Python debug configurations in Visual Studio Code

The Python extension supports debugging of a number of types of Python applications. For a short walkthrough of basic debugging, see Tutorial - Configure and run the debugger. Also see the Flask tutorial. Both tutorials demonstrate core skills like setting breakpoints and stepping through code.

For general debugging features such as inspecting variables, setting breakpoints, and other activities that aren't language dependent, review VS Code debugging.

This present article addresses only those considerations that are specific to Python, mainly Python-specific debugging configurations, including the necessary steps for specific app types and remote debugging.

Initialize configurations

A configuration drives VS Code's behavior during a debugging session. Configurations are defined in a launch.json file that's stored in a .vscode folder in your workspace.

To initialize debug configurations, first select the Debug View in the sidebar:

Debug icon

If you don't yet have any configurations defined, you'll see "No Configurations" in the drop-down list, and a dot on the settings icon:

Debug toolbar settings command

To generate a launch.json file with Python configurations, do the following steps:

  1. Select the settings button (circled in the image above) or use the Debug > Open configurations menu command.

  2. In the Select Environment drop-down list that appears, select Python.

  3. The Python extension then creates and opens a launch.json file that contains number of pre-defined configurations. You can modify configurations (to add arguments, for example), and also add custom configurations.

The details of configuration properties are covered later in this article under Standard configuration and options. Additional configurations are also described in this article under Debugging specific app types.

Choose a configuration

To select a debugging configuration, select the Debug View in the sidebar, then select an option of your choice from the drop-down list:

Selecting a debug configuration

Note: Previous versions of the Python extension used slightly different names for the configurations as shown in the above graphic, but they generally work the same.

By default, VS Code shows only the most common configurations provided by the Python extension. You can select other configurations to include in launch.json by using the Add Configuration command shown in the list and in the launch.json editor. When you use the command, VS Code prompts you with a list of all available configurations (be sure to scroll down to see all the Python options):

Adding a new Python debugging configurations

See Debugging specific app types for details on all of these configurations.

While debugging, the Status Bar shows the current configuration on the lower left, with the current debugging interpreter to the right. Selecting the configuration brings up the list from which you can choose a different configuration:

Debugging Status Bar

By default, the debugger uses the same python.pythonPath workspace setting as for other features of VS Code. To use a different interpreter for debugging specifically, set the value for pythonPath in the applicable debugger configuration. Alternately, select the named interpreter on the Status Bar to select a different one.

Set configuration options

When you first create launch.json, there are two standard configurations that run the active file in the editor in either the integrated terminal (inside VS Code) or the external terminal (outside of VS Code):

{
    "name": "Python: Current File (Integrated Terminal)",
    "type": "python",
    "request": "launch",
    "program": "${file}",
    "console": "integratedTerminal"
},
{
    "name": "Python: Current File (External Terminal)",
    "type": "python",
    "request": "launch",
    "program": "${file}",
    "console": "externalTerminal"
}

The specific settings are described in the following sections. You can also add other settings, such as args, that aren't included in the standard configurations.

Tip: It's often helpful in a project to create a configuration that runs a specific startup file. For example, if you always want to always launch startup.py with the arguments --port 1593 whenever you start the debugger, create a configuration entry as follows:

{
    "name": "Python: startup.py",
    "type": "python",
    "request": "launch",
    "program": "${workspaceFolder}/startup.py",
    "args" : ["--port", "1593"]
},

name

Provides the name for the debug configuration that appears in the VS Code drop-down list.

type

Identifies the type of debugger to use; leave this set to python for Python code.

request

Specifies the mode in which to start debugging:

  • launch: start the debugger on the file specified in program
  • attach: attach the debugger to an already running process. See Remote debugging for an example.

program

Provides the fully qualified path to the python program's entry module (startup file). The value ${file}, often used in default configurations, uses the currently active file in the editor. By specifying a specific startup file, you can always be sure of launching your program with the same entry point regardless of which files are open. For example:

"program": "/Users/Me/Projects/PokemonGo-Bot/pokemongo_bot/event_handlers/__init__.py",

You can also rely on a relative path from the workspace root. For example, if the root is /Users/Me/Projects/PokemonGo-Bot then you can use the following:

"program": "${workspaceFolder}/pokemongo_bot/event_handlers/__init__.py",

pythonPath

Points to the Python interpreter to be used for debugging, which can be a folder containing a Python interpreter. The value can use variables like ${workspaceFolder} and ${workspaceFolder}/.venv.

