Regex for adding CDATA to XML nodes

[Tiago Fernandez]

Hello folks,

Say I have an XML like this:

<Node>Foo</Node><Node>Bar</Node>

I need to include CDATA prefix/sufix for every <Node/> element. The desired output would be:

<Node><![CDATA[Foo]]></Node><Node><![CDATA[Bar]]></Node>

Do you think it's possible to do this using regular expression without spliting the nodes and applying the transformation for each substring?

Thanks in advance,
Tiago

[Ilja Preuss]

It should be possible using String.replaceAll. Is that what you were looking for?

[Bill Cruise]

It would be tricky to try and do this with regular expression backreferences to replace the entire <Node>.*</Node> string with <Node><![CDATA[.*]]></Node>. You can get around this by doing it in two passes. First replace the opening tag, then the closing tag.

String str = "<Node>Foo</Node><Node>Bar</Node>";
str = str.replaceAll("<Node>", "<Node><![CDATA[");
str = str.replaceAll("</Node>", "]]></Node>");
System.out.println(str);

[Ilja Preuss]

Why would that be tricky? I haven't tried, but shouldn't something like the following work?

str.replaceAll("<Node>(.*?)</Node>", "<Node><![CDATA[\\1]]></Node>");

[Charles Lyons]

Be careful of greedy matching (the default matching mode) in the previous example: otherwise the .* will match everything between the very first <Node> and the very last </Node> in the document!

[Bill Cruise]

That's the first thing I tried too, but the backreference doesn't work with String's replaceAll method because the enclosing parentheses are in the regex and the backreference to the group are in the replacement String.

[Charles Lyons]

You need to use dollar $ signs for captured sub-sequences... See: Matcher.replaceAll(String)
[ July 03, 2008: Message edited by: Charles Lyons ]

[Paul Clapham]

Regex problems aside, you would also have to unescape things like ampersands and less-than symbols. For example you would want to convert

Marks & Spencer

to

<![CDATA[Marks & Spencer]]>

[ July 03, 2008: Message edited by: Paul Clapham ]

[Ilja Preuss]

Originally posted by Charles Lyons:
You need to use dollar $ signs for captured sub-sequences... See: Matcher.replaceAll(String)

[ July 03, 2008: Message edited by: Charles Lyons ]

Good point!

[Ilja Preuss]

Originally posted by Charles Lyons:
Be careful of greedy matching (the default matching mode) in the previous example: otherwise the .* will match everything between the very first <Node> and the very last </Node> in the document!

If you are talking about my example, those are exactly *not* greedy.

[Charles Lyons]

If you are talking about my example, those are exactly *not* greedy.

Nope - I did notice the extra "?" in your example. It's just something very easy to overlook, generally not well understood and I thought a good idea to point out!

[Mike Simmons]

Another complication: can node elements be nested inside other Nodes? That could get ugly:

<Node>Foo<Node>Bar</Node></Node>

That might become:

<Node><![CDATA[Foo<Node><![CDATA[Bar]]></Node>]]></Node>

The above is almost certainly not what you want here, but I don't know what you do want. With luck, nesting never occurs, and this will be easier.

[Carey Evans]

Paul Clapham:
... you would also have to unescape things like ampersands and less-than symbols.

While looking out for ]]&gt;, which can't be part of a CDATA section at all.
[ July 05, 2008: Message edited by: Carey Evans ]

[Tiago Fernandez]

Thanks everyone! I've taken the following solution for the problem I was having anyways:

String str = "<Node>ABC</Node><Node>XXX</Node>";
str = str.replaceAll("<Node>(.*?)</Node>", "<Node><![CDATA[$1]]></Node>");

Cheers!