Using Python to create UNIX command line tools
A Python command line interface manifesto
Can you write a command line tool? Maybe you can, but can you write a really good command line tool? This articles covers making a robust command line tool in Python, complete with built-in help menus, error handling, and option handling. For some strange reason, it is not widely known that the standard library in Python® has all of the tools necessary to make incredible powerful *NIX command line tools.
Arguably, Python is the best language for making *NIX command line tools, period, due to its batteries-included philosophy, and its emphasis on readable code. Just a warning, though; these are dangerous ideas, when you find out how easy it is to create a command line tool in Python, you might be spoiled for life. To my knowledge, there has never been an article published in this detail on creating command line tools in Python, so I hope you enjoy it.
The optparse module in the Python Standard Library does most of the dirty work of creating a command line tool. Optparse was included in Python 2.3, so this module will be included on many *NIX Operating Systems. If for some reason, the operating system you are on does not include a module you need, it is a great comfort to know the latest versions of Python have been tested and compiles on just about any *NIX Operating Systems. Python support includes IBM® AIX®, HP-UX, Solaris, Free BSD, Red Hat Linux®, Ubuntu, OS X, IRIX, and even a Nokia phone or two.
Creating a Hello World command line tool
The first step to writing a great command line tool is to define the problem you wish to solve. This is critical to the success of your tool. It is also important to solve the problem in the simplest way possible. The KISS principle, Keep It Simple Stupid, definitely applies here. Add options and increased functionality only after the planned functionality is implemented and tested.
Let's get started by creating a Hello World command line tool. Following the advice above, let's define the problem in the simplest possible terms.
Problem Definition: I would like to create a command line tool that prints Hello World by default, but takes an option to print the name of a different person.
Given that description, this is one possible solution that comes in at just a handful of lines of code.
Hello World command line interface (CLI)
#!/usr/bin/env python import optparse def main(): p = optparse.OptionParser() p.add_option('--person', '-p', default="world") options, arguments = p.parse_args() print 'Hello %s' % options.person if __name__ == '__main__': main()
If we run this code, we get the expected output of:
Our handful of code does quite a bit more than that, though. We get a help menu for free:
python hello_cli.py --help Usage: hello_cli.py [options] Options: -h, --help show this help message and exit -p PERSON, --person=PERSON
We can see from the help menu that we have two ways to change the output of Hello World:
python hello_cli.py -p guido Hello guido
We also get error handling for free:
python hello_cli.py --name matz Usage: hello_cli.py [options] hello_cli.py: error: no such option: --name
If you have not used the optparse module in Python, you might be jumping out of your seat right now, thinking about all of the incredible tools you can write in Python. If you are new to Python, it might surprise you know that everything is this easy in Python. There is a funny comic about this very subject of "how easy Python is," by "xkcd" that I have included in the Related topics.
Creating a useful command line tool
Now that we have the basics out of the way, we can move onto creating a tool to solve a specific problem. For this example, we will be using a Python network library and interactive tool called Scapy. Scapy works on most *NIX systems, can send packets on layer 2 and layer 3, and allows you to create incredibly sophisticated tools in just a few lines of Python. If you would like to follow along from home, make sure you have the proper prerequisites installed.
Let's define a new problem to solve.
Problem: I would like to create a command line tool takes an IP address or a subnet as an argument, and return a MAC address or a list of MAC addresses with their respective IP address to standard out.
Now that we have a clearly defined problem, let's try to separate the problem in the simplest possible pieces, and solve those pieces individually. For this problem I see two separate pieces. The first piece is to write a function that takes an IP address or a subnet range, and returns a MAC address or MAC addresses. We can worry about integrating this into a command line tool after we solve this problem.
Solution Part 1: Creating a Python function to determine MAC addresses from IP addresses
from scapy import srp,Ether,ARP,conf conf.verb=0 ans,unans=srp(Ether(dst="ff:ff:ff:ff:ff:ff")/ARP(pdst="10.0.1.1"), timeout=2) for snd, rcv in ans: print rcv.sprintf(r"%Ether.src% %ARP.psrc%")
The output of the command is:
sudo python arping.py 00:00:00:00:00:01 10.0.1.1
Note that doing operations with scapy requires escalated privileges, so we must use sudo. I also change the real output to include a fake MAC address for the purpose of this article. Now that we have proven we can find out a MAC address from an IP address. We need clean up this code to take an IP address or subnet and return a MAC address and IP address pair.
