parsing nested unbounded XML fields with ElementTree

[Stefan Behnel]

Larry Martell, 26.11.2013 13:23:
> On Tue, Nov 26, 2013 at 2:38 AM, Stefan Behnel wrote:
>> Larry.Martell... at gmail.com, 25.11.2013 23:22:
>>> I have an XML file that has an element called "Node". These can be nested to any depth and the depth of the nesting is not known to me. I need to parse the file and preserve the nesting. For exmaple, if the XML file had:
>>>
>>> <Node Name="A">
>>>    <Node Name="B">
>>>       <Node Name="C">
>>>         <Node Name="D">
>>>           <Node Name="E">
>>>
>>> When I'm parsing Node "E" I need to know I'm in A/B/C/D/E. Problem is I don't know how deep this can be. This is the code I have so far:
>>>
>>> nodes = []
>>>
>>> def parseChild(c):
>>>     if c.tag == 'Node':
>>>         if 'Name' in c.attrib:
>>>             nodes.append(c.attrib['Name'])
>>>         for c1 in c:
>>>             parseChild(c1)
>>>     else:
>>>         for node in nodes:
>>>             print node,
>>>         print c.tag
>>>
>>> for parent in tree.getiterator():
>>>     for child in parent:
>>>         for x in child:
>>>             parseChild(x)
>>
>> This seems hugely redundant. tree.getiterator() already returns a recursive
>> iterable, and then, for each nodes in your document, you are running
>> recursively over its entire subtree. Meaning that you'll visit each node as
>> many times as its depth in the tree.
>>
>>
>>> My problem is that I don't know when I'm done with a node and I should
>>> remove a level of nesting. I would think this is a fairly common
>>> situation, but I could not find any examples of parsing a file like
>>> this. Perhaps I'm going about it completely wrong.
>>
>> Your recursive traversal function tells you when you're done. If you drop
>> the getiterator() bit, reaching the end of parseChild() means that you're
>> done with the element and start backing up. So you can simply pass down a
>> list of element names that you append() at the beginning of the function
>> and pop() at the end, i.e. a stack. That list will then always give you the
>> current path from the root node.
> 
> Thanks for the reply. How can I remove getiterator()? Then I won't be
> traversing the nodes of the tree. I can't iterate over tree. I am also
> unclear on where to do the pop(). I tried putting it just after the
> recursive call to parseChild() and I tried putting as the very last
> statement in parseChild() - neither one gave the desired result. Can
> you show me in code what you mean?

untested:

  nodes = []

  def process_subtree(c, path):
      name = c.get('Name') if c.tag == 'Node' else None
      if name:
          path.append(name)
          nodes.append('/'.join(path))

      for c1 in c:
          process_subtree(c1, path)

      if name:
          path.pop()

  process_subtree(tree.getroot(), [])


Stefan