how to remove an element in lxml

You're looking for the remove function. Call the tree's remove method and pass it a subelement to remove.

import lxml.etree as et

xml="""
<groceries>
  <fruit state="rotten">apple</fruit>
  <fruit state="fresh">pear</fruit>
  <punnet>
    <fruit state="rotten">strawberry</fruit>
    <fruit state="fresh">blueberry</fruit>
  </punnet>
  <fruit state="fresh">starfruit</fruit>
  <fruit state="rotten">mango</fruit>
  <fruit state="fresh">peach</fruit>
</groceries>
"""

tree=et.fromstring(xml)

for bad in tree.xpath("//fruit[@state='rotten']"):
    bad.getparent().remove(bad)

print et.tostring(tree, pretty_print=True)

Result:

<groceries>
  <fruit state="fresh">pear</fruit>
  <fruit state="fresh">starfruit</fruit>
  <fruit state="fresh">peach</fruit>
</groceries>

Use the remove method of an xmlElement :

tree=et.fromstring(xml)

for bad in tree.xpath("//fruit[@state=\'rotten\']"):
  bad.getparent().remove(bad)     # here I grab the parent of the element to call the remove directly on it

print et.tostring(tree, pretty_print=True, xml_declaration=True)

If I had to compare with the @Acorn version, mine will work even if the elements to remove are not directly under the root node of your xml.

I met one situation:

<div>
    <script>
        some code
    </script>
    text here
</div>

div.remove(script) will remove the text here part which I didn't mean to.

following the answer here, I found that etree.strip_elements is a better solution for me, which you can control whether or not you will remove the text behind with with_tail=(bool) param.

But still I don't know if this can use xpath filter for tag. Just put this for informing.

Here is the doc:

strip_elements(tree_or_element, *tag_names, with_tail=True)

Delete all elements with the provided tag names from a tree or subtree. This will remove the elements and their entire subtree, including all their attributes, text content and descendants. It will also remove the tail text of the element unless you explicitly set the with_tail keyword argument option to False.

Tag names can contain wildcards as in _Element.iter.

Note that this will not delete the element (or ElementTree root element) that you passed even if it matches. It will only treat its descendants. If you want to include the root element, check its tag name directly before even calling this function.

Example usage::

   strip_elements(some_element,
       'simpletagname',             # non-namespaced tag
       '{http://some/ns}tagname',   # namespaced tag
       '{http://some/other/ns}*'    # any tag from a namespace
       lxml.etree.Comment           # comments
       )

As already mentioned, you can use the remove() method to delete (sub)elements from the tree:

for bad in tree.xpath("//fruit[@state=\'rotten\']"):
  bad.getparent().remove(bad)

But it removes the element including its tail, which is a problem if you are processing mixed-content documents like HTML:

<div><fruit state="rotten">avocado</fruit> Hello!</div>

Becomes

<div></div>

Which is I suppose what you not always want :) I have created helper function to remove just the element and keep its tail:

def remove_element(el):
    parent = el.getparent()
    if el.tail.strip():
        prev = el.getprevious()
        if prev:
            prev.tail = (prev.tail or '') + el.tail
        else:
            parent.text = (parent.text or '') + el.tail
    parent.remove(el)

for bad in tree.xpath("//fruit[@state=\'rotten\']"):
    remove_element(bad)

This way it will keep the tail text:

<div> Hello!</div>