If not specified, this setting defaults to the interpreter identified in the python.pythonPath setting, which is equivalent to using the value ${config:python.pythonPath}. To use a different interpreter, specify its path instead.

You can specify platform-specific paths by placing pythonPath within a parent object named osx, windows, or linux. For example, the configuration for PySpark uses the following values:

"osx": {
    "pythonPath": "^\"\\${env:SPARK_HOME}/bin/spark-submit\""
},
"windows": {
    "pythonPath": "^\"\\${env:SPARK_HOME}/bin/spark-submit.cmd\""
},
"linux": {
    "pythonPath": "^\"\\${env:SPARK_HOME}/bin/spark-submit\""
},

Alternately, you can use a custom environment variable that's defined on each platform to contain the full path to the Python interpreter to use, so that no additional folder paths are needed.

args

Specifies arguments to pass to the Python program. Each element of the argument string that's separated by a space should be contained within quotes, for example:

"args": ["--quiet", "--norepeat", "--port", "1593"],

stopOnEntry

When set to true, breaks the debugger at the first line of the program being debugged. If omitted (the default) or set to false, the debugger runs the program to the first breakpoint.

console

Specifies how program output is displayed.

Value Where output is displayed
"none" VS Code debug console
"integratedTerminal" (default) VS Code Integrated Terminal
"externalTerminal" Separate console window

cwd

Specifies the current working directory for the debugger, which is the base folder for any relative paths used in code. If omitted, defaults to ${workspaceFolder} (the folder open in VS Code).

As an example, say ${workspaceFolder} contains a py_code folder containing app.py, and a data folder containing salaries.csv. If you start the debugger on py_code/app.py, then the relative paths to the data file vary depending on the value of cwd:

cwd Relative path to data file
Omitted or ${workspaceFolder} data/salaries.csv
${workspaceFolder}/py_code ../data/salaries.csv
${workspaceFolder}/data salaries.csv

debugOptions

An array of additional options that may contain the following:

Option Description
"RedirectOutput" (default) Causes the debugger to print all output from the program into the VS Code debug output window. If this setting is omitted, all program output is not displayed in the debugger output window. This option is typically omitted when using "console": "integratedTerminal" or "console": "externalTerminal" because there's no need to duplicate the output in the debug console.
"DebugStdLib" Enabled debugging of standard library functions.
"Django" Activates debugging features specific to the Django web framework.
"Sudo" When used with "console": "externalTerminal", allows for debugging apps that require elevation. Using an external console is necessary to capture the password.
"Pyramid" When set to true, ensures that a Pyramid app is launched with the necessary pserve command.

env

Sets optional environment variables for the debugger process beyond system environment variables, which the debugger always inherits.

envFile

Optional path to a file that contains environment variable definitions. See Configuring Python environments - environment variable definitions file.

gevent

If set to true, enables debugging of gevent monkey-patched code.

Invoking a breakpoint in code

In your Python code, you can call breakpoint() at any point where you want to pause the debugger during a debugging session.

Attach to a local script

In some scenarios you need to debug a Python script that's invoked locally by another process. For example, you may be debugging a web server that runs different Python scripts for specific processing jobs. In such cases, you need to attach the VS Code debugger to the script once it's been launched:

  1. Run VS Code, open the folder or workspace containing the script, and create a launch.json for that workspace if one doesn't exist already.

