I am using cElementTree.iterparse in order to parse through a huge XML
document and filter out sections of interest.

The usage pattern is that I wait for an "end" event for a element of
interest and then if it matches a some criterium I write it out using
cElementTree.tostring().

My code had bug in it because the cElementTree.tostring methods prints
the element including its tail. The element retreived from the iterparse
iterator sometimes contains the tail by the time it emits the end event
but sometimes it does not.

In my document the tail just consisted of the newline '\n' character and
about 98% of the time it was attached to the element during its end event.

This is rather confusing behavior. 

Could ElementTree/cElementTree.iterparse be changed so that if you
respond to the end event for an element its tail is never set?

"ET.tostring(elem)" works as documented.

Proposed workaround:

import copy

elem_copy = copy.copy(elem)
elem_copy.tail = ''
print ET.tostring(elem_copy)

Footnote: "iterparse" does things this way mostly to keep the implementation simple and fast; due to buffering, the tree builder are usually ahead of the event generation with up to 16k.  See the note on this page:

http://effbot.org/zone/element-iterparse.htm

and the message it links to for more on this topic.

Your case is a very common use case for "tostring", so it would probably have made sense to make "tostring" skip the tail on the element itself, at least if it's whitespace only.  Guess we could add an option...

But in your case, you can probably just nuke or normalize the "tail" element before writing it out (i.e. set it to None or "\n").