I have written a system for rendering webpages which takes a document called .axml and renders it into html. The axml document contains xml tags which correspond to perl plugins which get eval'd and the result pasted back into the doc. The upshot is that I can develop documents which are very similar to html but with all the power of a perl CGI system behind it.

action.pl?ref=somedata

a <qd>ref</qd> tag in the document will trigger the qd plugin which then returns the value "somedata". This value is then pasted into the document in the place where the <qd> tag was. I have defined several dozen such tags and have a set large enough to build most small CGI applications with.

a simple axml case statement action.pl?ref=file1

<case>
  <c><qd>ref</qd></c>
  <file1><insertfile>data/file1.xml</insertfile></file1>
  <file2><insertfile>data/file2.xml</insertfile></file2>
  <else>
    <insertfile>data/file3.xml</insertfile>
  </else>
</case>

update a value in an xml file

<changexmlvalue>
  <d>actions/publicpages/db/publicpages.xml</d>
  <key>page</key>
  <match>id=<qd>edit</qd></match>
  <change>stylesheet</change>
  <newvalue><qd>file</qd></newvalue> 
</changexmlvalue>
  
foreach...

<foreach>
  <s>publicpages</s>
  <key>page</key>
    <mask>
      <d>pagename</d><br/>
     </mask>
</foreach>

<p>where s can be a file or some data, key is the xml primary key to t
+he data, mask is the output for each instance, d is a data item.</p>

given the data 

<page>
  <pagename>welcome</pagename>
  <title>Welcome to Widgets.com</title>
</page>
<page>
  <pagename>login</pagename>
  <title>Login to Widgets.com</title>
</page>

the foreach in this example will produce;

welcome<br/>
login<br/>
[download]

------------------------

i have built dozens of useful little plugins for the system and the main parser allows for these plugins to be assembled like little lego blocks for just about any complex output to be built.

axml allows for 3 types of tags, delimted by the standard <>, and two non-standards... and ( )

axml tags are processed from the inside outwards. the tags are left to the final pass, whilst ( ) tags are processed before < > tags. using these three bracket types ive been able to construct complex processing orders. for instance (case) (qd)ref(/qd) <....

in this case, the qd is run first, and then the case, both of which are in ( ) brackets and therefore run prior to any < > tags which the case contains.