This documentation covers IPython versions 6.0 and higher. Beginning with version 6.0, IPython stopped supporting compatibility with Python versions lower than 3.3 including all versions of Python 2.7.
If you are looking for an IPython version compatible with Python 2.7, please use the IPython 5.x LTS release and refer to its documentation (LTS is the long term support release).
Specific config details¶
Changed in version 5.0.
From IPython 5, prompts are produced as a list of Pygments tokens, which are tuples of (token_type, text). You can customise prompts by writing a method which generates a list of tokens.
There are four kinds of prompt:
- The in prompt is shown before the first line of input
- The continuation prompt is shown before further lines of input
- The rewrite prompt is shown to highlight how special syntax has been
interpreted (default like
- The out prompt is shown before the result from evaluating the input
Custom prompts are supplied together as a class. If you want to customise only
some of the prompts, inherit from
which defines the defaults. The required interface is like this:
Prompt style definition. shell is a reference to the
continuation_prompt_tokens(self, cli=None, width=None)¶
Return the respective prompts as lists of
For continuation prompts, width is an integer representing the width of the prompt area in terminal columns.
cli, where used, is the prompt_toolkit
CommandLineInterfaceinstance. This is mainly for compatibility with the API prompt_toolkit expects.
Here is an example Prompt class that will show the current working directory in the input prompt:
from IPython.terminal.prompts import Prompts, Token import os class MyPrompt(Prompts): def in_prompt_tokens(self, cli=None): return [(Token, os.getcwd()), (Token.Prompt, ' >>>')]
To set the new prompt, assign it to the
prompts attribute of the IPython
In : ip = get_ipython() ...: ip.prompts = MyPrompt(ip) /home/bob >>> # it works
IPython/example/utils/cwd_prompt.py for an example of how to write an
extensions to customise prompts.
Inside IPython or in a startup script, you can use a custom prompts class
get_ipython().prompts to an instance of the class.
TerminalInteractiveShell.prompts_class may be set to
either the class object, or a string of its full importable name.
To include invisible terminal control sequences in a prompt, use
Token.ZeroWidthEscape as the token type. Tokens with this type are ignored
when calculating the width.
Colours in the prompt are determined by the token types and the highlighting
style; see below for more details. The tokens used in the default prompts are
Changed in version 5.0.
There are two main configuration options controlling colours.
InteractiveShell.colors sets the colour of tracebacks and object info (the
output from e.g.
zip?). It may also affect other things if the option below
is set to
'legacy'. It has four case-insensitive values:
'nocolor', 'neutral', 'linux', 'lightbg'. The default is neutral, which
should be legible on either dark or light terminal backgrounds. linux is
optimised for dark backgrounds and lightbg for light ones.
TerminalInteractiveShell.highlighting_style determines prompt colours and
syntax highlighting. It takes the name (as a string) or class (as a subclass of
pygments.style.Style) of a Pygments style, or the special value
to pick a style in accordance with
You can see the Pygments styles available on your system by running:
import pygments list(pygments.styles.get_all_styles())
TerminalInteractiveShell.highlighting_style_overrides can override
specific styles in the highlighting. It should be a dictionary mapping Pygments
token types to strings defining the style. See Pygments’ documentation for the language used
to define styles.
Colors in the pager¶
On some systems, the default pager has problems with ANSI colour codes. To configure your default pager to allow these:
- Set the environment PAGER variable to
- Set the environment LESS variable to
-r(plus any other options you always want to pass to less by default). This tells less to properly interpret control sequences, which is how color information is given to your terminal.
IPython can integrate with text editors in a number of different ways:
- Editors (such as (X)Emacs, vim and TextMate) can send code to IPython for execution.
%editmagic command can open an editor of choice to edit a code block.
