pdb++, a drop-in replacement for pdb
What is it?
This module is an extension of the pdb module of the standard library. It is meant to be fully compatible with its predecessor, yet it introduces a number of new features to make your debugging experience as nice as possible.
pdb++ features include:
- colorful TAB completion of Python expressions (through fancycompleter)
- optional syntax highlighting of code listings (through Pygments)
- sticky mode
- several new commands to be used from the interactive
(Pdb++)prompt- smart command parsing (hint: have you ever typed
rorcat the prompt to print the value of some variable?)- additional convenience functions in the
pdbmodule, to be used from your program
pdb++ is meant to be a drop-in replacement for pdb. If you find some unexpected behavior, please report it as a bug.
Installation
Since pdb++ is not a valid package name the package is named pdbpp:
$ pip install pdbpp
pdb++ is also available via conda:
$ conda install -c conda-forge pdbpp
Alternatively, you can just put pdb.py somewhere inside your PYTHONPATH.
Usage
Note that the module is called pdb.py so that pdb++ will automatically be used in all places that do import pdb (e.g. pytest --pdb will give you a pdb++ prompt).
The old pdb module is still available by doing e.g. import pdb; pdb.pdb.set_trace().
New interactive commands
The following are new commands that you can use from the interative (Pdb++) prompt.
-
sticky [start end] -
Toggle sticky mode. When in this mode, every time the current position changes, the screen is repainted and the whole function shown. Thus, when doing step-by-step execution you can easily follow the flow of the execution. If
startandendare given, sticky mode is enabled and only lines within that range (extremes included) will be displayed. -
longlist(ll) -
List source code for the current function. Different from the normal pdb
listcommand,longlistdisplays the whole function. The current line is marked with->. In case of post-mortem debugging, the line which actually raised the exception is marked with>>. If thehighlightconfig option is set and Pygments is installed, the source code is highlighted. -
interact - Start an interative interpreter whose global namespace contains all the names found in the current scope.
-
track EXPRESSION -
Display a graph showing which objects are the value of the expression refers to and are referred by. This command requires the
pypysource code to be importable. -
display EXPRESSION - Add an expression to the display list; expressions in this list are evaluated at each step, and printed every time its value changes. WARNING: since these expressions are evaluated multiple times, make sure not to put expressions with side-effects in the display list.
-
undisplay EXPRESSION: -
Remove
EXPRESSIONfrom the display list. -
source EXPRESSION - Show the source code for the given function/method/class.
-
edit EXPRESSION - Open the editor in the right position to edit the given function/method/class. The editor used is specified in a config option.
-
hf_unhide,hf_hide,hf_list -
Some frames might be marked as "hidden" by e.g. using the @pdb.hideframe function decorator. By default, hidden frames are not shown in the stack trace, and cannot be reached using
upanddown. You can usehf_unhideto tell pdb to ignore the hidden status (i.e., to treat hidden frames as normal ones), andhf_hideto hide them again.hf_listprints a list of hidden frames. The config optionenable_hidden_framescan be used to disable handling of hidden frames in general.
Smart command parsing
By default, pdb tries hard to interpret what you enter at the command prompt as one of its builtin commands. However, this is inconvenient if you want to just print the value of a local variable which happens to have the same name as one of the commands. E.g.:
(Pdb) list 1 2 def fn(): 3 c = 42 4 import pdb;pdb.set_trace() 5 -> return c (Pdb) c
In the example above, instead of printing 42 pdb interprets the input as the command continue, and then you lose your prompt. It's even worse than that, because it happens even if you type e.g. c.__class__.
pdb++ fixes this unfriendly (from the author's point of view, of course :-)) behavior by always prefering variable in scope, if it exists. If you really want to execute the corresponding command, you can prefix it with !!. Thus, the example above becomes:
(Pdb++) list 1 2 def fn(): 3 c = 42 4 import pdb;pdb.set_trace() 5 -> return c (Pdb++) c 42 (Pdb++) !!c
Note that the "smart" behavior takes place only when there is ambiguity, i.e. if there exists a variable with the same name as a command: in all other cases, everything works as usual.
