ElementTree iterparse strategy

Question

I have to handle xml documents that are big enough (up to 1GB) and parse them with python. I am using the iterparse() function (SAX style parsing).

My concern is the following, imagine you have an xml like this

<?xml version="1.0" encoding="UTF-8" ?>
<families>
  <family>
    <name>Simpson</name>
    <members>
        <name>Homer</name>
        <name>Marge</name>
        <name>Bart</name>
    </members>
  </family>
  <family>
    <name>Griffin</name>
    <members>
        <name>Peter</name>
        <name>Brian</name>
        <name>Meg</name>
    </members>
  </family>
</families>

The problem is, of course to know when I am getting a family name (as Simpsons) and when I am getting the name of one of that family member (for example Homer)

What I have been doing so far is to use "switches" which will tell me if I am inside a "members" tag or not, the code will look like this

import xml.etree.cElementTree as ET

__author__ = 'moriano'

file_path = "test.xml"
context = ET.iterparse(file_path, events=("start", "end"))

# turn it into an iterator
context = iter(context)
on_members_tag = False
for event, elem in context:
    tag = elem.tag
    value = elem.text
    if value :
        value = value.encode('utf-8').strip()

    if event == 'start' :
        if tag == "members" :
            on_members_tag = True

        elif tag == 'name' :
            if on_members_tag :
                print "The member of the family is %s" % value
            else :
                print "The family is %s " % value

    if event == 'end' and tag =='members' :
        on_members_tag = False
    elem.clear()

And this works fine as the output is

The family is Simpson 
The member of the family is Homer
The member of the family is Marge
The member of the family is Bart
The family is Griffin 
The member of the family is Peter
The member of the family is Brian
The member of the family is Meg

My concern is that with this (simple) example i had to create an extra variable to know in which tag i was (on_members_tag) imagine with the true xml examples that I have to handle, they have more nested tags.

Also note that this is a very reduced example, so you can assume that i may be facing an xml with more tags, more inner tags and trying to get different tag names, attributes and so on.

So question is. Am I doing something horribly stupid here? I feel like there must be a more elegant solution to this.

Answer 1

Here's one possible approach: we maintain a path list and peek backwards to find the parent node(s).

path = []
for event, elem in ET.iterparse(file_path, events=("start", "end")):
    if event == 'start':
        path.append(elem.tag)
    elif event == 'end':
        # process the tag
        if elem.tag == 'name':
            if 'members' in path:
                print 'member'
            else:
                print 'nonmember'
        path.pop()

Answer 2

pulldom is excellent for this. You get a sax stream. You can iterate through the stream, and when you find a node that your are interested in, load that node in to a dom fragment.

import xml.dom.pulldom as pulldom
import xpath # from http://code.google.com/p/py-dom-xpath/

events = pulldom.parse('families.xml')
for event, node in events:
    if event == 'START_ELEMENT' and node.tagName=='family':
        events.expandNode(node) # node now contains a dom fragment
        family_name = xpath.findvalue('name', node)
        members = xpath.findvalues('members/name', node)
        print('family name: {0}, members: {1}'.format(family_name, members))

output:

family name: Simpson, members: [u'Hommer', u'Marge', u'Bart']
family name: Griffin, members: [u'Peter', u'Brian', u'Meg']