c e c a r )             #=> ["r","a","c","e","c","a","r"]
arr.select { |c| arr.count(c).odd? }  #=> ["e"]
arr.select! { |c| arr.count(c).odd? } #=> ["e","r"] <<<<<<<<<<<<<<< ??????

I think the confusing part for me is clearly marked and if anyone can explain if this is a bug or if there's some logic to it, I'd greatly appreciate it. Thanks!

Upvotes: 0

Views: 63

Answers (2)

the Tin Man
the Tin Man

Reputation: 160571

Just to give you some other ways of accomplishing what you're doing:

arr = %w( r a c e c a r )
arr.group_by{ |c| arr.count(c).odd? }        
# => {false=>["r", "a", "c", "c", "a", "r"], true=>["e"]}
arr.group_by{ |c| arr.count(c).odd? }.values 
# => [["r", "a", "c", "c", "a", "r"], ["e"]]
arr.partition{ |c| arr.count(c).odd? }       
# => [["e"], ["r", "a", "c", "c", "a", "r"]]

And if you want more readable keys:

arr.group_by{ |c| arr.count(c).odd? ? :odd : :even } 
# => {:even=>["r", "a", "c", "c", "a", "r"], :odd=>["e"]}

partition and group_by are basic building blocks for separating elements in an array into some sort of grouping, so it is good to be familiar with them.

Upvotes: 0

user229044
user229044

Reputation: 239382

You're modifying the array while you're read from it while you iterate over it. I'm not sure the result is defined behavior. The algorithm isn't required to keep the object in any kind of sane state while it's running.

Some debug printing during the iteration shows why your particular result happens:

irb(main):005:0> x
=> ["r", "a", "c", "e", "c", "a", "r"]
irb(main):006:0> x.select! { |c| p x; x.count(c).odd? }
["r", "a", "c", "e", "c", "a", "r"]
["r", "a", "c", "e", "c", "a", "r"]
["r", "a", "c", "e", "c", "a", "r"]
["r", "a", "c", "e", "c", "a", "r"] # "e" is kept...
["e", "a", "c", "e", "c", "a", "r"] # ... and moved to the start of the array
["e", "a", "c", "e", "c", "a", "r"]
["e", "a", "c", "e", "c", "a", "r"] # now "r" is kept
=> ["e", "r"]

You can see by the final iteration, there is only one r, and that the e has been moved to the front of the array. Presumably the algorithm modifies the array in-place, moving matched elements to the front, overwriting elements that have already failed your test. It keeps track of how many elements are matched and moved, and then truncates the array down to that many elements.

So, instead, use select.

A longer example that matches more elements makes the problem a little clearer:

irb(main):001:0> nums = (1..10).to_a
=> [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
irb(main):002:0> nums.select! { |i| p nums; i.even? }
[1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
[1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
[2, 2, 3, 4, 5, 6, 7, 8, 9, 10]
[2, 2, 3, 4, 5, 6, 7, 8, 9, 10]
[2, 4, 3, 4, 5, 6, 7, 8, 9, 10]
[2, 4, 3, 4, 5, 6, 7, 8, 9, 10]
[2, 4, 6, 4, 5, 6, 7, 8, 9, 10]
[2, 4, 6, 4, 5, 6, 7, 8, 9, 10]
[2, 4, 6, 8, 5, 6, 7, 8, 9, 10]
[2, 4, 6, 8, 5, 6, 7, 8, 9, 10]
=> [2, 4, 6, 8, 10]

You can see that it does indeed move matched elements to the front of the array, overwriting non-matched elements, and then truncate the array.

Upvotes: 6

Related Questions