Matt Sergeant


Apache::AxKit::XSP::Language::SimpleTaglib - alternate XSP taglib helper


    package Your::XSP::Package;
    use Apache::AxKit::Language::XSP::SimpleTaglib;

    ... more initialization stuff, start_document handler, utility functions, whatever
        you like, but no parse_start/end handler needed - if in doubt, just leave empty ...

    package Your::XSP::Package::Handlers;

    sub some_tag : attrib(id) attribOrChild(some-param) node(result) keepWhitespace {
        my ($e, $tag, %attr) = @_;
        return 'do_something($attr_some_param,'.$attr{'id'}.');';


This taglib helper allows you to easily write tag handlers with most of the common behaviours needed. It manages all 'Design Patterns' from the XSP man page plus several other useful tag styles.

Simple handler subs

A tag "<yourNS:foo>" will trigger a call to sub "foo" during the closing tag event. What happens in between can be configured in many ways using Perl function attributes. In the rare cases where some action has to happen during the opening tag event, you may provide a sub "foo__open" (double underscore) which will be called at the appropriate time. Usually you would only do that for 'if'- style tags which enclose some block of code.

It is important to understand that your tag handler is called during the XSP parse stage, when the XSP script is being constructed. Therefore, it is the responsibility of the handler to return a Perl code fragment to be appended to the XSP script, as shown above. Contrast this behaviour to TaglibHelper, where the handler is called when the XSP script is being run, and it returns data to be included in the XML output.

Context sensitive handler subs

A sub named "foo___bar" (triple underscore) gets called on the following XML input: "<yourNS:foo><yourNS:bar/></yourNS:foo>". Handler subs may have any nesting depth. The rule for opening tag handlers applies here as well. The sub name must represent the exact tag hierarchy (within your namespace).

Names, parameters, return values

Names for subs and variables get created by replacing any non-alphanumeric characters in the original tag or attribute to underscores. For example, 'get-id' becomes 'get_id'.

The called subs get passed 3 parameters: The parser object, the tag name, and an attribute hash. This hash only contains XML attributes declared using the 'attrib()' Perl function attribute. (Try not to confuse these two meanings of 'attribute' - unfortunately XML and Perl both call them that way.) The other declared parameters get converted into local variables with prefix 'attr_', or, in the case of 'childStruct', converted into the '%_' hash. These local variables are only available inside your code fragment which becomes part of the XSP script, unlike the attribute hash which is passed directly to your handler as the third parameter.

If a sub has an output attribute ('node', 'expr', etc.), the code fragment will be run in list context. If necessary, returned lists get converted to scalars by joining them without separation. Code fragments from plain subs (without an output attribute) inherit their context and have their return value left unmodified.


If more than one handler matches a tag, the following rules determine which one is chosen. Remember, though, that only tags in your namespace are considered.

  1. If the innermost tag has a 'childStruct' spec which matches, the internal childStruct handler takes precedence.

  2. Otherwise, if any surrounding tag has a matching 'child' or 'attribOrChild' attribute, the internal handler for the innermost matching tag gets chosen.

  3. Otherwise, the handler sub with the deepest tag hierarchy gets called.

Utility functions

Apache::AxKit::Language::XSP contains a few handy utility subs to help build your code fragment:

start_elem, end_elem, start_attr, end_attr

these create elements and attributes in the output document. Call them just like you call start_expr and end_expr.


given a scalar as input, it returns a scalar which yields the exact input value when evaluated; handy when using unknown text as-is in code fragments.


creates a valid, readable perl identifier from arbitrary input text. The return values might overlap.


Perl function attributes are used to define how XML output should be generated from your code fragment and how XML input should be presented to your handler. Note that parameters to attributes get handled as if 'q()' enclosed them (explicit quote marks are not needed). Furthermore, commas separate parameters (except for childStruct), so a parameter cannot contain a comma.

Output attributes

Choose none or one of these to select output behaviour.


Makes this tag behave like an '<xsp:expr>' tag, creating text nodes or inline text as appropriate. Choose this if you create plain text which may be used everywhere, including inside code. This attribute has no parameters.


Makes this tag create an XML node named name. The tag encloses all content as well as the results of the handler sub. Choose this if you want to create one XML node with all your output.


Makes this tag create a list of XML nodes named name. The tag(s) do not enclose content nodes, which become preceding siblings of the generated nodes. The return value gets converted to a node list by enclosing each element with an XML node named name. Choose this if you want to create a list of uniform XML nodes with all your output.


Makes this tag behave described under either 'node()' or 'expr', depending on the value of XML attribute attrname. If that value matches attrvalue, it will work like 'node()', otherwise it will work like 'expr'. attrname defaults to 'as', attrvalue defaults to 'node', thus leaving out both parameters means that 'as="node"' will select 'node()' behaviour. Choose this if you want to let the XSP author decide what to generate.


