XML::Toolkit Tools to Ease the Pain

Chris Prather

HD Publishing Group

Who Am I?

• Perl Developer
• Moose Developer
• Closet XML Lover

The Backstory

• Customer had an Flash site
• Flash was controlled by XML
• “Build us a CMS for our XML!”

The Backstory (cont.)

• Aha! I can whip out my XML Knowledge!
  • This was naughty, I didn’t check CPAN
  • XML::Compile
  • XML::Pastor
• Not all bad
  • my wheel is round in a different way
• I’ll get to this later

The Backstory (cont.)

• Me: Send me the DTD or Schema, this is cake!
• Them: … what’s a DTD?
• Me: Well send me your XML, if it’s well formed I can deal with that too
• Them: … Well formed?
• Me: … [redacted]

The Backstory (cont.)

• After much hassleling between Stevan and them
  • And me and stevan
• XML::Toolkit was born

The Requirements

• Had to Round Trip existing documents
• Elements in the same order because they weren’t using XPath
• Had to infer the schema from existing documents
• Had to be able to generate new documents that fit the above

The Architecture

• SAX
  • SAX ➢ Moose Class
  • SAX ➢ Moose Object
  • Moose ➢ XML (via SAX)

SAX ➢ Moose Class

• xmltk generate --input [XML]
• Parse XML
  • For every Element generate a class
  • Attributes
  • Children become a collection

SAX ➢ Moose Class

• Won’t work properly with mixed elements
• Doesn’t accurately reflect the real schema
  • “Only One Child”
  • One or Many
  • Data Types
• Works pretty damn well for an 80%

SAX ➢ Moose Object

• xmltk validate --input [XML]
• Parse XML
  • For every Element create an object
  • populate Attributes
  • Push children into collections

Moose ➢ XML (via SAX)

• xmltk validate --input [XML]
• Parse Moose Objects in memory
  • For every object fire start_element
  • for every text attribute fire characters
  • Let XML::SAX::Writer do the real work

What’s Next

• Finish Namespace Support
  • Serialization
• DTD / RelaxNG support
• Performance enhancements
• Class::XML rewrite

Colophon

• Titling: Goudy Bookletter @ 1911 48pt
• Body: Goudy Bookletter 1911 @ 24pt
• Code: Inconsolata @ 24pt
• Presentation: Markdown + S5