path: root/docs/transient.org

#+TITLE: Transient User and Developer Manual
:PREAMBLE:
#+AUTHOR: Jonas Bernoulli
#+EMAIL: jonas@bernoul.li
#+DATE: 2018-2021
#+LANGUAGE: en

#+TEXINFO_DIR_CATEGORY: Emacs
#+TEXINFO_DIR_TITLE: Transient: (transient).
#+TEXINFO_DIR_DESC: Transient Commands
#+SUBTITLE: for version 0.3.6

#+TEXINFO_DEFFN: t
#+OPTIONS: H:4 num:4 toc:2
#+PROPERTY: header-args :eval never
#+BIND: ox-texinfo+-before-export-hook ox-texinfo+-update-copyright-years
#+BIND: ox-texinfo+-before-export-hook ox-texinfo+-update-version-strings

Taking inspiration from prefix keys and prefix arguments, Transient
implements a similar abstraction involving a prefix command, infix
arguments and suffix commands.  We could call this abstraction a
"transient command", but because it always involves at least two
commands (a prefix and a suffix) we prefer to call it just a
"transient".

When the user calls a transient prefix command, then a transient
(temporary) keymap is activated, which binds the transient's infix
and suffix commands, and functions that control the transient state
are added to ~pre-command-hook~ and ~post-command-hook~.  The available
suffix and infix commands and their state are shown in a popup buffer
until the transient is exited by invoking a suffix command.

Calling an infix command causes its value to be changed, possibly by
reading a new value in the minibuffer.

Calling a suffix command usually causes the transient to be exited
but suffix commands can also be configured to not exit the transient.

#+TEXINFO: @noindent
This manual is for Transient version 0.3.6.

#+BEGIN_QUOTE
Copyright (C) 2018-2021 Jonas Bernoulli <jonas@bernoul.li>

You can redistribute this document and/or modify it under the terms
of the GNU General Public License as published by the Free Software
Foundation, either version 3 of the License, or (at your option) any
later version.

This document is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
General Public License for more details.
#+END_QUOTE
:END:
* Introduction

Taking inspiration from prefix keys and prefix arguments, Transient
implements a similar abstraction involving a prefix command, infix
arguments and suffix commands.  We could call this abstraction a
"transient command", but because it always involves at least two
commands (a prefix and a suffix) we prefer to call it just a
"transient".

#+BEGIN_QUOTE
Transient keymaps are a feature provided by Emacs.  Transients as
implemented by this package involve the use of transient keymaps.

Emacs provides a feature that it calls "prefix commands".  When we
talk about "prefix commands" in this manual, then we mean our own kind
of "prefix commands", unless specified otherwise.  To avoid ambiguity
we sometimes use the terms "transient prefix command" for our kind and
"regular prefix command" for Emacs' kind.
#+END_QUOTE

When the user calls a transient prefix command, then a transient
(temporary) keymap is activated, which binds the transient's infix and
suffix commands, and functions that control the transient state are
added to ~pre-command-hook~ and ~post-command-hook~.  The available suffix
and infix commands and their state are shown in a popup buffer until
the transient state is exited by invoking a suffix command.

Calling an infix command causes its value to be changed.  How that is
done depends on the type of the infix command.  The simplest case is
an infix command that represents a command-line argument that does not
take a value.  Invoking such an infix command causes the switch to be
toggled on or off.  More complex infix commands may read a value from
the user, using the minibuffer.

Calling a suffix command usually causes the transient to be exited;
the transient keymaps and hook functions are removed, the popup buffer
no longer shows information about the (no longer bound) suffix
commands, the values of some public global variables are set, while
some internal global variables are unset, and finally the command is
actually called.  Suffix commands can also be configured to not exit
the transient.

A suffix command can, but does not have to, use the infix arguments in
much the same way any command can choose to use or ignore the prefix
arguments.  For a suffix command that was invoked from a transient the
variable ~transient-current-suffixes~ and the function ~transient-args~
serve about the same purpose as the variables ~prefix-arg~ and
~current-prefix-arg~ do for any command that was called after the prefix
arguments have been set using a command such as ~universal-argument~.

