A replacement implementation of DAP for Python is debugpy, which is based on the development branch (5.x) of ptvsd. All future development is going to be happening in debugpy; there will be no further commits to or releases of ptvsd.

If you're using the Python extension for Visual Studio Code, you should be switched to debugpy automatically, so long as you're using an up-to-date version of the extension. Please file any bug reports directly in the debugpy repository.

If you're using Visual Studio 2019 16.5 or later, you should be switched to debugpy automatically. Visual Studio 2019 16.4 or earlier, including Visual Studio 2017, are not compatible with debugpy. We recommend that you upgrade to the latest version of Visual Studio in order to migrate to debugpy.

If you're using ptvsd directly, consider migrating to debugpy. This is required for continued support from the Python extension, and to receive any bug fixes. Issues with ptvsd that can also be reproduced in debugpy should be filed directly in the debugpy repository. Any issues that are specific to ptvsd, and cannot be reproduced with debugpy, will not be fixed in ptvsd.

This debugger is based on the Debug Adapter Protocol for VS Code: debugProtocol.json

To run a script file with debugging enabled, but without waiting for the debugger to attach (i.e. code starts executing immediately):

-m ptvsd --host localhost --port 5678 myfile.py

To wait until the debugger attaches before running your code, use the --wait switch.

-m ptvsd --host localhost  --port 5678 --wait myfile.py

The --host option specifies the interface on which the debug server is listening for connections. To be able to attach from another machine, make sure that the server is listening on a public interface - using 0.0.0.0 will make it listen on all available interfaces:

-m ptvsd --host 0.0.0.0 --port 5678 myfile.py

This should only be done on secure networks, since anyone who can connect to the specified port can then execute arbitrary code within the debugged process.

To pass arguments to the script, just specify them after the filename. This works the same as with Python itself - everything up to the filename is processed by ptvsd, but everything after that becomes sys.argv of the running process.

To run a module, use the -m switch instead of filename:

-m ptvsd --host localhost --port 5678 -m mymodule

Same as with scripts, command line arguments can be passed to the module by specifying them after the module name. All other ptvsd switches work identically in this mode; in particular, --wait can be used to block execution until the debugger attaches.

The following command injects the debugger into a process with a given PID that is running Python code. Once the command returns, a ptvsd server is running within the process, as if that process was launched via -m ptvsd itself.

-m ptvsd --host localhost --port 5678 --pid 12345

At the beginning of your script, import ptvsd, and call ptvsd.enable_attach() to start the debug server. The default hostname is 0.0.0.0, and the default port is 5678; these can be overridden by passing a (host, port) tuple as the first argument of enable_attach().

import ptvsd
ptvsd.enable_attach()
...

Use the ptvsd.wait_for_attach() function to block program execution until the debugger is attached.

import ptvsd
ptvsd.enable_attach()
ptvsd.wait_for_attach()  # blocks execution until debugger is attached
...

In Python 3.7 and above, ptvsd supports the standard breakpoint() function. Use ptvsd.break_into_debugger() function for similar behavior and compatibility with older versions of Python (3.6 and below). If the debugger is attached when either of these functions is invoked, it will pause execution on the calling line, as if it had a breakpoint set. If there's no debugger attached, the functions do nothing, and the code continues to execute normally.

import ptvsd
ptvsd.enable_attach()

while True:
    ...
    breakpoint() # or ptvsd.break_into_debugger() on <3.7
    ...

{
    "debugOptions":  [
            "RedirectOutput",       // Whether to redirect stdout and stderr (see pydevd_comm.CMD_REDIRECT_OUTPUT)
            "WaitOnNormalExit",     // Wait for user input after user code exits normally
            "WaitOnAbnormalExit",   // Wait for user input after user code exits with error
            "Django",               // Enables Django Template debugging
            "Jinja",                // Enables Jinja (Flask) Template debugging
            "FixFilePathCase",      // See FIX_FILE_PATH_CASE in wrapper.py
            "DebugStdLib",          // Whether to enable debugging of standard library functions
            "StopOnEntry",          // Whether to stop at first line of user code
            "ShowReturnValue",      // Show return values of functions
    ]
}

{
    "debugOptions":  [
            "RedirectOutput",       // Whether to redirect stdout and stderr (see pydevd_comm.CMD_REDIRECT_OUTPUT)
            "Django",               // Enables Django Template debugging
            "Jinja",                // Enables Jinja (Flask) Template debugging
            "FixFilePathCase",      // See FIX_FILE_PATH_CASE in wrapper.py
            "DebugStdLib",          // Whether to enable debugging of standard library functions
            "WindowsClient",        // Whether client OS is Windows
            "UnixClient",           // Whether client OS is Unix
            "ShowReturnValue",      // Show return values of functions
    ],
    "pathMappings": [
        {
            "localRoot": "C:\\Project\\src",   // Local root  (where source and debugger running)
            "remoteRoot": "/home/smith/proj"   // Remote root (where remote code is running)
        },
        // Add more path mappings
    ]
}

To enable debugger internal logging via CLI, the --log-dir switch can be used:

-m ptvsd --log-dir path/to/logs ...

When using enable_attach, the same can be done with log_dir argument:

ptvsd.enable_attach(log_dir='path/to/logs')

In both cases, the environment variable PTVSD_LOG_DIR can also be set to the same effect.

When logging is enabled, ptvsd will create a log file with a name ptvsd-<pid>.log in the specified directory, where <pid> is the ID of the process being debugged. When subprocess debugging is enabled, a separate log is created for every subprocess.