#!/usr/bin/env python from scapy import srp,Ether,ARP,conf def arping(iprange="10.0.1.0/24"): conf.verb=0 ans,unans=srp(Ether(dst="ff:ff:ff:ff:ff:ff")/ARP(pdst=iprange), timeout=2) collection =  for snd, rcv in ans: result = rcv.sprintf(r"%ARP.psrc% %Ether.src%").split() collection.append(result) return collection #Print results values = arping() for ip,mac in values: print ip,mac
As you can see, we have a function that takes an IP address or network and returns a nested list of IP/MAC addresses. We are now ready for part two, which is creating a command line interface for our tool.
Solution Part 2: Creating a command line tool from our arping function
In this example, we combine ideas from the previous portions of the article to make a complete command line tool that solves our original specification.
#!/usr/bin/env python import optparse from scapy import srp,Ether,ARP,conf def arping(iprange="10.0.1.0/24"): """Arping function takes IP Address or Network, returns nested mac/ip list""" conf.verb=0 ans,unans=srp(Ether(dst="ff:ff:ff:ff:ff:ff")/ARP(pdst=iprange), timeout=2) collection =  for snd, rcv in ans: result = rcv.sprintf(r"%ARP.psrc% %Ether.src%").split() collection.append(result) return collection def main(): """Runs program and handles command line options""" p = optparse.OptionParser(description=' Finds MAC Address of IP address(es)', prog='pyarping', version='pyarping 0.1', usage='%prog [10.0.1.1 or 10.0.1.0/24]') options, arguments = p.parse_args() if len(arguments) == 1: values = arping(iprange=arguments) for ip, mac in values: print ip, mac else: p.print_help() if __name__ == '__main__': main()
A few points of clarification on the script above might help our understanding of how optparse works.
First, an instance of optparse.OptionParser() must be created and it takes the optional parameters shown above:
description, prog, version, and usage
These should be self explanatory for the most part, but I want to make sure that it is understood that while optparse is incredible, it is not magic. It has a clearly defined interface that enables rapid creation of command line tools.
Second, in the line:
options, arguments = p.parse_args()
this is where values for options and arguments get divided into distinct bits. In the above code, there will be exactly one argument that we expect, so I have specified that there must be only one argument value, and that value gets passed into our arping function.
if len(arguments) == 1: values = arping(iprange=arguments)
To clarify even further, let's run the command and see how this works:
sudo python arping.py 10.0.1.1 10.0.1.1 00:00:00:00:00:01
In the above example, the argument is 10.0.1.1 and because there is only one argument, as I specified in the conditional statement, it is passed into the arping function. If we had options, they would be handed to options in the options, arguments = p.parse_args() idiom. Let's look at what happens when we break the intended use case of our command line tool and assign it two arguments:
sudo python arping.py 10.0.1.1 10.0.1.3 Usage: pyarping [10.0.1.1 or 10.0.1.0/24] Finds MAC Address or IP address(es) Options: --version show program's version number and exit -h, --help show this help message and exit
Because of how I structured the conditional statement for arguments, it will automatically bring up a help menu if the number of arguments is below one or greater than one:
if len(arguments) == 1: values = arping(iprange=arguments) for ip, mac in values: print ip, mac else: p.print_help()
This is an important way to control how your tool works as you can use the number of arguments, or name of a specific option as the mechanism to control the flow of your command line tool. Since we have covered creating options in the very first Hello World example, let's add a couple of options to our command line tool by changing the main function slightly:
arping CLI main function
def main(): """Runs program and handles command line options""" p = optparse.OptionParser(description='Finds MAC Address of IP address(es)', prog='pyarping', version='pyarping 0.1', usage='%prog [10.0.1.1 or 10.0.1.0/24]') p.add_option('-m', '--mac', action ='store_true', help='returns only mac address') p.add_option('-v', '--verbose', action ='store_true', help='returns verbose output') options, arguments = p.parse_args() if len(arguments) == 1: values = arping(iprange=arguments) if options.mac: for ip, mac in values: print mac elif options.verbose: for ip, mac in values: print "IP: %s MAC: %s " % (ip, mac) else: for ip, mac in values: print ip, mac else: p.print_help()
The main thing to take away from the changes is that we have created conditional statements based on whether an option has been specified or not. Notice, unlike the Hello World command line tool, we are only using the options as a true/false signal to our tool. In the case of the --MAC option, if it is specified, our conditional elif statement only prints the MAC address.