  2. In the script code, add the following and save the file:

    import ptvsd
    
    # 5678 is the default attach port in the VS Code debug configurations
    print("Waiting for debugger attach")
    ptvsd.enable_attach(address=('localhost', 5678), redirect_output=True)
    ptvsd.wait_for_attach()
    breakpoint()
    
  3. Open a terminal using Terminal: Create New Integrated Terminal, which activates the script's selected environment.

  4. In the terminal, install the ptvsd package with python -m pip install --upgrade ptvsd.

  5. In the terminal, start Python with the script, for example, python3 myscript.py. You should see the "Waiting for debugger attach" message that's included in the code, and the script halts at the ptvsd.wait_for_attach() call.

  6. Switch to the Debug view, select Python: Attach from the debugger drop-down list, and start the debugger.

  7. The debugger should stop on the breakpoint() call, from which point you can use the debugger normally. You can, of course, set other breakpoints in the script code instead of using breakpoint().

Note: In the future, the Python extension will support attaching the debugger by process ID instead of host and port. See Issue 1078.

Remote debugging

Remote debugging allows you to step through a program locally within VS Code while it runs on a remote computer. It is not necessary to install VS Code on the remote computer.

  1. Both computers: make sure that identical source code is available.

  2. Both computers: install ptvsd using python -m pip install --upgrade ptvsd into your environment (while using a form of virtual environment is not required, it is strongly recommended).

  3. Remote computer: open the port you wish to use for debugging in the appropriate firewall or other networking configuration.

  4. Remote computer: there are two ways to specify how to attach to the remote process. Do note that you may need to specify the remote computer's private IP address, if applicable (Linux VMs on Azure, for example, have both a public and private address). If you use the public IP address, you might see the error "Cannot assign requested address".

    1. In the source code, add the following lines, replacing address with the remote computer's IP address and port number (IP address 1.2.3.4 is shown here for illustration only).

      import ptvsd
      
      # Allow other computers to attach to ptvsd at this IP address and port.
      ptvsd.enable_attach(address=('1.2.3.4', 3000), redirect_output=True)
      
      # Pause the program until a remote debugger is attached
      ptvsd.wait_for_attach()
      

      The IP address used in enable_attach should be the remote computer's private IP address. You can then launch the program normally, causing it to pause until the debugger attaches.

    2. Launch the remote process through ptvsd, for example:

      python3 -m ptvsd --host 1.2.3.4 --port 3000 -m myproject
      

      This starts the package myproject using python3, with the remote computer's private IP address of 1.2.3.4 and listening on port 3000 (you can also start the remote Python process specifying a file path instead of using -m).

  5. Local computer: Only if you modified the source code on the remote computer as outlined above, then in the source code, add a commented-out copy of the same code added on the remote computer. Adding these lines makes sure that the source code on both computers matches line by line.

    #import ptvsd
    
    # Allow other computers to attach to ptvsd at this IP address and port.
    #ptvsd.enable_attach(address=('1.2.3.4', 3000), redirect_output=True)
    
    # Pause the program until a remote debugger is attached
    #ptvsd.wait_for_attach()
    
  6. Local computer: switch to Debug View in VS Code, select the Python: Attach configuration, then select the settings (gear) icon to open launch.json to that configuration.

  7. Local computer: Modify the configuration so that remoteRoot provide the location of the program on the remote computer's file system. Also modify host and port to match the values in the ptvsd.enable_attach call added to the source code, except that you need to use the remote computer's public IP address for host. You might also change name to specifically identify the configuration. For example:

    {
        "name": "Python Attach (Remote Debug 192.168.34.156)",
        "type": "python",
        "request": "attach",
        "pathMappings": [
            {
                "localRoot": "${workspaceFolder}",  // You may also manually specify the directory containing your source code.
                "remoteRoot": "~/hello" // Linux example; adjust as necessary for your OS and situation.
            }
        ],
        "port": 3000,                   // Set to the remote port.
        "host": "1.2.3.4"               // Set to your remote host's public IP address.
    }
    
  8. Local computer: set a breakpoint in the code where you want to start debugging.

    Tip: setting a single breakpoint on the statement immediately following the ptvsd.wait_for_attach() line may not work. Set at least one other breakpoint on another statement.

