Skip to content

Latest commit

 

History

History
477 lines (327 loc) · 17.6 KB

shell-integration.rst

File metadata and controls

477 lines (327 loc) · 17.6 KB

Shell integration

kitty has the ability to integrate closely within common shells, such as zsh, fish and bash to enable features such as jumping to previous prompts in the scrollback, viewing the output of the last command in :program:`less`, using the mouse to move the cursor while editing prompts, etc.

.. versionadded:: 0.24.0

Features

  • Open the output of the last command in a pager such as :program:`less` (:sc:`show_last_command_output`)
  • Jump to the previous/next prompt in the scrollback (:sc:`scroll_to_previous_prompt` / :sc:`scroll_to_next_prompt`)
  • Click with the mouse anywhere in the current command to move the cursor there
  • Hold Ctrl+Shift and right-click on any command output in the scrollback to view it in a pager
  • The current working directory or the command being executed are automatically displayed in the kitty window titlebar/tab title
  • The text cursor is changed to a bar when editing commands at the shell prompt
  • :ref:`clone_shell` with all environment variables and the working directory copied
  • :ref:`Edit files in new kitty windows <edit_file>` even over SSH
  • Glitch free window resizing even with complex prompts. Achieved by erasing the prompt on resize and allowing the shell to redraw it cleanly.
  • Sophisticated completion for the :program:`kitty` command in the shell.
  • When confirming a quit command if a window is sitting at a shell prompt, it is not counted (for details, see :opt:`confirm_os_window_close`)

Configuration

Shell integration is controlled by the :opt:`shell_integration` option. By default, all integration features are enabled. Individual features can be turned off or it can be disabled entirely as well. The :opt:`shell_integration` option takes a space separated list of keywords:

disabled
Turn off all shell integration. The shell's launch environment is not modified and :envvar:`KITTY_SHELL_INTEGRATION` is not set. Useful for :ref:`manual integration <manual_shell_integration>`.
no-rc
Do not modify the shell's launch environment to enable integration. Useful if you prefer to load the kitty shell integration code yourself, either as part of :ref:`manual integration <manual_shell_integration>` or because you have some other software that sets up shell integration. This will still set the :envvar:`KITTY_SHELL_INTEGRATION` environment variable when kitty runs the shell.
no-cursor
Turn off changing of the text cursor to a bar when editing shell command line.
no-title
Turn off setting the kitty window/tab title based on shell state. Note that for the fish shell kitty relies on fish's native title setting functionality instead.
no-cwd
Turn off reporting the current working directory. This is used to allow new_window_with_cwd and similar to open windows logged into remote machines using the :doc:`ssh kitten <kittens/ssh>` automatically with the same working directory as the current window. Note that for the fish shell this will not disable its built-in current working directory reporting.
no-prompt-mark
Turn off marking of prompts. This disables jumping to prompt, browsing output of last command and click to move cursor functionality. Note that for the fish shell this does not take effect, since fish always marks prompts.
no-complete
Turn off completion for the kitty command. Note that for the fish shell this does not take effect, since fish already comes with a kitty completion script.
no-sudo
Do not alias :program:`sudo` to ensure the kitty terminfo files are available in the sudo environment. This is needed if you have sudo configured to disable setting of environment variables on the command line. By default, if sudo is configured to allow all commands for the current user, setting of environment variables at the command line is also allowed. Only if commands are restricted is this needed.

More ways to browse command output

You can add further key and mouse bindings to browse the output of commands easily. For example to select the output of a command by right clicking the mouse on the output, define the following in :file:`kitty.conf`:

mouse_map right press ungrabbed mouse_select_command_output

Now, when you right click on the output, the entire output is selected, ready to be copied.

The feature to jump to previous prompts ( :sc:`scroll_to_previous_prompt` and :sc:`scroll_to_next_prompt`) and mouse actions (mouse_select_command_output and mouse_show_command_output) can be integrated with browsing command output as well. For example, define the following mapping in :file:`kitty.conf`:

map f1 show_last_visited_command_output

Now, pressing F1 will cause the output of the last jumped to command or the last mouse clicked command output to be opened in a pager for easy browsing.

In addition, You can define shortcut to get the first command output on screen. For example, define the following in :file:`kitty.conf`:

map f1 show_first_command_output_on_screen

Now, pressing F1 will cause the output of the first command output on screen to be opened in a pager.

You can also add shortcut to scroll to the last jumped position. For example, define the following in :file:`kitty.conf`:

map f1 scroll_to_prompt 0

How it works

At startup, kitty detects if the shell you have configured (either system wide or the :opt:`shell` option in :file:`kitty.conf`) is a supported shell. If so, kitty injects some shell specific code into the shell, to enable shell integration. How it does so varies for different shells.

.. tab:: zsh

   For zsh, kitty sets the :envvar:`ZDOTDIR` environment variable to make zsh
   load kitty's :file:`.zshenv` which restores the original value of
   :envvar:`ZDOTDIR` and sources the original :file:`.zshenv`. It then loads
   the shell integration code. The remainder of zsh's startup process proceeds
   as normal.