Regarding the list command itself, using list(… is a special case that gets handled as the Python builtin:
(Pdb++) list([1, 2]) [1, 2]
Additional functions in the pdb module
The pdb module that comes with pdb++ includes all the functions and classes that are in the module from the standard library. If you find any difference, please report it as a bug.
In addition, there are some new convenience functions that are unique to pdb++.
-
pdb.xpm() -
eXtended Post Mortem: it is equivalent to
pdb.post_mortem(sys.exc_info()[2]). If used inside anexceptclause, it will start a post-mortem pdb prompt from the line that raised the exception being caught. -
pdb.disable() -
Disable
pdb.set_trace(): any subsequent call to it will be ignored. -
pdb.enable() -
Re-enable
pdb.set_trace(), in case it was disabled bypdb.disable().
-
@pdb.hideframe -
A function decorator that tells pdb++ to hide the frame corresponding to the function. Hidden frames do not show up when using interactive commands such as
up,downorwhere, unlesshf_unhideis invoked. -
@pdb.break_on_setattr(attrname, condition=always) -
class decorator: break the execution of the program every time the attribute
attrnameis set on any instance of the class.conditionis a callable that takes the target object of thesetattrand the actual value; by default, it breaks every time the attribute is set. E.g.:@break_on_setattr('bar') class Foo(object): pass f = Foo() f.bar = 42 # the program breaks hereIf can be used even after the class has already been created, e.g. if we want to break when some attribute of a particular object is set:
class Foo(object): pass a = Foo() b = Foo() def break_if_a(obj, value): return obj is a break_on_setattr('bar', condition=break_if_a)(Foo) b.bar = 10 # no break a.bar = 42 # the program breaks hereThis can be used after
pdb.set_trace()also:(Pdb++) import pdb (Pdb++) pdb.break_on_setattr('tree_id')(obj.__class__) (Pdb++) continue
Configuration and customization
To customize pdb++, you can put a file named .pdbrc.py in your home directory. The file must contain a class named Config inheriting from pdb.DefaultConfig and override the desired values.
The following is a list of the options you can customize, together with their default value:
-
prompt = '(Pdb++) ' - The prompt to show when in interactive mode.
-
highlight = True -
Highlight line numbers and the current line when showing the
longlistof a function or when in sticky mode. -
encoding = 'utf-8' - File encoding. Useful when there are international characters in your string literals or comments.
-
sticky_by_default = False - Determine whether pdb++ starts in sticky mode or not.
-
line_number_color = pdb.Color.turquoise - The color to use for line numbers. See Notes on color options.
-
filename_color = pdb.Color.yellow - The color to use for file names when printing the stack entries. See Notes on color options.
-
current_line_color = "39;49;7" -
The SGR parameters for the ANSI escape sequence to highlight the current line. The default uses the default foreground (
39) and background (49) colors, inversed (7). See Notes on color options. -
editor = None -
The command to invoke when using the
editcommand. By default, it uses$EDITORif set, elsevimorvi(if found). If only the editor command is specified, theemacsandvinotation will be used to specify the line number:COMMAND +n filename. It's otherwise possible to use another syntax by using the placeholders{filename}and{lineno}. For example with sublime text, specifyeditor = "subl {filename}:{lineno}". -
truncate_long_lines = True - Truncate lines which exceed the terminal width.
-
exec_if_unfocused = None - Shell command to execute when starting the pdb prompt and the terminal window is not focused. Useful to e.g. play a sound to alert the user that the execution of the program stopped. It requires the wmctrl module.