  9. Local computer: start the VS Code debugger using the modified Python Attach configuration. VS Code should stop on your locally-set breakpoints, allowing you to step through the code, examine variables, and perform all other debugging actions. Expressions that you enter in the Debug Console are run on the remote computer as well.

    Text output to stdout, as from print statements, appear on both computers. Other output, such as graphical plots from a package like matplotlib, however, appear only on the remote computer.

  10. During remote debugging, the debugging toolbar appears as below:

    Debugging toolbar during remote debugging

    On this toolbar, the disconnect button (⇧F5 (Windows, Linux Shift+F5)) stops the debugger and allows the remote program to run to completion. The restart button (⇧⌘F5 (Windows, Linux Ctrl+Shift+F5)) restarts the debugger on the local computer but does not restart the remote program. You use the restart button only when you've already restarted the remote program and need to reattach the debugger.

Debugging over SSH

In some cases you may want or need to use a secure connection to the remote computer when debugging. On Windows computers, you may need to install OpenSSH to have the ssh command.

On the remote computer:

  1. Enable port forwarding by opening the sshd_config config file (found under /etc/ssh/ on Linux and under %programfiles(x86)%/openssh/etc on Windows) and adding or modifying the following setting:

    AllowTcpForwarding yes
    
  2. Restart the SSH server. On Linux/macOS, run sudo service ssh restart; on Windows, run services.msc, locate and select OpenSSH in the list of services, then select Restart.

  3. Start the Python program and let it wait for the debugger to attach as described in the previous section.

On the local computer:

  1. Create an SSH tunnel by running ssh -L sourceport:localhost:destinationport user@remoteaddress, using a selected port for destinationport and the appropriate username and the remote computer's IP address in user@remoteaddress. For example, to use port 3000 on IP address 1.2.3.4, the command would be ssh -L 3000:localhost:3000 user@1.2.3.4.

  2. Verify that you can see a prompt in the SSH session.

  3. In VS Code, set the port in the debug configuration of launch.json to match the port used in the ssh command and set the host to localhost. You use localhost here because you've set up the SSH tunnel.

  4. Launch the program and attach the debugger as described in the previous section.

Debugging specific app types

The configuration drop-down provides a variety of different options for general app types:

Configuration Description
Attach See Remote debugging in the previous section.
Django Specifies "program": "${workspaceFolder}/manage.py", "args": ["runserver", "--noreload", "--nothreading"], and "console": "integratedTerminal". Also adds "django": true to enable debugging of Django HTML templates. Note that automatic reloading of Django apps is not possible while debugging.
Flask See Flask debugging below.
Gevent Adds "gevent": true to the standard integrated terminal configuration.
Pyramid Removes program, adds "args": ["${workspaceFolder}/development.ini"], adds "jinja": true for enabling template debugging, and adds "pyramid": true to ensure that the program is launched with the necessary pserve command.
PySpark Runs the program using PySpark instead of the default interpreter, using platform-specific values for pythonPath as shown earlier under the pythonPath option.
Scrapy Specifies "module": "scrapy" and adds "args": ["crawl", "specs", "-o", "bikes.json"].
Watson Specifies "program": "${workspaceFolder}/console.py" and "args": ["dev", "runserver", "--noreload=True"]

Specific steps are also needed for remote debugging and Google App Engine. For details on debugging unit tests (including nosetest), see Unit testing.

To debug an app that requires administrator privileges, use "console": "externalTerminal" and include "Sudo" in debugOptions.