The information shown in the popup buffer while a transient is active
looks a bit like this:

#+BEGIN_EXAMPLE
,-----------------------------------------
|Arguments
| -f Force (--force)
| -a Annotate (--annotate)
|
|Create
| t tag
| r release
`-----------------------------------------
#+END_EXAMPLE

#+BEGIN_QUOTE
This is a simplified version of ~magit-tag~.  Info manuals do not
support images or colored text, so the above "screenshot" lacks some
information; in practice you would be able to tell whether the
arguments ~--force~ and ~--annotate~ are enabled or not based on their
color.
#+END_QUOTE

Transient can be used to implement simple "command dispatchers".  The
main benefit then is that the user can see all the available commands
in a popup buffer.  That is useful by itself because it frees the user
from having to remember all the keys that are valid after a certain
prefix key or command.  Magit's ~magit-dispatch~ (on ~C-x M-g~) command is
an example of using Transient to merely implement a command
dispatcher.

In addition to that, Transient also allows users to interactively pass
arguments to commands.  These arguments can be much more complex than
what is reasonable when using prefix arguments.  There is a limit to
how many aspects of a command can be controlled using prefix
arguments.  Furthermore what a certain prefix argument means for
different commands can be completely different, and users have to read
documentation to learn and then commit to memory what a certain prefix
argument means to a certain command.

Transient suffix commands on the other hand can accept dozens of
different arguments without the user having to remember anything.
When using Transient, then one can call a command with arguments that
are just as complex as when calling the same function non-interactively
using code.

Invoking a transient command with arguments is similar to invoking a
command in a shell with command-line completion and history enabled.
One benefit of the Transient interface is that it remembers history
not only on a global level ("this command was invoked using these
arguments and previously it was invoked using those other arguments"),
but also remembers the values of individual arguments independently.
See [[*Using History]].

After a transient prefix command is invoked ~C-h <key>~ can be used to
show the documentation for the infix or suffix command that ~<key>~ is
bound to (see [[*Getting Help for Suffix Commands]]) and infixes and
suffixes can be removed from the transient using ~C-x l <key>~.  Infixes
and suffixes that are disabled by default can be enabled the same way.
See [[*Enabling and Disabling Suffixes]].

Transient ships with support for a few different types of specialized
infix commands.  A command that sets a command line option for example
has different needs than a command that merely toggles a boolean flag.
Additionally Transient provides abstractions for defining new types,
which the author of Transient did not anticipate (or didn't get around
to implementing yet).

* Usage
** Invoking Transients

A transient prefix command is invoked like any other command by
pressing the key that is bound to that command.  The main difference
to other commands is that a transient prefix command activates a
transient keymap, which temporarily binds the transient's infix and
suffix commands.  Bindings from other keymaps may, or may not, be
disabled while the transient state is in effect.

There are two kinds of commands that are available after invoking a
transient prefix command; infix and suffix commands.  Infix commands
set some value (which is then shown in a popup buffer), without
leaving the transient.  Suffix commands on the other hand usually quit
the transient and they may use the values set by the infix commands,
i.e. the infix *arguments*.

Instead of setting arguments to be used by a suffix command, infix
commands may also set some value by side-effect, e.g. by setting the
value of some variable.

** Aborting and Resuming Transients

To quit the transient without invoking a suffix command press ~C-g~.

Key bindings in transient keymaps may be longer than a single event.
After pressing a valid prefix key, all commands whose bindings do not
begin with that prefix key are temporarily unavailable and grayed out.
To abort the prefix key press ~C-g~ (which in this case only quits the
prefix key, but not the complete transient).

A transient prefix command can be bound as a suffix of another
transient.  Invoking such a suffix replaces the current transient
state with a new transient state, i.e. the available bindings change
and the information displayed in the popup buffer is updated
accordingly.  Pressing ~C-g~ while a nested transient is active only
quits the innermost transient, causing a return to the previous
transient.

~C-q~ or ~C-z~ on the other hand always exits all transients.  If you use
the latter, then you can later resume the stack of transients using
~M-x transient-resume~.