Like exprOrNode, selecting between 'expr' and 'nodelist()' behaviour.


Makes this tag create a more complex XML fragment. You may return a single hashref or an array of hashrefs, which get converted into an XML structure. Each hash element may contain a scalar, which gets converted into an XML tag with the key as name and the value as content. Alternatively, an element may contain an arrayref, which means that an XML tag encloses each single array element. Finally, you may use hashrefs in place of scalars to create substructures. To create attributes on tags, use a hashref that contain the attribute names prefixed by '@'. A '' (empty string) as key denotes the text contents of that node.

You can also use a XML::LibXML::Document or XML::LibXML::Node object in place of a hashref. You can, for example, simply return an XML::LibXML::Document object and it gets inserted at the current location. You may also return an array of documents/nodes, and you may even mix plain hashrefs with DOM objects as you desire.

In an expression context, passes on the unmodified return value.

Other output attributes

These may appear more than once and modify output behaviour.


Adds an XML attribute named name to all generated nodes. expr gets evaluated at run time. Evaluation happens once for each generated node. Of course, this tag only makes sense with 'node()' type handlers.

Input attributes

These tags specify how input gets handled. Most may appear more than once, if that makes sense.


Declares name as a (non-mandatory) XML attribute. All attributes declared this way get passed to your handler sub in the attribute hash (the third argument to your handler).


Declares a child tag name. It always lies within the same namespace as the taglib itself. The contents of the tag, if any, get saved in a local variable named $attr_name and made available to your code fragment. If the child tag appears more than once, the last value overrides any previous value.


Declares an attribute or child tag named name. A variable is created just like for 'child()', containing the attribute or child tag contents. If both appear, the contents of the child tag take precedence.


Makes this tag preserve contained whitespace.


Makes this tag store the enclosed content in '$_' for later retrieval in your code fragment instead of adding it to the enclosing element. Non-text nodes will not work as expected.


Marks this tag to take a complex XML fragment as input. The resulting data structure is available as %_ in your code fragment. Whitespace is always preserved.

spec has the following syntax:

  1. A spec consists of a list of tag names, separated by whitespace (not commas!).

  2. Tags may appear in any order.

  3. A tag name prefixed by '@' may appear more than once in the XML document. A tag name prefixed by '$' or without any prefix may only appear once.

  4. If a '{' follows a tag name, that tag has child tags. A valid spec and a closing '}' must follow.

  5. A tag name prefixed by '*' does not indicate an input tag but specifies the name for the text contents of the surrounding tag in the resulting data structure. Such a tag name may not bear a '{...}' block.

  6. Any tag without child tags may also appear as attribute of the parent tag.

  7. A tag name followed by one or more parameter specs in parentheses means a hash gets created with the value of the corresponding attribute (or child tag) as key. This usage does not forbid appending a '{...}' block, which would result in a nested hash.

  8. A tag name prefixed by '&' denotes a recursive structure. The tag name must appear as the name of one of the surrounding '{...}'-blocks. The innermost matching block gets chosen.


Given the following handler sub:

    set_permission : childStruct(add{@permission{$type *name} $target $comment(lang)(day)} remove{@permission{$type *name} $target})

and the following XML as input:

            <permission type="user">
            <comment lang="en" day="Sun">Test entry</comment>
            <comment lang="en" day="Wed">Test entry 2</comment>
        <remove target="/test2.html">
            <permission type="user">

then the local variable '%_' will be made available to your code fragment (returned by your set_permission handler sub). It will be initialized like this:

    %_ = (
        add => {
            permission => [
                { type => "user", name => 'foo' },
                { type => "group", name => 'bar' },
            target => '/test.html',
            comment => {
                'en' => { 'Sun' => 'Test entry', 'Wed' => 'Test entry 2' },
                'de' => { '' => 'Testeintrag' },
        remove => {
            permission => [
                { type => "user", name => 'baz' },
            target => '/test2.html',


By default, all output element nodes are placed in the same namespace as the tag library. To specify a different namespace or no namespace, the desired namespace can be placed within curly braces before the node name in an output attribute:


To specify a prefix, place it after the namespace:


For example, to create an XML node named otherNS:name and associate the prefix 'otherNS' with the namespace 'http://mydomain/NS/other/v1':


To create an XML node with no namespace, use an empty namespace:


This notation for specifying namespaces can also be used in the struct output attribute. Alternatively, the standard "xmlns" XML attribute may be used to specify namespaces. For example, the following are equivalent:

  sub sample_struct : struct {
    return "{ '{http://mydomain/NS/other/v1}otherNS:name' => 'value' }";

  sub sample_struct : struct {
    return q{{
        'otherNS:name' =>
        { '@xmlns:otherNS' => 'http://mydomain/NS/other/v1',
          '' => 'value' }

Namespace scoping in the hashref is patterned after XML documents. You may refer to previously declared namespaces by using the same prefix, and you may override previously declared namespaces with new declarations (either with the curly-braced notation or by using "xmlns" XML attributes).

