We can remove elements using Element.remove(). Let’s say we want to remove all countries with a rank higher than 50:

>>>
>>> for country in root.findall('country'):
...     rank = int(country.find('rank').text)
...     if rank > 50:
...         root.remove(country)
...
>>> tree.write('output.xml')

When the original xml has over 2 country with  rank>50,and they are one by one neighborly siblings element,the upper code will delete the 1st 3rd 5th and more odd No. country.
A proper example should be:
index=0
while index < len(root.findall("./*")):
    rank = int (root[index].find("rank").text)
    if rank>50:
        root.remove(root[index])
        index=index+0
        continue
    index=index+1


I think "for each in list" should not work by index,but should work by pointer like thing,could be a list of pointers.
A finial solution should be like this --- when the "for each in list" was acting,the pointers list would be fixed,and you need not to worry about the "list" changing.

#new code to avoid an err
index=0
count=len(root.findall("./*"))
while index <count:
    rank = int (root[index].find("rank").text)
    if rank>50:
        root.remove(root[index])
        index=index+0
        count=count-1  # avoid index err
        continue
    index=index+1

My fault.
"for country in root.findall('country')“ is not working as same as "for country in root"
all 3 method below:

import xml.etree.ElementTree as ET 
xmlstr=\
r'''<?xml version="1.0"?>
<data>
    <country name="Liechtenstein">
        <rank updated="yes">2</rank>
        <year>2008</year>
        <gdppc>141100</gdppc>
        <neighbor name="Austria" direction="E"/>
        <neighbor name="Switzerland" direction="W"/>
    </country>
    <country name="Singapore">
        <rank updated="yes">5</rank>
        <year>2011</year>
        <gdppc>59900</gdppc>
        <neighbor name="Malaysia" direction="N"/>
    </country>
    <country name="Panama1">
        <rank updated="yes">69</rank>
        <year>2011</year>
        <gdppc>13600</gdppc>
        <neighbor name="Costa Rica" direction="W"/>
        <neighbor name="Colombia" direction="E"/>
    </country>
    <country name="Panama2">
        <rank updated="yes">69</rank>
        <year>2011</year>
        <gdppc>13600</gdppc>
        <neighbor name="Costa Rica" direction="W"/>
        <neighbor name="Colombia" direction="E"/>
    </country>
</data>'''
print(xmlstr)

#orginal code
root = ET.fromstring(xmlstr)
for country in root.findall('country'):
    rank = int(country.find('rank').text)
    if rank > 50:
        root.remove(country)
print("___orginal___")
for country in root.findall('country'):
        print (country.get("name"))
print("^^^orginal^^^^\n")

#wrong code in my mind
root = ET.fromstring(xmlstr)
for country in root:
    rank = int(country.find('rank').text)
    if rank > 50:
        root.remove(country)
print("___bad___")
for country in root.findall('country'):
        print (country.get("name"))
print("^^^bad^^^^\n")

#my code
root = ET.fromstring(xmlstr)
index=0
count=len(root.findall("./*"))
while index <count:
    rank = int (root[index].find("rank").text)
    if rank>50:
        root.remove(root[index])
        index=index+0
        count=count-1  # avoid index err
        continue
    index=index+1
print("___new___")
for country in root.findall('country'):
        print (country.get("name"))
print("^^^new^^^^\n")

The docs should specially tell that when need root.remove(child) must works with "for child  in root.findall()".

And my code with "while if continue" could make root.insert(index,newchild) easily.

As you've seen, the example is correct. I made the same mistake earlier today.

For others: see also #41891 for a suggestion to improve the documentation.

As was pointed out in that issue, it's generally true in Python that you should not mutate a sequence while iterating over it.

Could I say the mutable sequence containing not the object but the pointer like C++.
So they can changed in def functions.

Closing as duplicate of issue 41892. Let's keep the discussion there.