- Key: C-g, transient-quit-seq
- Key: C-g, transient-quit-one

  This key quits the currently active incomplete key sequence, if any,
  or else the current transient.  When quitting the current transient,
  then it returns to the previous transient, if any.

Transient's predecessor bound ~q~ instead of ~C-g~ to the quit command.
To learn how to get that binding back see ~transient-bind-q-to-quit~'s
doc string.

- Key: C-q, transient-quit-all

  This command quits the currently active incomplete key sequence, if
  any, and all transients, including the active transient and all
  suspended transients, if any.

- Key: C-z, transient-suspend

  Like ~transient-quit-all~, this command quits an incomplete key
  sequence, if any, and all transients.  Additionally it saves the
  stack of transients so that it can easily be resumed (which is
  particularly useful if you quickly need to do "something else" and
  the stack is deeper than a single transient and/or you have already
  changed the values of some infix arguments).

  Note that only a single stack of transients can be saved at a time.
  If another stack is already saved, then saving a new stack discards
  the previous stack.

- Key: M-x transient-resume, transient-resume

  This command resumes the previously suspended stack of transients,
  if any.

** Common Suffix Commands

A few shared suffix commands are available in all transients.  These
suffix commands are not shown in the popup buffer by default.

This includes the aborting commands mentioned in the previous node as
well as some other commands that are all bound to ~C-x <key>~.  After
~C-x~ is pressed, a section featuring all these common commands is
temporarily shown in the popup buffer.  After invoking one of them,
the section disappears again.  Note however that one of these commands
is described as "Show common permanently"; invoke that if you want the
common commands to always be shown for all transients.

- Key: C-x t, transient-toggle-common

  This command toggles whether the generic commands that are common to
  all transients are always displayed or only after typing the
  incomplete prefix key sequence ~C-x~.  This only affects the current
  Emacs session.

- User Option: transient-show-common-commands

  This option controls whether shared suffix commands are shown
  alongside the transient-specific infix and suffix commands.  By
  default the shared commands are not shown to avoid overwhelming
  the user with to many options.

  While a transient is active, pressing ~C-x~ always shows the common
  command.  The value of this option can be changed for the current
  Emacs session by typing ~C-x t~ while a transient is active.

The other common commands are described in either the previous node or
in one of the following nodes.

Some of Transient's key bindings differ from the respective bindings
of Magit-Popup; see [[*FAQ]] for more information.

** Saving Values

After setting the infix arguments in a transient, the user can save
those arguments for future invocations.

Most transients will start out with the saved arguments when they are
invoked.  There are a few exceptions though.  Some transients are
designed so that the value that they use is stored externally as the
buffer-local value of some variable.  Invoking such a transient again
uses the buffer-local value. [fn:1]

If the user does not save the value and just exits using a regular
suffix command, then the value is merely saved to the transient's
history.  That value won't be used when the transient is next invoked
but it is easily accessible (see [[*Using History]]).

- Key: C-x s, transient-set

  This command saves the value of the active transient for this Emacs
  session.

- Key: C-x C-s, transient-save

  Save the value of the active transient persistently across Emacs
  sessions.

- User Option: transient-values-file

  This file is used to persist the values of transients between Emacs
  sessions.

[fn:1] ~magit-diff~ and ~magit-log~ are two prominent examples, and their
handling of buffer-local values is actually a bit more complicated
than outlined above and even customizable.

** Using History

Every time the user invokes a suffix command the transient's current
value is saved to its history.  These values can be cycled through the
same way one can cycle through the history of commands that read
user-input in the minibuffer.

- Key: C-M-p, transient-history-prev
- Key: C-x p, transient-history-prev

  This command switches to the previous value used for the active
  transient.

- Key: C-M-n, transient-history-next
- Key: C-x n, transient-history-next

  This command switches to the next value used for the active
  transient.

In addition to the transient-wide history, Transient of course
supports per-infix history.  When an infix reads user-input using the
minibuffer, then the user can use the regular minibuffer history
commands to cycle through previously used values.  Usually the same
keys as those mentioned above are bound to those commands.

Authors of transients should arrange for different infix commands that
read the same kind of value to also use the same history key (see
[[*Suffix Slots]]).