Here is the output of the new options:
sudo python arping2.py Password: Usage: pyarping [10.0.1.1 or 10.0.1.0/24] Finds MAC Address of IP address(es) Options: --version show program's version number and exit -h, --help show this help message and exit -m, --mac returns only mac address -v, --verbose returns verbose output [ngift@M-6][H:11184][J:0]> sudo python arping2.py 10.0.1.1 10.0.1.1 00:00:00:00:00:01 [ngift@M-6][H:11185][J:0]> sudo python arping2.py -m 10.0.1.1 00:00:00:00:00:01 [ngift@M-6][H:11186][J:0]> sudo python arping2.py -v 10.0.1.1 IP: 10.0.1.1 MAC: 00:00:00:00:00:01
Advanced studies in creating command line tools
Here are a few more ideas for further study. They will be covered in great detail in the book I am co-authoring on Python for *NIX Systems Administration, which is due out in mid-2008.
Using subprocess module in a command line tool
The subprocess module is included with Python 2.4 or greater and is a unified interface for dealing with system calls and processes. You could easily replace the arping function above to use an arping tool available to your specific *NIX operating system. Here is a rough example of what that could look like:
import subprocess import re def arping(ipaddress="10.0.1.1"): """Arping function takes IP Address or Network, returns nested mac/ip list""" #Assuming use of arping on Red Hat Linux p = subprocess.Popen("/usr/sbin/arping -c 2 %s" % ipaddress, shell=True, stdout=subprocess.PIPE) out = p.stdout.read() result = out.split() pattern = re.compile(":") for item in result: if re.search(pattern, item): print item arping()
Here is the output of this function run standalone:
[root@localhost]~# python pyarp.py
Notice the use of subprocess to grab the output of the arping command, and the use of a compiled regular expression match to the MAC address. Note, if you are on Python 2.3, you could use the popen module instead of subprocess, which became available in Python 2.4 or greater.
Using an Object Relational Mapper, such as SQLAlchemy or Storm with SQLite, in a command line tool
Another potential option for a command line tool is to use an ORM, or Object Relational Mapper, to store records for data generated by a command line tool. There are quite a few ORM's available for Python, but SQLAlchemy and Storm happen to be two of the most popular. I flipped a coin and decided to use Storm for this example:
Storm ORM arping
#!/usr/bin/env python import optparse from storm.locals import * from scapy import srp,Ether,ARP,conf class NetworkRecord(object): __storm_table__ = "networkrecord" id = Int(primary=True) ip = RawStr() mac = RawStr() hostname = RawStr() def arping(iprange="10.0.1.0/24"): """Arping function takes IP Address or Network, returns nested mac/ip list""" conf.verb=0 ans,unans=srp(Ether(dst="ff:ff:ff:ff:ff:ff")/ARP(pdst=iprange), timeout=2) collection =  for snd, rcv in ans: result = rcv.sprintf(r"%ARP.psrc% %Ether.src%").split() collection.append(result) return collection def main(): """Runs program and handles command line options""" p = optparse.OptionParser() p = optparse.OptionParser(description='Finds MACAddr of IP address(es)', prog='pyarping', version='pyarping 0.1', usage= '%prog [10.0.1.1 or 10.0.1.0/24]') options, arguments = p.parse_args() if len(arguments) == 1: database = create_database("sqlite:") store = Store(database) store.execute("CREATE TABLE networkrecord " "(id INTEGER PRIMARY KEY, ip VARCHAR,\ mac VARCHAR, hostname VARCHAR)") values = arping(iprange=arguments) machine = NetworkRecord() store.add(machine) #Creates Records for ip, mac in values: machine.mac = mac machine.ip = ip #Flushes to database store.flush() #Prints Record print "Record Number: %r" % machine.id print "MAC Address: %r" % machine.mac print "IP Address: %r" % machine.ip else: p.print_help() if __name__ == '__main__': main()
The main thing to look at in this example is the creation of a class called NetworkRecord that maps to an "in memory" SQLite database. In the main function, I changed the output of our arping function to map to our record objects, flushed them to the database, then pulled them back out again to print the results. This is obviously not a production tool, but an instructive example of the steps involved in using an ORM with our tool.
