Why is it not a good idea to dynamically create a lot of symbols in ruby (for versions before 2.2)? Why is it not a good idea to dynamically create a lot of symbols in ruby (for versions before 2.2)? ruby ruby

Why is it not a good idea to dynamically create a lot of symbols in ruby (for versions before 2.2)?


Symbols are like strings but they are immutable - they can't be modified.

They are only put into memory once, making them very efficient to use for things like keys in hashes but they stay in memory until the program exits. This makes them a memory hog if you misuse them.

If you dynamically create lots of symbols, you are allocating a lot of memory that can't be freed until your program ends. You should only dynamically create symbols (using string.to_sym) if you know you will:

  1. need to repeatedly access the symbol
  2. not need to modify them

As I said earlier, they are useful for things like hashes - where you care more about the identity of the variable than its value. Symbols, when correctly used, are a readable and efficient way to pass around identity.

I will explain what I mean about the immutability of symbols RE your comment.

Strings are like arrays; they can be modified in place:

12:17:44 ~$ irbirb(main):001:0> string = "Hello World!"=> "Hello World!"irb(main):002:0> string[5] = 'z'=> "z"irb(main):003:0> string=> "HellozWorld!"irb(main):004:0> 

Symbols are more like numbers; they can't be edited in place:

irb(main):011:0> symbol = :Hello_World=> :Hello_Worldirb(main):012:0> symbol[5] = 'z'NoMethodError: undefined method `[]=' for :Hello_World:Symbol    from (irb):12    from :0


A symbol is the same object and the same allocation of memory no matter where it is used:

>> :hello.object_id=> 331068>> a = :hello=> :hello>> a.object_id=> 331068>> b = :hello=> :hello>> b.object_id=> 331068>> a = "hello"=> "hello">> a.object_id=> 2149256980>> b = "hello"=> "hello">> b.object_id=> 2149235120>> b = "hell" + "o"

Two strings which are 'the same' in that they contain the same characters may not reference the same memory, which can be inefficient if you're using strings for, say, hashes.

So, symbols can be useful for reducing memory overhead. However - they are a memory leak waiting to happen, because symbols cannot be garbage collected once created. Creating thousands and thousands of symbols would allocate the memory and not be recoverable. Yikes!


It can be particularly bad to create symbols from user input without validating the input against some kind of a white-list (for example, for query string parameters in RoR). If user input is converted to symbols without validation, a malicious user can cause your program to consume large amounts of memory that will never be garbage collected.

Bad (a symbol is created regardless of user input):

name = params[:name].to_sym

Good (a symbol is only created if the user input is allowed):

whitelist = ['allowed_value', 'another_allowed_value']raise ArgumentError unless whitelist.include?(params[:name])name = params[:name].to_sym