Both kinds of history are saved to a file when Emacs is exited.

- User Option: transient-history-file

  This file is used to persist the history of transients and their
  infixes between Emacs sessions.

- User Option: transient-history-limit

  This option controls how many history elements are kept at the time
  the history is saved in ~transient-history-file~.

** Getting Help for Suffix Commands

Transients can have many suffixes and infixes that the user might not
be familiar with.  To make it trivial to get help for these, Transient
provides access to the documentation directly from the active
transient.

- Key: C-h, transient-help

  This command enters help mode.  When help mode is active, then
  typing ~<key>~ shows information about the suffix command that ~<key>~
  normally is bound to (instead of invoking it).  Pressing ~C-h~ a
  second time shows information about the /prefix/ command.

  After typing ~<key>~ the stack of transient states is suspended and
  information about the suffix command is shown instead.  Typing ~q~ in
  the help buffer buries that buffer and resumes the transient state.

What sort of documentation is shown depends on how the transient was
defined.  For infix commands that represent command-line arguments
this ideally shows the appropriate manpage.  ~transient-help~ then tries
to jump to the correct location within that.  Info manuals are also
supported.  The fallback is to show the command's doc string, for
non-infix suffixes this is usually appropriate.

** Enabling and Disabling Suffixes

The user base of a package that uses transients can be very diverse.
This is certainly the case for Magit; some users have been using it and
Git for a decade, while others are just getting started now.

For that reason a mechanism is needed that authors can use to classify a
transient's infixes and suffixes along the essentials...everything
spectrum.  We use the term "levels" to describe that mechanism.

Each suffix command is placed on a level and each transient has a
level (called transient-level), which controls which suffix commands
are available.  Integers between 1 and 7 (inclusive) are valid levels.
For suffixes, 0 is also valid; it means that the suffix is not
displayed at any level.

The levels of individual transients and/or their individual suffixes
can be changed interactively, by invoking the transient and then
pressing ~C-x l~ to enter the "edit" mode, see below.

The default level for both transients and their suffixes is 4.  The
~transient-default-level~ option only controls the default for
transients.  The default suffix level is always 4.  The authors of
transients should place certain suffixes on a higher level, if they
expect that it won't be of use to most users, and they should place
very important suffixes on a lower level, so that they remain
available even if the user lowers the transient level.

- User Option: transient-default-level

  This option controls which suffix levels are made available by
  default.  It sets the transient-level for transients for which the
  user has not set that individually.

- User Option: transient-levels-file

  This file is used to persist the levels of transients and their
  suffixes between Emacs sessions.

- Key: C-x l, transient-set-level

  This command enters edit mode.  When edit mode is active, then all
  infixes and suffixes that are currently usable are displayed along
  with their levels.  The colors of the levels indicate whether they
  are enabled or not.  The level of the transient is also displayed
  along with some usage information.

  In edit mode, pressing the key that would usually invoke a certain
  suffix instead prompts the user for the level that suffix should be
  placed on.

  Help mode is available in edit mode.

  To change the transient level press ~C-x l~ again.

  To exit edit mode press ~C-g~.

  Note that edit mode does not display any suffixes that are not
  currently usable.  ~magit-rebase~ for example shows different suffixes
  depending on whether a rebase is already in progress or not.  The
  predicates also apply in edit mode.

  Therefore, to control which suffixes are available given a certain
  state, you have to make sure that that state is currently active.

** Other Commands

When invoking a transient in a small frame, the transient window may
not show the complete buffer, making it necessary to scroll, using the
following commands.  These commands are never shown in the transient
window, and the key bindings are the same as for ~scroll-up-command~ and
~scroll-down-command~ in other buffers.

- Command: transient-scroll-up arg

  This command scrolls text of transient popup window upward ARG
  lines.  If ARG is ~nil~, then it scrolls near full screen.  This
  is a wrapper around ~scroll-up-command~ (which see).

- Command: transient-scroll-down arg

  This command scrolls text of transient popup window down ARG
  lines.  If ARG is ~nil~, then it scrolls near full screen.  This
  is a wrapper around ~scroll-down-command~ (which see).

