Ruby dup/clone recursively

Question

I have a hash like:

h = {'name' => 'sayuj', 
     'age' => 22, 
     'project' => {'project_name' => 'abc', 
                   'duration' => 'prq'}}

I need a dup of this hash, the change should not affect the original hash.

When I try,

d = h.dup # or d = h.clone
d['name'] = 'sayuj1'
d['project']['duration'] = 'xyz'

p d #=> {"name"=>"sayuj1", "project"=>{"duration"=>"xyz", "project_name"=>"abc"}, "age"=>22}
p h #=> {"name"=>"sayuj", "project"=>{"duration"=>"xyz", "project_name"=>"abc"}, "age"=>22}

Here you can see the project['duration'] is changed in the original hash because project is another hash object.

I want the hash to be duped or cloned recursively. How can I achieve this?

Answer 1

If you are in Rails: Hash.deep_dup

Answer 2

Another alternative is to use the full_dup gem (full disclosure: I am the author of that gem) that handles arrays, hashes, structs, and is extendable to user defined classes.

To use:

require 'full_dup'
# Other code omitted ...
d = h.full_dup

Also note that full_dup handles complex data relationships including those with loops or recursion.

Answer 3

In case the Marchal #dump/load pair isn't work, for rails there is a Hash's method #deep_dup, so you can:

h = {'name' => 'sayuj', 
 'age' => 22, 
 'project' => {'project_name' => 'abc', 
               'duration' => 'prq'}}

h1 = h.deep_dup

Answer 4

Here's how you make deep copies in Ruby

d = Marshal.load( Marshal.dump(h) )

Answer 5

This is an answer to a reasonably old question, but I happened upon it while implementing something similar, thought I'd chime in for a more efficient method.

For the simple, two level deep hash like above, you can also do something like this:

d = h.inject({}) {|copy, (key, value)| 
    copy[key] = value.dup rescue value; copy
}

I ran a test on a hash of hashes with 4k elements, each a few hundred bytes, and it was about 50% faster than the Marshal.dump/load

Of course, it's not as complete, as it won't work if you have a hash as, e.g., the value of the 'project_name' field, but for a simple 2 level hash, it works great / faster.