Flask debugging

{
    "name": "Python: Flask",
    "type": "python",
    "request": "launch",
    "module": "flask",
    "env": {
        "FLASK_APP": "app.py"
    },
    "args": [
        "run",
        "--no-debugger",
        "--no-reload"
    ],
    "jinja": true
},

As you can see, this configuration specifies "env": {"FLASK_APP": "app.py"} and "args": ["run", "--no-debugger","--no-reload"]. The "module": "flask" property is used instead of program. (You may see "FLASK_APP": "${workspaceFolder}/app.py" in the env property, in which case modify the configuration to refer to only the filename. Otherwise you may see "Cannot import module C" errors where C is a drive letter.)

The "jinja": true setting also enables debugging for Flask's default Jinja templating engine.

If you want to run Flask's development server in development mode, use the following configuration:

{
    "name": "Python: Flask (development mode)",
    "type": "python",
    "request": "launch",
    "module": "flask",
    "env": {
        "FLASK_APP": "app.py",
        "FLASK_ENV": "development"
    },
    "args": [
        "run"
    ],
    "jinja": true
},

Google App Engine debugging

Google App Engine launches an app by itself, so launching it in the VS Code debugger isn't directly possible. What's required instead is to use ptvsd in the app and then launch Google App Engine in a mode that allows VS Code to attach its debugger.

Debugging Google App Engine

  1. Download ptvsd and extract its files into a ptvsd folder in your working folder. (If using a different folder, modify the path in the pydev_startup.py file created in step 4).

  2. Create a tasks.json file with the following contents:

    {
        "version": "2.0.0",
        "tasks": [
            {
                "label": "Launch Google App Engine",
                "command": "python",
                "type": "shell",
                "args": [
                    "/usr/local/google_appengine/dev_appserver.py",
                    "--python_startup_script=${workspaceFolder}/pydev_startup.py",
                    "--automatic_restart=no",
                    "--max_module_instances=default:1",
                    "${workspaceFolder}/app.yaml"
                ]
            }
        ]
    }
    
  3. On Windows and Linux, replace the first item in args with the path to wherever Google App Engine is installed (the path shown in the source code above is for macOS).

  4. Create a file named pydev_startup.py in your project root with the following contents, modified as noted:

    import sys
    import os
    
    # Assuming that pdvsd is located in the working folder
    sys.path.append(os.getcwd())
    
    import ptvsd
    # Modify the port number as desired; you're debugging locally so the values don't matter.
    # However, be sure the port is not blocked on your computer.
    ptvsd.enable_attach(address=('0.0.0.0', 3000), redirect_output=True)
    
    # The debug server has started and you can now use VS Code to attach to the application for debugging
    print("Google App Engine has started, ready to attach the debugger")
    
  5. Create a launch.json configuring using the Attach (Remote Debug) configuration as a template. Make sure the port value match what's in the source code above.

  6. Add "preLaunchTask": "python" to launch.json.

  7. From the Command Palette, run the Run Build Task command. This opens the Tasks output window where you see various messages.

  8. Once you see the message "Google App Engine has started, ready to attach the debugger", start the VS Code debugger using the remote debugging configuration.

  9. Set breakpoints where you want, then start the browser to start the app.

Troubleshooting

There are many reasons why the debugger may not work. Oftentimes the debug console reveals specific causes, but two specific reasons are as follows:

  • The path to the python executable is incorrect: check the value of pythonPath in your user settings.

  • Invalid expressions in the watch window (see example below): clear all expressions from the Watch window and restart the debugger.

    Traceback (most recent call last):
      File ".../visualstudio_py_debugger.py", line 1646, in loop
        cmd()
      File ".../visualstudio_py_debugger.py", line 1918, in command_execute_code
        thread.run_on_thread(text, cur_frame, eid, frame_kind, repr_kind)
      File ".../visualstudio_py_debugger.py", line 1246, in run_on_thread
        self.schedule_work(lambda : self.run_locally(text, cur_frame, execution_id, frame_kind, repr_kind))
      File ".../visualstudio_py_debugger.py", line 1238, in schedule_work
        self.unblock()
      File ".../visualstudio_py_debugger.py", line 1234, in unblock
        self._block_lock.release()
    RuntimeError: release unlocked lock
    

Next steps