** Other Options

- User Option: transient-show-popup

  This option controls whether the current transient's infix and
  suffix commands are shown in the popup buffer.

  - If ~t~ (the default) then the popup buffer is shown as soon as a
    transient prefix command is invoked.

  - If ~nil~, then the popup buffer is not shown unless the user
    explicitly requests it, by pressing an incomplete prefix key
    sequence.

  - If a number, then the a brief one-line summary is shown instead of
    the popup buffer.  If zero or negative, then not even that summary
    is shown; only the pressed key itself is shown.

    The popup is shown when the user explicitly requests it by
    pressing an incomplete prefix key sequence.  Unless this is zero,
    then the popup is shown after that many seconds of inactivity
    (using the absolute value).

- User Option: transient-enable-popup-navigation

  This option controls whether navigation commands are enabled in the
  transient popup buffer.

  While a transient is active the transient popup buffer is not the
  current buffer, making it necessary to use dedicated commands to act
  on that buffer itself.  This is disabled by default.  If this option
  is non-nil, then the following features are available:

  - ~<up>~ moves the cursor to the previous suffix.
    ~<down>~ moves the cursor to the next suffix.
    ~RET~ invokes the suffix the cursor is on.
  - ~<mouse-1>~ invokes the clicked on suffix.
  - ~C-s~ and ~C-r~ start isearch in the popup buffer.

- User Option: transient-display-buffer-action

  This option specifies the action used to display the transient popup
  buffer.  The transient popup buffer is displayed in a window using
  ~(display-buffer BUFFER transient-display-buffer-action)~.

  The value of this option has the form ~(FUNCTION . ALIST)~, where
  FUNCTION is a function or a list of functions.  Each such function
  should accept two arguments: a buffer to display and an alist of the
  same form as ALIST.  See [[info:elisp#Choosing Window]].

  The default is ~(display-buffer-in-side-window (side . bottom))~.
  This displays the window at the bottom of the selected frame.
  Another useful value is ~(display-buffer-below-selected)~.  This is
  what ~magit-popup~ used by default.  For more alternatives see
  [[info:elisp#Display Action Functions]].

  It may be possible to display the window in another frame, but
  whether that works in practice depends on the window-manager.
  If the window manager selects the new window (Emacs frame),
  then it doesn't work.

  If you change the value of this option, then you might also want
  to change the value of ~transient-mode-line-format~.

- User Option: transient-mode-line-format

  This option controls whether the transient popup buffer has a
  mode-line, separator line, or neither.

  If ~nil~, then the buffer has no mode-line.  If the buffer is not
  displayed right above the echo area, then this probably is not a
  good value.

  If ~line~ (the default), then the buffer also has no mode-line, but a
  thin line is drawn instead, using the background color of the face
  ~transient-separator~.  Termcap frames cannot display thin lines and
  therefore fallback to treating ~line~ like ~nil~.

  Otherwise this can be any mode-line format.  See [[info:elisp#Mode
  Line Format]] for details.

- User Option: transient-read-with-initial-input

  This option controls whether the last history element is used as the
  initial minibuffer input when reading the value of an infix argument
  from the user.  If ~nil~, then there is no initial input and the first
  element has to be accessed the same way as the older elements.

- User Option: transient-highlight-mismatched-keys

  This option controls whether key bindings of infix commands that do
  not match the respective command-line argument should be highlighted.
  For other infix commands this option has no effect.

  When this option is non-nil, then the key binding for an infix argument
  is highlighted when only a long argument (e.g. ~--verbose~) is
  specified but no shorthand (e.g ~-v~).  In the rare case that a
  shorthand is specified but the key binding does not match, then it
  is highlighted differently.

  Highlighting mismatched key bindings is useful when learning the
  arguments of the underlying command-line tool; you wouldn't want to
  learn any short-hands that do not actually exist.

  The highlighting is done using one of the faces
  ~transient-mismatched-key~ and ~transient-nonstandard-key~.

- User Option: transient-substitute-key-function

  This function is used to modify key bindings.  If the value of this
  option is nil (the default), then no substitution is performed.

