Perl PXML intends to be a simple, Perl based representation for XML, or at least the subset that's necessary for doing most tasks. Currently it doesn't support XML namespaces properly (manually prefixing element names may be a workable solution, though?). It is primarily meant to produce XML output; parsing of XML is of secondary interest (but Htmlgen already has some code to parse by way of HTML::TreeBuilder).
Its in-memory representation are PXML::Element (or subclassed) objects. Serialization is done using functions/procedures from PXML::Serialize (also, using PXML::Preserialize if necessary for performance).
The body of elements can be a mix of standard Perl arrays, FP::PureArrays, linked lists based on FP::List, and promises (FP::Lazy, FP::Stream), the latter of which allow for the generation of streaming output (which means the document is generated while it is being serialized, thus there's no memory needed to hold the whole document at once).
Direct creation of XML elements:
use PXML::Element; my $element = PXML::Element->new ("a", {href=> "http://myserver.com"}, ["my server"]);
Using 'tag functions' for shorter code:
use PXML::XHTML; my $element= A({href=> "http://myserver.com"}, "my server");
See test and testlazy for complete examples, and examples/csv_to_xml for a simple real example, and htmlgen/gen for the program that generates this website. FP::DBI is supposed to fit well with PXML.
PXML, PXML::XHTML, PXML::HTML5, PXML::SVG, PXML::Tags, PXML::Serialize, PXML::Preserialize, PXML::Util
When generating HTML, CGI's tag functions seem similar, what are the differences?
PXML::XHTML chooses upper-case constructor names to reduce the chances for conflicts; for example using "tr" for <TR></TR> conflicts with the tr builtin Perl operator.
CGI's creators return strings, whereas PXML::XHTML returns PXML::Element objects. The former might have O(n^2) complexity with the size of documents (getting slower to concatenate big strings), while the latter should have constant overhead. Also, PXML can be inspected after creation, an option not possible with CGI (without using an XML parser).
PXML serialization always escapes strings, hence is safe against XSS, while CGI does/is not.
PXML has chosen not to support dashes on attributes, like {-href=> "foo"}, as the author feels that this is unnecessary clutter both for the eyes and for the programmer wanting to access attributes from such hashes, and added complexity/runtime cost for the serializer.
CGI's tag functions are actually deprecated now.
Perhaps PXML should be renamed to FXML. The idea behind PXML was originally to provide something similar to SXML, using Perl arrays and hashes (hence 'P' instead of 'S'), but that has proven to be pretty impractical, wrapper functions producing blessed objects is a much better user interface.