insert method in ElementTree

[mirandacascade]

O/S: Win2K
Vsn of Python: 2.4

Example:

<a>
<b createAnotherWhenCondition="x">
<c>text for c</c>
<d>text for d</d>
</b>
<e>
<f>text for f</f>
<g>text for g</g>>
</e>
<h>
<i>text for i</i>
<j createAnotherWhenCondition="y">
<k>text for k</k>
<l>text for l</l>
</j>
<m>text for m</m>
</h>
</a>

Python script reads XML document above into ElementTree. The script
outputs an XML document that is based on the XML document above. The
output XML will contain all of the elements in the XML document above.
If there is a parent element that does not have the
createAnotherWhencondition attribute, then that element and its
descendants will be part of the output XML. If there is a parent
element with atribute createAnotherWhenCondition and that condition is
true, then the output XML should contain another instance of the parent
element below the original instance. For example if x were true when
the script ran, the output XML document would be:

<a>
<b>
<c>text for c</c>
<d>text for d</d>
</b>
<b>
<c>text for c</c>
<d>text for d</d>
</b>
<e>
<f>text for f</f>
<g>text for g</g>>
</e>
<h>
<i>text for i</i>
<j>
<k>text for k</k>
<l>text for l</l>
</j>
<m>text for m></m>
</h>
</a>

The example attempts to illustrate that the createAnotherWhenCondition
attribute may appear in parent elements that are at different levels in
the hierarchy; the <belement is at the 2nd level in the hierarchy,
and the <jelement is at the 3rd level. There will never be
'nesting', i.e. a parent element with the
createAnotherWhenCondition attribute that is a descendant of a parent
element with a createAnotherWhenCondition attribute.

I'm pretty sure I can figure out how to create the output XML by
creating a second XML document. It would be created by iterating
through the input XML. When a new element is encountered, an element
or subelement would be created in the output XML. When a parent
element with the createAnotherWhenCondition is encountered and that
condition is true, I think I can figure out how to propagate another
instance of that parent element and all its descendants.

My request for advice is this: instead of creating a second XML
document that represents the output, would it be possible to expand the
input XML document as needed? I was thinking that the program could
iterate through all the elements. As it is iterating, it would check
for the createAnotherWhenCondition attribute. If encountered and if
the condition were true, the program would:
- make a copy of the parent element (perhaps with copy.copy)
- use the insert method to insert the just-created copy
Where I'm struggling is figuring out what the index argument should
be in the insert method. Using the example above

# assume rootElement is the root of the input XML
xList = rootElement.getiterator()
idx = 0
for x in xList:
# mix of pseudo-code and Python code
if (this element has createAnotherWhenCondition attribute)
and
(y is true):
jcopy = copy.copy(x)
??.insert(??, jcopy)
idx = idx + 1

If the program were run when y was true, then I believe it would
encounter the <jelement when idx has a value of 9. Conceptually, I
think that jcopy should be inserted after the <jelement and before
the <melement. But I'm not sure that 9 (the index in the list
created from rootElement.getiterator()) has any relevance for this
insert task. Assuming that I want to insert jcopy after the <j>
element and before the <melement:
a) would the insert need to be done relative to the <helement, which
is the parent of <j>
b) if so, would the index argument in the insert method be relative
index within the <helement?

[Stefan Behnel]

mi************@yahoo.com wrote:

Where I'm struggling is figuring out what the index argument should
be in the insert method. Using the example above

# assume rootElement is the root of the input XML
xList = rootElement.getiterator()
idx = 0
for x in xList:
# mix of pseudo-code and Python code
if (this element has createAnotherWhenCondition attribute)
and
(y is true):
jcopy = copy.copy(x)
??.insert(??, jcopy)
idx = idx + 1

ElementTree does not have parent pointers. You have to look one element level
ahead to add the child. You can use a parent stack for this, although
getiterator() is not very helpful in that case as it does not tell you when it
backtracks.

If you want to use lxml instead, you can either
* access the parent directly (element.getparent()) and add the child there
or
* use the iterwalk() function to add the child when backtracking up the tree or
* implement the whole thing in XSLT (with a custom evaluator function in Python).

There may also be other ways to do it. Just give it a try:
http://codespeak.net/lxml/
http://cheeseshop.python.org/pypi/lxml/1.1alpha

Stefan

[Carl Banks]

mi************@yahoo.com wrote:

My request for advice is this: instead of creating a second XML
document that represents the output, would it be possible to expand the
input XML document as needed? I was thinking that the program could
iterate through all the elements. As it is iterating, it would check
for the createAnotherWhenCondition attribute. If encountered and if
the condition were true, the program would:
- make a copy of the parent element (perhaps with copy.copy)
- use the insert method to insert the just-created copy
Where I'm struggling is figuring out what the index argument should
be in the insert method.

I recommend not using the insert method; instead, build a new list of
elements and set the parent to the new list. Another problem is that
you have to have the parent node around, since children don't indicate
their parents. Using getiterator won't work. I recommend a recursive
function instead. Something like this should do it:

def expand_children(parent):
if len(parent) == 0:
return
newchildren = []
for child in parent:
expand_children(child) # recursively process child nodes
if <you're supposed to create another child>:
<remove createAnotherWhenCondition attr>
newchildren.append(child) # or copy.copy(child)
newchildren.append(child)
parent[:] = newchildren # slice assign

So, basically, for each node, you build a list of children in parallel,
making duplicates as necessary, then use slice assignment to set the
parent's children to be the new list. No more mucking around with
indexes.
Carl Banks