To specify a default namespace for all unqualified node names in the hashref, state it as a parameter to the struct output attribute:


You may also specify a prefix:


For example, the following is equivalent to the previous example:

  sub sample_struct : struct({http://mydomain/NS/other/v1}otherNS) {
    return "{ 'name' => 'value' }";

To turn off the default namespace for all node names, use an empty namespace:

  sub sample_struct : struct({}) {
    return "{ 'name' => 'value' }";

By default, XML attributes created with the nodeAttr output attribute are not in a namespace. The curly-braced notation can be used to specify a namespace. For example:


If you are specifying more than one attribute in the same namespace, you can refer to previously declared namespaces by using the same prefix:


A prefix is required to associate a namespace with an attribute. Default namespaces (those without a prefix) do not apply to attributes and are ignored.


Refer to the Demo tag libraries included in the AxKit distribution and look at the source code of AxKit::XSP::Sessions and AxKit::XSP::Auth for full-featured examples.



Because of the use of perl attributes, SimpleTaglib will only work with Perl 5.6.0 or later. This software is already tested quite well and works for a number of simple and complex taglibs. Still, you may have to experiment with the attribute declarations, as the differences can be quite subtle but decide between 'it works' and 'it doesn't'. XSP can be quite fragile if you start using heavy trickery.

If some tags don't work as expected, try surrounding the offending tag with <xsp:content>, this is a common gotcha (but correct and intended). If you find yourself needing <xsp:expr> around a tag, please contact the author, as that is probably a bug.

If you use the '&' flag of childStruct and are reloading your taglib through Apache::StatINC or a similar method, consider installing the 'WeakRef' module from CPAN to prevent memory leaks. If you never use '&' or don't reload the taglib in the running server, this is not necessary.

TODO: to be fixed: childStruct currently does not allow hash keys to be child nodes, they must be attributes of their parent node. For example, given childStruct(text(lang)), this is valid: <text lang="en">foo</text> but this is not: <text><lang>en</lang>foo</text>

Request-time handler

TODO: This shall be enhanced in a future release.

If you pine for the TaglibHelper-style handlers that get called at request time, and you do not need the flexibility of custom-generated code fragments provided by SimpleTaglib, you can define a subroutine in your tag library to be called at request time instead of at parse time. Just place a call to your subroutine inside the code fragment returned by your handler. You can even pass it some request-time variables such as $r and $cgi. For example,

    package Your::XSP::Package;
    use Apache::AxKit::Language::XSP::SimpleTaglib;

    sub some_tag {
        my($r, $cgi, $some_param) = @_;
        # define code here to be run at request time

    package Your::XSP::Package::Handlers;

    sub some_tag : attribOrChild(some-param) node(result) {

Using attrib and childStruct together

TODO: to be fixed.

You may need a list-valued parameter to be specified by XML child tags for your tag handler, but you also want the option that a single value can be passed in as an XML attribute. For example:


  <yourNS:some_tag format="XML"/>

The 'attribOrChild' Perl attribute will not suffice here because the child tag overwrites the previous value each time instead of creating a list (format will be set to 'PDF'). What you need is a combination of 'attrib' and 'childStruct':

    sub some_tag : attrib(format) childStruct(@format) node(result) {
        my ($e, $tag, %attr) = @_;
        my $code = '';
        if ( defined $attr{format} ) {
            my $quoted = Apache::AxKit::Language::XSP::makeSingleQuoted($attr{format});
            $code .= '$_{format} = ' . $quoted . ' unless defined $_{format};';
        $code .= 'Your::XSP::Package::some_tag($r,$cgi,%_);';

This technique can be generalized to support any number of parameters. In your handler, iterate over the '%attr' hash (defined by 'attrib') and merge the values into the '%_' hash (defined by 'childStruct') inside your code fragment. Remember that parameters defined in the childStruct attribute are separated by spaces, not commas.

    sub some_tag : attrib(format,option) childStruct(@format @option) node(result) {
        my ($e, $tag, %attr) = @_;
        my $code = '';
        while ( my($key, $value) = each %attr ) {
            next unless defined $value;
            $value = Apache::AxKit::Language::XSP::makeSingleQuoted($value);
            $code .= "\$_{'_$key'} = $value unless defined \$_{'_$key'};\n"
        $code .= 'Your::XSP::Package::some_tag($r,$cgi,%_);';


Jörg Walter <>


All rights reserved. This program is free software; you can redistribute it and/or modify it under the same terms as AxKit itself.


AxKit, Apache::AxKit::Language::XSP, Apache::AxKit::Language::XSP::TaglibHelper

1 POD Error

The following errors were encountered while parsing the POD:

Around line 1321:

Non-ASCII character seen before =encoding in 'Jörg'. Assuming ISO8859-1