Table of contents
We’re Earthly. We make building software simpler and therefore faster. If you’re interested in a simple and containerized approach to CI that works well with Python and everything else then check us out.
Python ships with built-in modules such as os and sys that provide some functionality to interact with the underlying operating system. However, it may sometimes be more convenient to run system programs from within a Python script. Python’s subprocess module provides ready-to-use functionality to run external commands, capture and process outputs, redirect output to files and I/O streams, and much more!
This guide will help you get started with the subprocess module in Python. By the end of this tutorial, you’ll have learned how to:
- Run Bash commands inside Python scripts
- Analyze what happens when an external command fails
- Pipe the output of one subprocess to the input of another subprocess
- Change the environment in which a subprocess runs
Python Subprocess Module: The Basics
Prerequisites: To follow along with this tutorial, you should have Python 3.7 or a later version installed on your machine.
⚠️ If you’re on a Windows machine, we recommend using Git Bash or installing the Windows Subsystem for Linux (WSL) to code along. If you prefer using other code editors or IDEs on Windows, you should run the Windows equivalents of some of the Bash commands used in this tutorial.
subprocess module is built into the Python standard library, so you can import it into your working environment:
How Does Subprocess Execution Occur in Python?
After importing the
subprocess module, you can call the
run() function with the syntax:
subprocess.run(command) to run an external command. This call to the
run() function invokes a subprocess.
When a subprocess is invoked, the following actions occur under the hood:
- The Python script inside which you’re running the external command is the parent process and the subprocess is the child process.
- Upon starting a subprocess, the control is transferred from the parent process to the child process.
- The child process executes the external command inside a secondary environment.
- After the child process finishes executing, the control returns to the parent process.
How to Run a Subprocess With
subprocess.run() function with an external command or program as the argument prints out the output of the command onto the console. The function call also returns a
CompletedProcess object which has a set of useful attributes.
As our first subprocess, let’s run the
ls command that lists the contents of the working directory:
= subprocess.run("ls") process_1 print(process_1) print(process_1.args) print(process_1.returncode) print(process_1.stdout)
py_logging main.py py_unit_testing string_manipulation CompletedProcess(args='ls', returncode=0) ls 0 None
In the above output:
- The contents of the working directory are printed onto the console.
argsattribute contains the arguments passed to the
returncodeattribute indicates whether the command ran successfully; A return code of zero indicates successful execution.
- If the output of the command is captured, the
stdoutattribute contains the output; if we do not capture the output,
Subprocess Popen Constructor
In the subprocess module, the Popen class handles the creation and management of subprocesses. However, the
run() function is recommended for invoking subprocesses and suffices for most common use cases.
Running Bash Commands With Arguments
To run Bash commands with arguments, you can specify the name of the command followed by the arguments that it takes. The name of the command and its arguments should be passed in as a list of strings.
We’ll now run the
ls command to list the contents of the
= subprocess.run(["ls","py_logging"]) process_2 print(process_2) print(process_2.args) print(process_2.returncode) print(process_2.stdout)
custom_logger.py logger.py CompletedProcess(args=['ls', 'py_logging'], returncode=0) ['ls', 'py_logging'] 0 None
You can also specify the command as a long string and call Python’s
split() method on it. By default, the
split() method splits the string on all occurrences of whitespace and returns a list of strings. This list of strings can be used in the call to the
= ... command = command.split()command_list
To parse more complex commands—where whitespace is not the correct separator to split on—use
shlex.split() that splits the command string using a shell-like syntax.
import shlex = ... command shlex.split(command)
Let’s take an example:
Suppose you’d like to create a directory named ‘Foo Bar’. The code cells below show how the command string is split when we use the
split() method and
>>> command_str = "mkdir 'Foo Bar'" >>> command_str.split() 'mkdir', "'Foo", "Bar'"][
As seen, the
split() method splits
command_str on all whitespaces. When run as a subprocess, the above command will create two directories,
Bar', which is not what we want!
>>> import shlex >>> shlex.split(command_str) 'mkdir', 'Foo Bar'][
shlex.split() returns the command list
['mkdir', 'Foo Bar']. Running this command list as a subprocess creates the ‘Foo Bar’ directory.
When you set the
shell parameter to
True, you can pass in the command as a single string—without splitting it into a list of strings.
= subprocess.run("ls py_logging",shell=True) process_3 print(process_3) print(process_3.args) print(process_3.stdout)
custom_logger.py logger.py CompletedProcess(args='ls py_logging', returncode=0) ls py_logging None
⚠️ If you set
shell = True, an instance of the underlying shell is used to run the command. Because this is susceptible to shell injection attacks, avoid setting
shell = True to run commands that take in user inputs.