  This function is called with one argument, the prefix object, and
  must return a key binding description, either the existing key
  description it finds in the ~key~ slot, or the key description that
  replaces the prefix key.  It could be used to make other
  substitutions, but that is discouraged.

  For example, ~=~ is hard to reach using my custom keyboard layout,
  so I substitute ~(~ for that, which is easy to reach using a layout
  optimized for lisp.

  #+BEGIN_SRC emacs-lisp
    (setq transient-substitute-key-function
          (lambda (obj)
            (let ((key (oref obj key)))
              (if (string-match "\\`\\(=\\)[a-zA-Z]" key)
                  (replace-match "(" t t key 1)
                key))))
  #+END_SRC

- User Option: transient-detect-key-conflicts

  This option controls whether key binding conflicts should be
  detected at the time the transient is invoked.  If so, then this
  results in an error, which prevents the transient from being used.
  Because of that, conflicts are ignored by default.

  Conflicts cannot be determined earlier, i.e. when the transient is
  being defined and when new suffixes are being added, because at that
  time there can be false-positives.  It is actually valid for
  multiple suffixes to share a common key binding, provided the
  predicates of those suffixes prevent that more than one of them is
  enabled at a time.

- User Option: transient-force-fixed-pitch

  This option controls whether to force the use of a monospaced font
  in popup buffer.  Even if you use a proportional font for the
  ~default~ face, you might still want to use a monospaced font in
  transient's popup buffer.  Setting this option to t causes ~default~
  to be remapped to ~fixed-pitch~ in that buffer.

* Modifying Existing Transients

To an extent transients can be customized interactively, see [[*Enabling
and Disabling Suffixes]].  This section explains how existing transients
can be further modified non-interactively.

The following functions share a few arguments:

- PREFIX is a transient prefix command, a symbol.

- SUFFIX is a transient infix or suffix specification in the same form
  as expected by ~transient-define-prefix~.  Note that an infix is a
  special kind of suffix.  Depending on context "suffixes" means
  "suffixes (including infixes)" or "non-infix suffixes".  Here it
  means the former.  See [[*Suffix Specifications]].

  SUFFIX may also be a group in the same form as expected by
  ~transient-define-prefix~.  See [[*Group Specifications]].

- LOC is a command, a key vector, a key description (a string as
  returned by ~key-description~), or a list specifying coordinates (the
  last element may also be a command or key). For example ~(1 0 -1)~
  identifies the last suffix (~-1~) of the first subgroup (~0~) of the
  second group (~1~).

  If LOC is a list of coordinates, then it can be used to identify a
  group, not just an individual suffix command.

  The function ~transient-get-suffix~ can be useful to determine whether
  a certain coordination list identifies the suffix or group that you
  expect it to identify.  In hairy cases it may be necessary to look
  at the definition of the transient prefix command.

These functions operate on the information stored in the
~transient--layout~ property of the PREFIX symbol.  Suffix entries in
that tree are not objects but have the form ~(LEVEL CLASS PLIST)~, where
plist should set at least ~:key~, ~:description~ and ~:command~.

- Function: transient-insert-suffix prefix loc suffix

  This function inserts suffix or group SUFFIX into PREFIX before LOC.

- Function: transient-append-suffix prefix loc suffix

  This function inserts suffix or group SUFFIX into PREFIX after LOC.

- Function: transient-replace-suffix prefix loc suffix

  This function replaces the suffix or group at LOC in PREFIX with
  suffix or group SUFFIX.

- Function: transient-remove-suffix prefix loc

  This function removes the suffix or group at LOC in PREFIX.

- Function: transient-get-suffix prefix loc

  This function returns the suffix or group at LOC in PREFIX.  The
  returned value has the form mentioned above.

- Function: transient-suffix-put prefix loc prop value

  This function edits the suffix or group at LOC in PREFIX, by setting
  the PROP of its plist to VALUE.

Most of these functions do not signal an error if they cannot perform
the requested modification.  The functions that insert new suffixes
show a warning if LOC cannot be found in PREFIX, without signaling an
error.  The reason for doing it like this is that establishing a key
binding (and that is what we essentially are trying to do here) should
not prevent the rest of the configuration from loading.  Among these
functions only ~transient-get-suffix~ and ~transient-suffix-put~ may
signal an error.

