I'm working on a function that determines if the content of a given html element - el - in an lxml ElementTree is the leading content of a line in a rendered HTML page. To do this, I'm trying to find the right-most block level element that is left of el, and then determine if there is content between these two.
I figure this can happen via a traversal in the reverse order of a DFS, with the reverse traversal starting at el. But I've also been trying to find if a simpler method exists using lxml or xpath to do this. So far I've found ways to find elements that are ancestors or left siblings of a given element with some criteria, but I haven't spotted anything that works on the entire tree right (or left) of a specific node.
Does anybody know of a simpler way to do this search using lxml or xpath?
Example
<html>
<body class="first">
root
<!-- A span that does not have its own content, but does have several levels of children-->
<span>
<a>
<b>
<h1 class="first">
A block level that is the decendant of several non block levels
</h1>
</b>
</a>
<span class="first" id="tricky">
A non-block level that has no block levels among its ancestors, but a block level element among its left cousins
</span>
<span>
A non-block level that has no block levels among its ancestors, and content between itself and its nearest left-cousin block level
</span>
</span>
<div class="first">
a block level
</div>
<div>
<span class="first">first content in a non block level in a block level</span>
<span>following content in a non block level in a block level</span>
</div>
<div>
<i> </i><bclass="first">a non block level that contains the first content within a block level, but follows an empty non-block level</b>
</div>
</body>
</html>
In the above I've added a "first" class to any element that, when rendered, would appear to present the leading content of a line. Of particular interest is the element with id "tricky", because that element will present the first content of a line even though none of its ancestors nor its siblings are block level elements. "tricky" will be on a new line because a descendant of one of its siblings (the h1) is a block level, and there is no other content following that h1.
Follow Up At this point I have written a function in Python that does a type of backwards traversal. Its a bit complicated, but it seems to work:
block_level = {'blockquote','br','dd','div','dl','dt','h1','h2','h3','h4','h5','h6','hr','li','ol','p','pre','td','ul'}
# Returns true if the content of the provided element is the leading content of a line
# This function runs on HTML elements before any translation occurs
# Here 'content' refers to non-whiespace characters
def is_first_in_line_html(self, el):
# This element contains no content, so it can't be the leading content of a line.
if el.text is None or el.text.strip() == '': return False
# This element has content and is a block level, so its content is the leading content of a line.
if el.tag in block_level: return True
# This element has content, is not a block level, and is the body element. Definitely leading content of a line.
if el.tag == 'body': return True
# Final case - is there content between the present element and the nearest block level element to the left of the present
# element.
def traverse_children(element, bound_text):
children = element.iterchildren(reversed=True)
for child in children:
if child.tail is not None: bound_text = child.tail + bound_text
if bound_text.strip() != '': return False
if child.tag in block_level: return bound_text.strip() == ''
rst_children = traverse_children(child, bound_text)
if rst_children is not None: return rst_children
if child.text is not None: bound_text = child.text + bound_text
if bound_text.strip() != '': return False
return None
def traverse_left_sibs_and_ancestors(element, bound_text):
left_sibs = element.itersiblings(preceding=True)
for sib in left_sibs:
if sib.tail is not None: bound_text = sib.tail + bound_text
if bound_text.strip() != '': return False
if sib.tag in block_level: return bound_text.strip() == ''
rst_children = traverse_children(sib, bound_text)
if rst_children is not None: return rst_children
if sib.text is not None: bound_text = sib.text + bound_text
if bound_text.strip() != '': return False
parent = element.getparent()
if parent.tail is not None: bound_text = parent.tail + bound_text
if parent.tag == 'body': return bound_text.strip() == ''
if parent.tag in block_level: return bound_text.strip() == ''
return traverse_left_sibs_and_ancestors(parent)
return traverse_left_sibs_and_ancestors(el, '')