# Hash with array as value type

**URL:** <https://www.ruby-forum.com/t/hash-with-array-as-value-type/53141>\
**Category:** Ruby\
**Created:** [January 26, 2006, 12:22pm UTC](https://www.ruby-forum.com/t/hash-with-array-as-value-type/53141 "2006-01-26T12:22:40Z")\
**Posts on this page:** 7\
**Page:** 1

<div class="post-metadata">

**Author:** ![Michael\_McGreevy](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/m/eb8c5e/32.png) [@Michael\_McGreevy](https://www.ruby-forum.com/u/Michael_McGreevy)\
**Post date:** [January 26, 2006, 12:22pm UTC](https://www.ruby-forum.com/t/hash-with-array-as-value-type/53141/1 "2006-01-26T12:22:40Z")

</div>

Hello all,

I would like to store arrays in a hash, indexed by a string key. I  
would like to have the hash create an empty array as the default value,  
when it sees a new key, something like the code involving “hash1” below.  
However, this code is giving some strange results – it claims that the  
hash is empty, even though there is an array stored in it, and I can  
then retrieve that array.

I am new to ruby, so maybe I am just doing something stupid (… I am  
not sure about that “Hash.new( [] )” for example… )

Can anyone explain these results?

Thanks.

CODE:

#!/usr/bin/ruby

hash1 = Hash.new( [] )  
hash1[“hello”].push(1.0)  
hash1[“hello”].push(2.0)  
$stderr.write “hash1.size() = #{hash1.size()}\n”  
$stderr.write “hash1.empty() = #{hash1.empty?()}\n”  
$stderr.write “hash1[“hello”].size() = #{hash1[“hello”].size()}\n”  
$stderr.write “hash1[“hello”] = #{hash1[“hello”]}\n”

hash2 = {“hello” =\> [1.0,2.0]}  
$stderr.write “\nhash2.size() = #{hash2.size()}\n”  
$stderr.write “hash2.empty() = #{hash2.empty?()}\n”  
$stderr.write “hash2[“hello”].size() = #{hash2[“hello”].size()}\n”  
$stderr.write “hash2[“hello”] = #{hash2[“hello”]}\n”

OUTPUT:

hash1.size() = 0  
hash1.empty() = true  
hash1[“hello”].size() = 2  
hash1[“hello”] = 1.02.0

hash2.size() = 1  
hash2.empty() = false  
hash2[“hello”].size() = 2  
hash2[“hello”] = 1.02.0

VERSION:  
ruby 1.8.3 (2005-09-21) [i686-linux]

---

<div class="post-metadata">

**Author:** ![Michael\_McGreevy](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/m/eb8c5e/32.png) [@Michael\_McGreevy](https://www.ruby-forum.com/u/Michael_McGreevy)\
**Post date:** [January 26, 2006, 2:19pm UTC](https://www.ruby-forum.com/t/hash-with-array-as-value-type/53141/2 "2006-01-26T14:19:39Z")

</div>

Michael McGreevy wrote:

> Hello all,
> 
> I would like to store arrays in a hash, indexed by a string key. I  
> would like to have the hash create an empty array as the default value,  
> when it sees a new key, something like the code involving “hash1” below.  
> However, this code is giving some strange results – it claims that the  
> hash is empty, even though there is an array stored in it, and I can  
> then retrieve that array.
> 
> I am new to ruby, so maybe I am just doing something stupid (… I am  
> not sure about that “Hash.new( [] )” for example… )
> 
> Can anyone explain these results?
> 
> Thanks.
> 
> CODE:
> 
> #!/usr/bin/ruby
> 
> hash1 = Hash.new( [] )

This makes a new Hash with the default value being the Array reference  
you’ve passed it. This is only executed _once_, so each new key/value  
pair points to the _same_ Array instance. You want

hash1 = Hash.new { |h,k| Array.new }

That’ll make a new Array instance each time it needs a new default value  
for a new key.

---

<div class="post-metadata">

**Author:** ![Michael\_McGreevy](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/m/eb8c5e/32.png) [@Michael\_McGreevy](https://www.ruby-forum.com/u/Michael_McGreevy)\
**Post date:** [January 26, 2006, 2:19pm UTC](https://www.ruby-forum.com/t/hash-with-array-as-value-type/53141/3 "2006-01-26T14:19:48Z")

</div>

Michael McGreevy wrote:

> not sure about that “Hash.new( [] )” for example… )
> 
> Can anyone explain these results?

Yes. You ran into the typical Hash pitfal: the default value is the one  
returned if something is not found for the given key. But it never  
changes the hash and there is just this single instance. Consider this:

> > h=Hash.new([])  
> > =\> {}
> 
> > h[0]\<\<1  
> > =\> [1]
> 
> > h[0]  
> > =\> [1]
> 
> > h[“foo”]  
> > =\> [1]
> 
> > h[“bar”] \<\< 2  
> > =\> [1, 2]
> 
> > h[:x]  
> > =\> [1, 2]
> 
> > h.default  
> > =\> [1, 2]

It works for numeric values because then one usually assigns:

> > h=Hash.new(0)  
> > =\> {}
> 
> > h[:foo] += 1  
> > =\> 1
> 
> > h[:bar] += 10  
> > =\> 10
> 
> > h  
> > =\> {:bar=\>10, :foo=\>1}

Note the “+=” contains an assignment and it’s equivalent to

h[:foo] = h[:foo] + 1

You on the other hand want the block form because that can do arbitrary  
things when a key is not found:

> > h=Hash.new() {|ha,key| puts “missing #{key}”; ha[key] = []}  
> > =\> {}
> 
> > h[:foo] \<\< “foo”  
> > missing foo  
> > =\> [“foo”]
> 
> > h[:bar] \<\< “foo”  
> > missing bar  
> > =\> [“foo”]
> 
> > h[:foo] \<\< “foo end”  
> > =\> [“foo”, “foo end”]
> 
> > h[:foo] \<\< “foo more”  
> > =\> [“foo”, “foo end”, “foo more”]
> 
> > h  
> > =\> {:bar=\>[“foo”], :foo=\>[“foo”, “foo end”, “foo more”]}

Kind regards

```
robert
```

---

<div class="post-metadata">

**Author:** ![Michael\_McGreevy](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/m/eb8c5e/32.png) [@Michael\_McGreevy](https://www.ruby-forum.com/u/Michael_McGreevy)\
**Post date:** [January 26, 2006, 2:33pm UTC](https://www.ruby-forum.com/t/hash-with-array-as-value-type/53141/4 "2006-01-26T14:33:14Z")

</div>

Robert K. wrote:

> You on the other hand want the block form because that can do arbitrary  
> things when a key is not found:
> 
> > > h=Hash.new() {|ha,key| puts “missing #{key}”; ha[key] = []}

Thanks a lot :-). I had actually already tried the block form, but in a  
way similar to what Mike F. posted. I now see that simply  
returning something from the block does not modify the hash… you have  
to explicitly modify the hash within the block.

Thanks to both of you for responding.

Michael.

---

<div class="post-metadata">

**Author:** ![Michael\_McGreevy](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/m/eb8c5e/32.png) [@Michael\_McGreevy](https://www.ruby-forum.com/u/Michael_McGreevy)\
**Post date:** [January 26, 2006, 3:00pm UTC](https://www.ruby-forum.com/t/hash-with-array-as-value-type/53141/5 "2006-01-26T15:00:21Z")

</div>

Mike F. wrote:  
[…]

> hash1 = Hash.new { |h,k| Array.new }
> 
> That’ll make a new Array instance each time it needs a new default value  
> for a new key.

GAH, that should be

hash1 = Hash.new { | h, k | h[k] = Array.new }

MEMO TO SELF: Don’t post in the mornings before caffeine has chance to  
take effect.

---

<div class="post-metadata">

**Author:** ![Michael\_McGreevy](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/m/eb8c5e/32.png) [@Michael\_McGreevy](https://www.ruby-forum.com/u/Michael_McGreevy)\
**Post date:** [January 26, 2006, 3:05pm UTC](https://www.ruby-forum.com/t/hash-with-array-as-value-type/53141/6 "2006-01-26T15:05:08Z")

</div>

On Jan 26, 2006, at 6:22 AM, Michael McGreevy wrote:

> I am new to ruby, so maybe I am just doing something stupid (… I am  
> not sure about that “Hash.new( [] )” for example… )  
> Can anyone explain these results?

It is a bit confusing.

Hash.new([]) tucks away the newly created array as the default value:

d = []  
puts d.object\_id  
hash1 = Hash.new(d)  
puts hash1.default.object\_id

You can see from this that Hash has saved a reference to the default  
value.  
So when you call:

hash1[“hello”].push(1.0)

on the empty hash, a reference to the default value is returned because  
the key is not found. The float, 1.0, is pushed onto that default  
value.  
You still haven’t entered anything into the Hash itself but you have  
pushed  
a value into the default array:

puts hash1.default

This same default object will be returned each time a lookup fails in  
the Hash.

The solution is to use the block form of the Hash constructor.

hash1 = Hash.new { |h,k| h[k] = [] }

In this form, every lookup miss causes the block to be called with  
the hash  
and the key as the two arguments. The block allocates a new array  
and then  
stores it back into the hash using the key so that the next lookup  
finds the  
newly allocated array and doesn’t call the block.

Here is your example rewritten with that and adjusted to follow the  
usual ruby coding styles:

hash1 = Hash.new { |h,k| h[k] = [] }  
hash1[“hello”].push(1.0)  
hash1[“hello”].push(2.0)  
warn(“hash1.size = #{hash1.size}”)  
warn(“hash1.empty? = #{hash1.empty?}”)  
warn(“hash1[“hello”].size = #{hash1[“hello”].size}”)  
warn(“hash1[“hello”] = #{hash1[“hello”]}”)  
warn("")

hash2 = {“hello” =\> [1.0,2.0]}  
warn(“hash2.size = #{hash2.size}”)  
warn(“hash2.empty = #{hash2.empty?}”)  
warn(“hash2[“hello”].size = #{hash2[“hello”].size}”)  
warn(“hash2[“hello”] = #{hash2[“hello”]}”)

Gary W.

---

<div class="post-metadata">

**Author:** ![Michael\_McGreevy](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/m/eb8c5e/32.png) [@Michael\_McGreevy](https://www.ruby-forum.com/u/Michael_McGreevy)\
**Post date:** [January 26, 2006, 3:50pm UTC](https://www.ruby-forum.com/t/hash-with-array-as-value-type/53141/7 "2006-01-26T15:50:37Z")

</div>

Robert K. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

> > I am new to ruby, so maybe I am just doing something stupid (… I am  
> > not sure about that “Hash.new( [] )” for example… )
> > 
> > Can anyone explain these results?
> 
> Yes. You ran into the typical Hash pitfal: the default value is the one  
> returned if something is not found for the given key. But it never  
> changes the hash and there is just this single instance. Consider this:

I find it instructive to think of the default as a #key\_missing method,  
analogous to #method\_missing.

martin