-
enable_hidden_frames = True -
Certain frames can be hidden by default. If enabled, the commands
hf_unhide,hf_hide, andhf_listcan be used to control display of them. -
show_hidden_frames_count = True -
If
enable_hidden_framesisTruethis controls if the number of hidden frames gets displayed. -
def setup(self, pdb): pass -
This method is called during the initialization of the
Pdbclass. Useful to do complex setup. -
show_traceback_on_error = True -
Display tracebacks for errors via
Pdb.error, that come fromPdb.default(i.e. the execution of an unrecognized pdb command), and are not a direct cause of the expression itself (e.g.NameErrorwith a command likedoesnotexist).With this option disabled only
*** exception stringgets printed, which often misses useful context. -
show_traceback_on_error_limit = None -
This option sets the limit to be used with
traceback.format_exception, whenshow_traceback_on_erroris enabled.
Options relevant for source code highlighting (using Pygments)
-
use_pygments = None -
By default Pygments is used for syntax highlighting of source code when it can be imported, e.g. when showing the
longlistof a function or when in sticky mode.
pygments_formatter_class = None
You can configure the Pygments formatter to use via the
pygments_formatter_classconfig setting as a string (dotted path). This should should be one of the following typically:"pygments.formatters.Terminal256Formatter","pygments.formatters.TerminalTrueColorFormatter", or"pygments.formatters.TerminalFormatter".The default is to auto-detect the best formatter based on the
$TERMvariable, e.g. it usesTerminal256Formatterif the$TERMvariable contains "256color" (e.g.xterm-256color), but also knows about e.g. "xterm-kitty" to support true colors (TerminalTrueColorFormatter).TerminalFormattergets used as a fallback.
pygments_formatter_kwargs = {}
A dictionary of keyword arguments to pass to the formatter's contructor.
The default arguments (updated with this setting) are:
{ "style": "default", "bg": self.config.bg, "colorscheme": self.config.colorscheme, } ``style = 'default'`` The style to use, can be a string or a Pygments Style subclass. E.g. ``"solarized-dark"``. See http://pygments.org/docs/styles/. ``bg = 'dark'`` Selects a different palette for dark/light backgrounds. Only used by ``TerminalFormatter``. ``colorscheme = None`` A dictionary mapping token types to (lightbg, darkbg) color names or ``None`` (default: ``None`` = use builtin colorscheme). Only used by ``TerminalFormatter``.
Example:
class Config(pdb.DefaultConfig):
pygments_formatter_class = "pygments.formatters.TerminalTrueColorFormatter"
pygments_formatter_kwargs = {"style": "solarized-light"}
Notes on color options
The values for color options will be used inside of the SGR escape sequence \e[%sm where \e is the ESC character and %s the given value. See SGR parameters.
The following means "reset all colors" (0), set foreground color to 18 (48;5;18), and background to 21: "0;48;5;18;38;5;21".
Constants are available via pdb.Color, e.g. pdb.Color.red ("31;01"), but in general any string can be used here.
Coding guidelines
pdb++ is developed using Test Driven Development, and we try to keep test coverage high.
As a general rule, every commit should come with its own test. If it's a new feature, it should come with one or many tests which excercise it. If it's a bug fix, the test should fail before the fix, and pass after.
The goal is to make refactoring easier in the future: if you wonder why a certain line of code does something, in principle it should be possible to comment it out and see which tests fail.
In exceptional cases, the test might be too hard or impossible to write: in that cases it is fine to do a commmit without a test, but you should explain very precisely in the commit message why it is hard to write a test and how to reproduce the buggy behaviour by hand.
It is fine NOT to write a test in the following cases:
- typos, documentation, and in general any non-coding commit
- code refactorings which do not add any feature
- commits which fix an already failing test
- commits to silence warnings
- purely cosmetic changes, such as change the color of the output
1 Jan 10, 2022
42 Dec 12, 2022
5 Feb 08, 2022
156 Jan 01, 2023
1 Nov 16, 2021
2 Jan 02, 2022
1 Nov 10, 2021
296 Dec 28, 2022
7.3k Dec 31, 2022
1.5k Jan 01, 2023
332 Dec 23, 2022
705 Dec 08, 2022
5.9k Dec 30, 2022
1.1k Dec 26, 2022
80 Nov 18, 2022
3.9k Jan 01, 2023
1.6k Jan 04, 2023
2 Dec 10, 2021
2.8k Jan 08, 2023