The %edit command (and its alias %ed) will invoke the editor set in your
EDITOR. If this variable is not set, it will default
to vi under Linux/Unix and to notepad under Windows. You may want to set this
variable properly and to a lightweight editor which doesn’t take too long to
start (that is, something other than a new instance of Emacs). This way you
can edit multi-line code quickly and with the power of a real editor right
You can also control the editor by setting
If you are a dedicated Emacs user, and want to use Emacs when IPython’s
%edit magic command is called you should set up the Emacs server so that
new requests are handled by the original process. This means that almost no
time is spent in handling the request (assuming an Emacs process is already
running). For this to work, you need to set your EDITOR environment variable
to ‘emacsclient’. The code below, supplied by Francois Pinard, can then be
used in your
.emacs file to enable the server:
(defvar server-buffer-clients) (when (and (fboundp 'server-start) (string-equal (getenv "TERM") 'xterm)) (server-start) (defun fp-kill-server-with-buffer-routine () (and server-buffer-clients (server-done))) (add-hook 'kill-buffer-hook 'fp-kill-server-with-buffer-routine))
Thanks to the work of Alexander Schmolck and Prabhu Ramachandran, currently (X)Emacs and IPython get along very well in other ways.
With (X)EMacs >= 24, You can enable IPython in python-mode with:
(require 'python) (setq python-shell-interpreter "ipython")
Changed in version 5.0.
You can customise keyboard shortcuts for terminal IPython. Put code like this in a startup file:
from IPython import get_ipython from prompt_toolkit.enums import DEFAULT_BUFFER from prompt_toolkit.keys import Keys from prompt_toolkit.filters import HasFocus, HasSelection, ViInsertMode, EmacsInsertMode ip = get_ipython() insert_mode = ViInsertMode() | EmacsInsertMode() def insert_unexpected(event): buf = event.current_buffer buf.insert_text('The Spanish Inquisition') # Register the shortcut if IPython is using prompt_toolkit if getattr(ip, 'pt_cli'): registry = ip.pt_cli.application.key_bindings_registry registry.add_binding(Keys.ControlN, filter=(HasFocus(DEFAULT_BUFFER) & ~HasSelection() & insert_mode))(insert_unexpected)
For more information on filters and what you can do with the
see the prompt_toolkit docs.
Enter to execute¶
In the Terminal IPython shell – which by default uses the
interface, the semantic meaning of pressing the Enter key can be
ambiguous. In some case Enter should execute code, and in others it
should add a new line. IPython uses heuristics to decide whether to execute or
insert a new line at cursor position. For example, if we detect that the current
code is not valid Python, then the user is likely editing code and the right
behavior is to likely to insert a new line. If the current code is a simple
ord('*'), then the right behavior is likely to execute. Though
the exact desired semantics often varies from users to users.
As the exact behavior of Enter is ambiguous, it has been special cased
to allow users to completely configure the behavior they like. Hence you can
have enter always execute code. If you prefer fancier behavior, you need to get
your hands dirty and read the
prompt_toolkit and IPython documentation
though. See PR #10500, set the
c.TerminalInteractiveShell.handle_return option and get inspiration from the
following example that only auto-executes the input if it begins with a bang or
a modulo character (
%). To use the following code, add it to your
def custom_return(shell): """This function is required by the API. It takes a reference to the shell, which is the same thing `get_ipython()` evaluates to. This function must return a function that handles each keypress event. That function, named `handle` here, references `shell` by closure.""" def handle(event): """This function is called each time `Enter` is pressed, and takes a reference to a Prompt Toolkit event object. If the current input starts with a bang or modulo, then the input is executed, otherwise a newline is entered, followed by any spaces needed to auto-indent.""" # set up a few handy references to nested items... buffer = event.current_buffer document = buffer.document text = document.text if text.startswith('!') or text.startswith('%'): # execute the input... buffer.accept_action.validate_and_handle(event.cli, buffer) else: # insert a newline with auto-indentation... if document.line_count > 1: text = text[:document.cursor_position] indent = shell.input_splitter.check_complete(text + '\n') or 0 buffer.insert_text('\n' + ' ' * indent) # if you just wanted a plain newline without any indentation, you # could use `buffer.insert_text('\n')` instead of the lines above return handle c.TerminalInteractiveShell.handle_return = custom_return