I’ve been playing with Aaron Swarz’s XML Tramp as an intuitive/pythonic way of processing XML. The model/syntax isn’t explicitly documented, but from the source and (mostly) examples, this is what I’ve figured out:

XML Tramp Conventions

• Elements have a list of children ‘[]’
  • iterate over children: for child in doc
  • get first and second (splice) child: doc[0:2]
  • get named child element book: doc.book or doc['book']
• Elements have a dict of attributes ‘()’
  • test for an attribute COLOR: if 'COLOR' in doc.book()
  • get attribute COLOR value: doc.book('COLOR')
  • assign attribute value: doc.book(COLOR='blue')
• Namespaces (NS) are indicated with a period, the period indicates its not a literal value (as with quotes), but a namespace corresponding to an arbitrary but specified prefix
  • NS qualified elments now appear unquoted within brackets: doc[ns.book]
  • NS qualified attributes now appear unquoted within parenthesis: doc(ns.COLOR)
• To reserialize an object use print doc.__repr__(True)

Ported/Archived Responses

Joseph Reagle on 2003-11-18

Yep. It’s not like COLOR is a (pythonic/object) attribute of some namespace instance. I think the pythonic approach is one of consistency, why remove quotes when dealing with a NS qualified element or attribute? Oh well, unless I convince Aaron, or change it for my own version, your preference will persist I presume!

I also really dig XPath (used it heavily in the C14N implementations), but again have found the “set-up” in the various toolkits to be a big pain… (What do you use?) I’ve recently been wondering if I could add a simple XPath query on top of XML Tramp.

anders on 2003-11-20

libxml2 has a pretty sweet XPath API. see: http://www.xmldatabases.org/WK/blog/607?t=item

anders on 2003-11-16

i think the dot notation is more pythonic, and the point does seem to be to make XML data accessible through a more native feeling API.

are you really suggesting doc(‘ns:COLOR’) instead of doc(ns.COLOR)?

personally, i find XPath to be intuitive enough and don’t really desire anything more pythonic, but if i did, i think i would agree with Schwarz’s approach.