Many people dread XML and would rather work with JSON. However, XPATH can be extremely powerful when dealing with the Juniper XML API output.

Even though the Juniper API can be made to return JSON instead of XML, I prefer sticking to XML. With little code, I tend to be able to find exactly what I am looking for.

This blog contains some of my commonly used snippets of code that leverage XPATH to deal with the RPC returned by the Juniper XML API.

Code that was used to execute the RPC:

The data I will be working with in the article is retrieved using the following code:

from lxml import etree
from jnpr.junos import Device
import getpass
import sys

username = sys.argv[1]
host = sys.argv[2]
password = getpass.getpass()

with Device(host=host, user=username, password=password, normalize=True) as dev:                                  
    rpc = dev.rpc.get_interface_information(extensive=True)

How the RPC was mapped from the Junos CLI command:

The rpc that is referred to in the rest of the examples is the device response to get_interface_information(extensive=True). This is the RPC to retrieve the information that can be displayed with the show interfaces extensive command:

admin@ar.lab> show interfaces extensive | display xml rpc 
<rpc-reply xmlns:junos="">

I stored the example script as and ran it with python -i. This will drop you in the interpreter and allows you to paste in the following snippets and examples.

Examples on how to use XPATH to navigate the RPC return:

XPATH getting all the interface names:


XPATH getting all the logical interface names:


XPATH getting all the interface names of the physical as well as the logical interfaces:


XPATH getting all the physical and logical interface XML data:


Using XPATH, we slice the XML into a list of items. To see the content of the items we can use the following:

for interface_xml in rpc.xpath('//physical-interface|//physical-interface/logical-interface'):    

That works, but it does not look very pretty. A (somewhat involved) way to make it look pretty:

import xml.dom.minidom
for interface_xml in rpc.xpath('//physical-interface|//physical-interface/logical-interface'):
    xml_minidom = xml.dom.minidom.parseString(etree.tostring(interface_xml))
    interface_xml_pretty = xml_minidom.toprettyxml()

Even easier then using the above would be to just check the XML on the device using show interfaces extensive | display xml.

XPATH getting the physical interface names in case they include ‘et’:

rpc.xpath('physical-interface[contains(name, "et-")]/name/text()')

XPATH getting the logical interface names in case they include ‘irb’:

rpc.xpath('physical-interface/logical-interface[contains(name, "irb")]/name/text()') 

Using regular expresions in lxml:

ns = {"re": ""}
# RE to get 100G:
rpc.xpath('physical-interface[re:match(name, "et")]/name/text()', namespaces=ns) 

# RE to get 100G on line card position 1 and 2:
rpc.xpath('physical-interface[re:match(name, "et-[1,2]")]/name/text()', namespaces=ns)

# RE to get 100G on line card position 1 and 2 while ignoring upper and lower case:
rpc.xpath('physical-interface[re:match(name, "ET-[1,2]", "i")]/name/text()', namespaces=ns)

XPATH to get all interfaces, physical as well as logical:

all_interfaces = rpc.xpath('//physical-interface|//physical-interface/logical-interface')

Now we can iterate all_interfaces to get the name of the interfaces using the find method:

for interface in all_interfaces:
    # extract the name using find:
    interface.find('./name').text if interface.find('./name') is not None else None

The if interface.find('./name') is not None else None was added to avoid running into exceptions.

Alternatively, we use an XPATH to accomplish the same thing again:

for interface in all_interfaces:
    # extract the name using XPATH:

We can also inspect the text of all the child nodes, just to see what is on offer:

for interface in all_interfaces:

Sometimes, the attributes offer more interesting data. To access these attributes, we can use the following:

for interface in all_interfaces:
    interface.find('./interface-flapped').attrib['seconds'] if interface.find('./interface-flapped') is not None else None

Safely accessing attributes of can also be done using the Python built-in getattr:

all_interfaces = rpc.xpath('//physical-interface|//physical-interface/logical-interface')
for interface in all_interfaces:
    getattr(interface.find('./name'), 'text' , '')
    getattr(interface.find('./description'), 'text' , '')    
    getattr(interface.find('./oper-status'), 'text', None)
    getattr(interface.find('./mtu'), 'text', None)
    interface.find('./interface-flapped').attrib['seconds'] if interface.find('./interface-flapped') is not None else None    
    getattr(interface.find('.//input-pps'), 'text' , '')
    getattr(interface.find('.//output-pps'), 'text' , '')

The getattr(object, attribute, default) returns the value of a specified attribute if it exists. If it does not exist, a default value is returned.

Short example where we put the collected information into a dictionary:

# to ensure ordering of dict key/values remains the same, not needed starting python 3.9
from collections import OrderedDict 

all_interfaces = rpc.xpath('//physical-interface|//physical-interface/logical-interface')
interfaces_list_of_dict = []

for interface in all_interfaces:
        'interface-name' : getattr(interface.find('./name'), 'text' , ''),
        'interface-description' : getattr(interface.find('./description'), 'text' , ''),
        'interface-status' : getattr(interface.find('./oper-status'), 'text', None),
        'interface-mtu' : getattr(interface.find('./mtu'), 'text', None),
        'interface-last-flapped' : interface.find('./interface-flapped').attrib['seconds'] if interface.find('./interface-flapped') is not None else None,
        'interface-input-pps' : getattr(interface.find('.//input-pps'), 'text' , ''),
        'interface-output-pps' : getattr(interface.find('.//output-pps'), 'text' , ''),


To look at what was gathered, use json.dumps (printing OrderedDicts looks silly):

from json import dumps
print(dumps(interfaces_list_of_dict, indent=4))

There are a lot of ‘uninteresting’ interfaces that Juniper has for internal use. Can be done with multiple letnghty XPATHs, but sometimes using any to filter them out is easier:

uninteresting = ["jsrv", "local", "igb", "ixlv", "bme", "32768", "16384", "32767"]

for interface in all_interfaces:
    if any(s in getattr(interface.find('./name'), 'text' , '') for s in uninteresting):
    if len(interface.xpath('./address-family/interface-address/ifa-local/text()')):
        getattr(interface.find('./name'), 'text' , '')

I have recorded some sample output right here and here.

More examples and explanation on how to work with lxml can be found here.


Python 3.8.3 was used with the following package versions:

junos-eznc            2.5.4    
lxml                  4.6.2    

The Juniper devices I used were the following:

QFX 15.1X53-D65.3
MX 16.1R3-S8     
MX10003 17.4R2-S11