filter out tag name in cdata sections or comments

[kanghao]

how can I filter out xml elements in cdata sections or comments.

I have to parse xml elements and manupulate their attributes.

but I want the text in the form of <TAG> ... </TAG> or <TAG/> in the cdata sections or comments untouched.

help me.

Thanks.

 

[kanghao]

I have to modify attributes values in the xml document below
to a string-added value.

<DOC>
<Drawing id="2"><![CDATA[<EMI id="1" height="95" width="95" file="utm00004.jpg">]]>
<EMI id="2" height="95" width="95" file="utm00004.jpg">ads</EMI>
</Drawing>
<!-- bla... <EMI id="3" height="95" width="95" file="utm00004.jpg"> ... bla -->
</DOC>

to

<DOC>
<Drawing id="2"><![CDATA[<EMI id="1" height="95" width="95" file="utm00004.jpg">]]>
<EMI id="2" height="95" width="95" file="56789-utm00004.jpg">ads</EMI>
</Drawing>
<!-- bla... <EMI id="3" height="95" width="95" file="utm00004.jpg"> ... bla -->
</DOC>

the value 56789 is from a database on the id value 2.

I can <EMI part using regex <EMI[\w\s\"\=\.\-\/]{0,}>.
how can I filter out the tags in the comments or cdata sections?