.. tab:: fish

    For fish, to make it automatically load the integration code provided by
    kitty, the integration script directory path is prepended to the
    :envvar:`XDG_DATA_DIRS` environment variable. This is only applied to the
    fish process and will be cleaned up by the integration script after startup.
    No files are added or modified.


.. tab:: bash

    For bash, kitty starts bash in POSIX mode, using the environment variable
    :envvar:`ENV` to load the shell integration script. This prevents bash from
    loading any startup files itself. The loading of the startup files is done
    by the integration script, after disabling POSIX mode. From the perspective
    of those scripts there should be no difference to running vanilla bash.

Then, when launching the shell, kitty sets the environment variable :envvar:`KITTY_SHELL_INTEGRATION` to the value of the :opt:`shell_integration` option. The shell integration code reads the environment variable, turns on the specified integration functionality and then unsets the variable so as to not pollute the system.

The actual shell integration code uses hooks provided by each shell to send special escape codes to kitty, to perform the various tasks. You can see the code used for each shell below:

Click to toggle shell integration code
.. tab:: zsh

    .. literalinclude:: ../shell-integration/zsh/kitty-integration
        :language: zsh



.. tab:: fish

    .. literalinclude:: ../shell-integration/fish/vendor_conf.d/kitty-shell-integration.fish
        :language: fish
        :force:


.. tab:: bash

    .. literalinclude:: ../shell-integration/bash/kitty.bash
        :language: bash

Shell integration over SSH

The easiest way to have shell integration work when SSHing into remote systems is to use the :doc:`ssh kitten <kittens/ssh>`. Simply run:

kitten ssh hostname

And, by magic, you will be logged into the remote system with fully functional shell integration. Alternately, you can :ref:`setup shell integration manually <manual_shell_integration>`, by copying the kitty shell integration scripts to the remote server and editing the shell rc files there, as described below.

Shell integration in a container

Install the kitten standalone binary in the container somewhere in the PATH, then you can log into the container with:

docker exec -ti container-id kitten run-shell --shell=/path/to/your/shell/in/the/container

The kitten will even take care of making the kitty terminfo database available in the container automatically.

Clone the current shell into a new window

You can clone the current shell into a new kitty window by simply running the :command:`clone-in-kitty` command, for example:

clone-in-kitty
clone-in-kitty --type=tab
clone-in-kitty --title "I am a clone"

This will open a new window running a new shell instance but with all environment variables and the current working directory copied. This even works over SSH when using :doc:`kittens/ssh`.

The :command:`clone-in-kitty` command takes almost all the same arguments as the :doc:`launch <launch>` command, so you can open a new tab instead or a new OS window, etc. Arguments of launch that can cause code execution or that don't make sense when cloning are ignored. Most prominently, the following options are ignored: :option:`--allow-remote-control <launch --allow-remote-control>`, :option:`--copy-cmdline <launch --copy-cmdline>`, :option:`--copy-env <launch --copy-env>`, :option:`--stdin-source <launch --stdin-source>`, :option:`--marker <launch --marker>` and :option:`--watcher <launch --watcher>`.

:command:`clone-in-kitty` can be configured to source arbitrary code in the cloned window using environment variables. It will automatically clone virtual environments created by the 🔗`Python venv module <https://docs.python.org/3/library/venv.html>` or 🔗`Conda <https://conda.io/>`. In addition, setting the env var :envvar:`KITTY_CLONE_SOURCE_CODE` to some shell code will cause that code to be run in the cloned window with eval. Similarly, setting :envvar:`KITTY_CLONE_SOURCE_PATH` to the path of a file will cause that file to be sourced in the cloned window. This can be controlled by :opt:`clone_source_strategies`.

:command:`clone-in-kitty` works by asking the shell to serialize its internal state (mainly CWD and env vars) and this state is transmitted to kitty and restored by the shell integration scripts in the cloned window.

Edit files in new kitty windows even over SSH

edit-in-kitty myfile.txt
edit-in-kitty --type tab --title "Editing My File" myfile.txt
# open myfile.txt at line 75 (works with vim, neovim, emacs, nano, micro)
edit-in-kitty +75 myfile.txt

The :command:`edit-in-kitty` command allows you to seamlessly edit files in your default :opt:`editor` in new kitty windows. This works even over SSH (if you use the :doc:`ssh kitten <kittens/ssh>`), allowing you to easily edit remote files in your local editor with all its bells and whistles.

The :command:`edit-in-kitty` command takes almost all the same arguments as the :doc:`launch <launch>` command, so you can open a new tab instead or a new OS window, etc. Not all arguments are supported, see the discussion in the :ref:`clone_shell` section above.

In order to avoid remote code execution, kitty will only execute the configured editor and pass the file path to edit to it.