Integrating config files into the CLI
Python INI config syntax
[AIX] MAC: 00:00:00:00:02 IP: 10.0.1.2 Hostname: aix.example.com [HPUX] MAC: 00:00:00:00:03 IP: 10.0.1.3 Hostname: hpux.example.com [SOLARIS] MAC: 00:00:00:00:04 IP: 10.0.1.4 Hostname: solaris.example.com [REDHAT] MAC: 00:00:00:00:05 IP: 10.0.1.5 Hostname: redhat.example.com [UBUNTU] MAC: 00:00:00:00:06 IP: 10.0.1.6 Hostname: ubuntu.example.com [OSX] MAC: 00:00:00:00:07 IP: 10.0.1.7 Hostname: osx.example.com
Next we need to parse this using the ConfigParser module:
#!/usr/bin/env python import ConfigParser def readConfig(file="config.ini"): Config = ConfigParser.ConfigParser() Config.read(file) sections = Config.sections() for machine in sections: #uncomment line below to see how this config file is parsed #print Config.items(machine) macAddr = Config.items(machine) print machine, macAddr readConfig()
The output of this function is:
OSX 00:00:00:00:07 SOLARIS 00:00:00:00:04 AIX 00:00:00:00:02 REDHAT 00:00:00:00:05 UBUNTU 00:00:00:00:06 HPUX 00:00:00:00:03
I will leave the rest of the problem solving as an exercise to the reader. What I would do next is to integrate this config file into my script such that I could compare an inventory of machines kept in my config file against the actual inventory of MAC addresses that appear in the ARP cache. An IP address or a hostname is only so useful in tracking down a machine, but our potential tool could be a very useful way to track the hardware address of a machine on your network and determine if it was on the network.
We started with a few lines of code to create a very basic but powerful Hello World command line tool. We then moved on to create a sophisticated network tool using a Python Networking Library. Finally, we proceeded to look at more advanced areas of study for the motivated reader. In the advanced research section, we looked at integrating the subprocess module, Object Relational Mappers, and finally configuration files.
Although it is apparently a secret, anyone with an IT background can create a command line tool with Python with a little effort. I hope this article has inspired you to go out on your own and create the next revolutionary command line tool.
- Read Noah Gift's PyCon 2008 talk on command line tools.
- See David Beazly's PyCon 2008 tutorial on using generators for system's programming.
- Visit the official Scapy website to find product documentation.
- The Scapy portability page provides installation information for various operating system
- Scapy Presentation Article in Science and Engineering
- The Style Guide for Python Code gives coding conventions for the Python code comprising the standard library in the main Python distribution.
- Python For Bash Scripters shows you how to write a function in Python that you would normally write in Bash.
- Visit The ConfigParser module for more configuration documentation.
- See Config Parsing Examples on using ConfigParser.
- The Subprocess Module lets you to spawn new processes, connect to their input/output/error pipes, and obtain their return codes.
- Python 2.3 Popen Module
- Using Net-SNMP and IPython (Noah Gift, developerWorks, February 2008) shows you how to use Net-SNMP, Python, and the IPython shell to interactively explore and manage a network.
- Unittests supports test automation, sharing of setup and shutdown code for tests, aggregation of tests into collections, and independence of the tests from the reporting framework.
- Doctest Module searches for pieces of text that look like interactive Python sessions.
- Python Is Easy Comic by XKCD.COM
- SQLAlchemy is the Python SQL toolkit and Object Relational Mapper.
- Storm is an object-relational mapper (ORM) for Python.