How to Capture and Redirect Subprocess Outputs
subprocess.run(command) prints the output of the command onto the console, and the
stdout attribute is
None by default if we don’t capture the output. When you run Bash commands such as
chmod to change file permissions or the
sleep command, you don’t need to process the output. However, you may sometimes need to capture the outputs and use them in next steps in your program.
To capture the output of a command for further processing, you can set the
capture_output argument to
True when calling the
= subprocess.run("ls",capture_output=True) process_4 print(process_4.stdout)
Note that the output of
process_4 is not printed onto the console anymore. The
stdout attribute of the
process_4 contains the output as a string of bytes.
You can call the
decode() method on the value of
stdout to access the output as a normal Python string.
= subprocess.run("ls",capture_output=True) process_4 print(process_4.stdout.decode())
py_logging py_subprocess.py py_unit_testing string_manipulation
If you set
True, the output is captured as a Python string, thereby eliminating the need for decoding the
= subprocess.run("ls",capture_output=True,text=True) process_5 print(process_5.stdout)
py_logging py_subprocess.py py_unit_testing string_manipulation
📋 A Note on Subprocess Pipe
Under the hood, setting
True redirects both the output and the errors to
subprocess.PIPE. Therefore, you can also set
subprocess.PIPE when calling the
= subprocess.run("ls",stdout=subprocess.PIPE,text=True) process_6 print(process_6.stdout)
py_logging main.py py_unit_testing string_manipulation
How to Redirect Subprocess Output to a File
You can redirect the output of a subprocess to a text file by setting the
stdout attribute to a valid file object. In this example, the contents of the working directory, including the
contents.txt file are listed in the
with open('contents.txt','w') as f_obj: "ls",stdout=f_obj,text=True) subprocess.run(
What Happens When a System Program Fails?
In the examples we’ve coded thus far, the external programs ran successfully. In practice, we’ll run into errors if the external programs do not exist in our development environment or are called with invalid arguments. For example, trying to run a git commit as an external command (subprocess) when I don’t have Git installed.
returncode attribute tells whether or not the execution of the external command was successful. A non-zero
returncode indicates an error in running the command. The
stderr attribute contains information on the error.
When we try to list the contents of a directory that doesn’t exist, the
returncode of the
process_7 is 2 (non-zero) and its
stderr attribute contains the output as a string of bytes.
= subprocess.run(["ls","non-existent-directory"],\ process_7 =True) capture_outputprint(process_7.returncode) print(process_7.stderr)
2 b"ls: cannot access 'non-existent-directory': No such file or directory\n"
You can also capture errors in the
stderr attribute by explicitly setting the
stderr argument to
subprocess.PIPE when calling the
= subprocess.run(["ls","non-existent-directory"],\ process_8 =subprocess.PIPE) stderrprint(process_8.stderr)
b"ls: cannot access 'non-existent-directory': No such file or directory\n"
To check if the external command ran successfully, here’s a summary of methods we’ve learned so far:
- Check the
returncodeattribute of the
CompletedProcessobject. A non-zero return code indicates that there was an error running the external command.
- Access the
stderrattribute of the
CompletedProcessobject. The error message in
stderrhas diagnostic information on what went wrong when the external command was run.
- If you anticipate that an external command may fail, you can explicitly set the
subprocess.PIPEto redirect errors to
In all of the above methods, Python does not throw an error when the external command fails. However, it might help to get an error message when running the Python script if the external command fails to run successfully. For this, you can set the optional
check parameter to
True in the
run() function call. If the external command fails to run, Python throws a
= subprocess.run(["ls","non-existent-directory"],\ process_9 =True,check=True)capture_output
Traceback (most recent call last): File "main.py", line 38, in <module> process_9 = subprocess.run(["ls","non-existent-directory"],\ capture_output=True,check=True) File "/usr/lib/python3.8/subprocess.py", line 516, in run raise CalledProcessError(retcode, process.args, subprocess.CalledProcessError: Command '['ls', \ 'non-existent-directory']' returned non-zero exit status 2.
How to Raise a
We’ve learned that when a subprocess is invoked, the control is transferred from the parent process (Python script) to the child process that runs the command.
The control returns to the parent process only after the child process finishes execution which can sometimes be challenging. For example, the subprocess may involve pinging a URL when you’re working offline. In this case, the parent process waits for transfer of control from the child process; but the child process never finishes its execution.
To address this, you can set a
timeout value (in seconds) when invoking a subprocess. If the subprocess does not finish executing within the value specified by the
timeout argument, a
TimeoutExpired exception is raised.