Note

To edit files using sudo the best method is to set the SUDO_EDITOR environment variable to kitten edit-in-kitty and then edit the file using the sudoedit or sudo -e commands.

Using shell integration in sub-shells, containers, etc.

.. versionadded:: 0.29.0

To start a sub-shell with shell integration automatically setup, simply run:

kitten run-shell

This will start a sub-shell using the same binary as the currently running shell, with shell-integration enabled. To start a particular shell use:

kitten run-shell --shell=/bin/bash

To run a command before starting the shell use:

kitten run-shell ls .

This will run ls . before starting the shell.

This will even work on remote systems where kitty itself is not installed, provided you use the :doc:`SSH kitten <kittens/ssh>` to connect to the system. Use kitten run-shell --help to learn more.

Manual shell integration

The automatic shell integration is designed to be minimally intrusive, as such it won't work for sub-shells, terminal multiplexers, containers, etc. For such systems, you should either use the :ref:`run-shell <run_shell>` command described above or setup manual shell integration by adding some code to your shells startup files to load the shell integration script.

First, in :file:`kitty.conf` set:

shell_integration disabled

Then in your shell's rc file, add the lines:

.. tab:: zsh

    .. code-block:: sh

        if test -n "$KITTY_INSTALLATION_DIR"; then
            export KITTY_SHELL_INTEGRATION="enabled"
            autoload -Uz -- "$KITTY_INSTALLATION_DIR"/shell-integration/zsh/kitty-integration
            kitty-integration
            unfunction kitty-integration
        fi


.. tab:: fish

    .. code-block:: fish

        if set -q KITTY_INSTALLATION_DIR
            set --global KITTY_SHELL_INTEGRATION enabled
            source "$KITTY_INSTALLATION_DIR/shell-integration/fish/vendor_conf.d/kitty-shell-integration.fish"
            set --prepend fish_complete_path "$KITTY_INSTALLATION_DIR/shell-integration/fish/vendor_completions.d"
        end



.. tab:: bash

    .. code-block:: sh

        if test -n "$KITTY_INSTALLATION_DIR"; then
            export KITTY_SHELL_INTEGRATION="enabled"
            source "$KITTY_INSTALLATION_DIR/shell-integration/bash/kitty.bash"
        fi

The value of :envvar:`KITTY_SHELL_INTEGRATION` is the same as that for :opt:`shell_integration`, except if you want to disable shell integration completely, in which case simply do not set the :envvar:`KITTY_SHELL_INTEGRATION` variable at all.

In a container, you will need to install the kitty shell integration scripts and make sure the :envvar:`KITTY_INSTALLATION_DIR` environment variable is set to point to the location of the scripts.

Integration with other shells

There exist third-party integrations to use these features for various other shells:

  • Jupyter console and IPython via a patch (:iss:`4475`)
  • xonsh
  • Nushell: Set $env.config.shell_integration = true in your config.nu to enable it.

Notes for shell developers

The protocol used for marking the prompt is very simple. You should consider adding it to your shell as a builtin. Many modern terminals make use of it, for example: kitty, iTerm2, WezTerm, DomTerm

Just before starting to draw the PS1 prompt send the escape code:

<OSC>133;A<ST>

Just before starting to draw the PS2 prompt send the escape code:

<OSC>133;A;k=s<ST>

Just before running a command/program, send the escape code:

<OSC>133;C<ST>

Optionally, when a command is finished its "exit status" can be reported as:

<OSC>133;D;exit status as base 10 integer<ST>

Here <OSC> is the bytes 0x1b 0x5d and <ST> is the bytes 0x1b 0x5c. This is exactly what is needed for shell integration in kitty. For the full protocol, that also marks the command region, see the iTerm2 docs.

kitty additionally supports several extra fields for the <OSC>133;A command to control its behavior, separated by semi-colons. They are:

redraw=0 - this tells kitty that the shell will not redraw the prompt on
resize so it should not erase it

special_key=1 - this tells kitty to use a special key instead of arrow keys
to move the cursor on mouse click. Useful if arrow keys have side-effects
like triggering auto complete. The shell integration script then binds the
special key, as needed.

k=s - this tells kitty that the secondary (PS2) prompt is starting at the
current line.

click_events=1 - this tells kitty that the shell is capable of handling
mouse click events. kitty will thus send a click event to the shell when
the user clicks somewhere in the prompt. The shell can then move the cursor
to that position or perform some other appropriate action. Without this,
kitty will instead generate a number of fake key events to move the cursor
to the clicked location, which is not fully robust.

kitty also optionally supports sending the cmdline going to be executed with <OSC>133;C as:

<OSC>133;C;cmdline=cmdline encoded by %q<ST>
or
<OSC>133;C;cmdline_url=cmdline as UTF-8 URL %-escaped text<ST>

Here, encoded by %q means the encoding produced by the %q format to printf in bash and similar shells. Which is basically shell escaping with the addition of using ANSI C quoting for control characters ($'' quoting).