How can I make empty tags self-closing with Nokogiri?

I've created an XML template in ERB. I fill it in with data from a database during an export process.

In some cases, there is a null value, in which case an element may be empty, like this:

<someitem>

</someitem>

In that case, the client receiving the export wants it to be converted into a self-closing tag:

<someitem/>

I'm trying to see how to get Nokogiri to do this, but I don't see it yet. Does anybody know how to make empty XML tags self-closing with Nokogiri?

Update

A regex was sufficient to do what I specified above, but the client now also wants tags whose children are all empty to be self-closing. So this:

<someitem>
  <subitem>

  </subitem>
  <subitem>

  </subitem>
</someitem>

... should also be

<someitem/>

I think that this will require using Nokogiri.

Search for

<([^>]+)>\s*</\1>

and replace with

<\1/>

In Ruby:

result = subject.gsub(/<([^>]+)>\s*<\/\1>/, '<\1/>')

Explanation:

<       # Match opening bracket
(       # Match and remember...
 [^>]+  # One or more characters except >
)       # End of capturing group
>       # Match closing bracket
\s*     # Match optional whitespace & newlines
<       # Match opening bracket
/       # Match /
\1      # Match the contents of the opening tag
>       # Match closing bracket

A couple questions:

1. <foo></foo> is the same as <foo />, so why worry about such a tiny detail? If it is syntactically significant because the text node between the two is a "\n", then put a test in your ERB template that checks for the value that would go there, and if it's not initialized output the self-closing tag instead? See "Yak shaving".
2. Why involve Nokogiri? You should be able to generate correct XML in ERB since you're in control of the template.

EDIT - Nokogiri's behavior is to not-rewrite parsed XML unless it has to. I suspect you'd have to remove the node in question, then reinsert it as an empty node to get Nokogiri to output what you want.