In this example, the Bash command
sleep 10 requires 10 seconds to finish but setting
timeout to 2 seconds raises the
= subprocess.run(["sleep","10"],capture_output=True,\ process10 =True,timeout=2)check
Traceback (most recent call last): File "main.py", line 40, in <module> process_10 = subprocess.run(["sleep","10"],capture_output=True,\ check=True,timeout=2) File "/usr/lib/python3.8/subprocess.py", line 495, in run stdout, stderr = process.communicate(input, timeout=timeout) File "/usr/lib/python3.8/subprocess.py", line 1028, in communicate stdout, stderr = self._communicate(input, endtime, timeout) File "/usr/lib/python3.8/subprocess.py", line 1869, in _communicate self._check_timeout(endtime, orig_timeout, stdout, stderr) File "/usr/lib/python3.8/subprocess.py", line 1072, in _check_timeout raise TimeoutExpired( subprocess.TimeoutExpired: Command '['sleep', '10']' timed \ out after 2 seconds
The above traceback for the
TimeoutExpired exception is difficult to parse. We can use Python’s
except statements to handle the
The general syntax to use
try-except statements is as follows:
try: # doing this except ThisError: # do this when ThisError occurs
TimeoutExpired exception occurs, we now get a simple error message instead of the hard-to-parse traceback.
try: = subprocess.run(["sleep","10"],capture_output=True,\ process10 =True,timeout=2) checkexcept subprocess.TimeoutExpired: print("subprocess timed out")
subprocess timed out
TimeoutExpired Exception Doesn’t Kill the Child Process
TimeoutExpired exception raised—when a subprocess fails to complete executing within a specified duration—does not implicitly kill the child process. If there are many such subprocesses in a Python script, it would help to kill the child processes after they’ve timed out.
For this, we need to use the
Popen constructor to run the external command. The
timeout value is passed to the
communicate() method when called on the subprocess object. This method returns the
stderr values as a tuple. When the
TimeoutExpiredexception is raised upon timeout, the child process is killed and the communication is deemed complete.
= subprocess.Popen(["sleep","10"]) p1 try: = p1.communicate(timeout=1) std_out,std_err except subprocess.TimeoutExpired: p1.kill()= p1.communicate() std_out,std_err
How to Pipe Subprocess Outputs to Inputs of Subprocesses
You may sometimes need to chain commands such that the output of one process is the input to the next command in the chain. We know that we can capture the output of a command in the
stdout attribute. The
run() function provides an optional
input parameter which you can set to the
stdout of another subprocess. This way, you can pipe subprocess outputs to the input of another subprocess.
Consider the following example, where we’d like to run each of the commands as a subprocess. The output of the first subprocess is the input to the second.
- Run the
catcommand to display the contents of a text file.
- Search through the file’s contents for the occurrence of a specific string using the
= subprocess.run(["cat","sample.txt"],capture_output=True,\ process1 =True) text= subprocess.run(["grep","-n","Python"],capture_output=True,\ process2 =True,input=process1.stdout) textprint(process_2.stdout)
3:with Python programming is to 4:start a Python REPL and work your way through simple exercises!
How to Change the Environment of a Subprocess
You can change the environment in which a subprocess runs by setting the
env parameter to a modified environment. The secondary environment in which a subprocess runs is inherited from the environment of the parent process. This is the default behavior and the
env parameter is set to its default value
None. To change this default execution environment, you should set the
env parameter to the modified environment variable dictionary, when calling the
In Python, the
environ() function in the
os module returns the environment variables dictionary. Instead of modifying the underlying environment variables directly, you can create a copy of the dictionary and update one or more environment variables.
In the following example, we update the
PATH environment variable by adding a new directory and set the
env parameter to
import os = os.environ.copy() new_env "PATH"] = os.pathsep.join(["/testapp/",new_env["PATH"]]) new_env[= subprocess.run(["ls"],env=new_env)process_now
I hope this tutorial helped you understand how to use Python’s
subprocess module. Here’s a summary of what you’ve learned.
- To run an external command within a Python script, use
commandis a string or a list of strings (when running commands with arguments). The
run()function returns a
- You can set
capture_output = Trueto capture the output as a string of bytes in the
stdoutattribute, which you can decode by calling the
text = Trueeliminates the decoding step. You can redirect output to files by setting the
stdoutargument to a valid file object.
- When an external program fails to run, the
returncodeis non-zero and the error information is available in the
stderrattribute of the
- Setting the
CalledProcessErrorwhen the external command fails. You can optionally set the
timeoutparameter to raise a
TimeoutExpiredexception when the subprocess fails to complete execution within a specified time-frame.
- To change the environment in which a subprocess runs, set the
envparameter to the modified environment. You can update one or more environment variables by creating a copy of the environment variables dictionary.
If you need to run external programs within the Python application, try to run them as subprocesses. You can also leverage Python’s threading capabilities to run subprocesses concurrently.