* Defining New Commands
** Defining Transients

A transient consists of a prefix command and at least one suffix
command, though usually a transient has several infix and suffix
commands.  The below macro defines the transient prefix command *and*
binds the transient's infix and suffix commands.  In other words, it
defines the complete transient, not just the transient prefix command
that is used to invoke that transient.

- Macro: transient-define-prefix name arglist [docstring] [keyword value]... group... [body...]

  This macro defines NAME as a transient prefix command and binds the
  transient's infix and suffix commands.

  ARGLIST are the arguments that the prefix command takes.
  DOCSTRING is the documentation string and is optional.

  These arguments can optionally be followed by keyword-value pairs.
  Each key has to be a keyword symbol, either ~:class~ or a keyword
  argument supported by the constructor of that class.  The
  ~transient-prefix~ class is used if the class is not specified
  explicitly.

  GROUPs add key bindings for infix and suffix commands and specify
  how these bindings are presented in the popup buffer.  At least one
  GROUP has to be specified.  See [[*Binding Suffix and Infix Commands]].

  The BODY is optional.  If it is omitted, then ARGLIST is ignored and
  the function definition becomes:

  #+BEGIN_SRC emacs-lisp
    (lambda ()
      (interactive)
      (transient-setup 'NAME))
  #+END_SRC

  If BODY is specified, then it must begin with an ~interactive~ form
  that matches ARGLIST, and it must call ~transient-setup~.  It may
  however call that function only when some condition is satisfied.

  All transients have a (possibly ~nil~) value, which is exported when
  suffix commands are called, so that they can consume that value.
  For some transients it might be necessary to have a sort of
  secondary value, called a "scope".  Such a scope would usually be
  set in the command's ~interactive~ form and has to be passed to the
  setup function:

  #+BEGIN_SRC emacs-lisp
    (transient-setup 'NAME nil nil :scope SCOPE)
  #+END_SRC

  For example, the scope of the ~magit-branch-configure~ transient is
  the branch whose variables are being configured.

** Binding Suffix and Infix Commands
*** _ :ignore:

The macro ~transient-define-prefix~ is used to define a transient.
This defines the actual transient prefix command (see [[*Defining
Transients]]) and adds the transient's infix and suffix bindings, as
described below.

Users and third-party packages can add additional bindings using
functions such as ~transient-insert-suffix~ (See [[*Modifying Existing
Transients]]).  These functions take a "suffix specification" as one of
their arguments, which has the same form as the specifications used in
~transient-define-prefix~.

*** Group Specifications

The suffix and infix commands of a transient are organized in groups.
The grouping controls how the descriptions of the suffixes are
outlined visually but also makes it possible to set certain properties
for a set of suffixes.

Several group classes exist, some of which organize suffixes in
subgroups.  In most cases the class does not have to be specified
explicitly, but see [[*Group Classes]].

Groups are specified in the call to ~transient-define-prefix~, using
vectors.  Because groups are represented using vectors, we cannot use
square brackets to indicate an optional element and instead use curly
brackets to do the latter.

Group specifications then have this form:

#+BEGIN_SRC emacs-lisp
  [{LEVEL} {DESCRIPTION} {KEYWORD VALUE}... ELEMENT...]
#+END_SRC

The LEVEL is optional and defaults to 4.  See [[*Enabling and Disabling
Suffixes]].

The DESCRIPTION is optional.  If present it is used as the heading of
the group.

The KEYWORD-VALUE pairs are optional.  Each keyword has to be a
keyword symbol, either ~:class~ or a keyword argument supported by the
constructor of that class.

- One of these keywords, ~:description~, is equivalent to specifying
  DESCRIPTION at the very beginning of the vector.  The recommendation
  is to use ~:description~ if some other keyword is also used, for
  consistency, or DESCRIPTION otherwise, because it looks better.

- Likewise ~:level~ is equivalent to LEVEL.

- Other important keywords include the ~:if...~ keywords.  These
  keywords control whether the group is available in a certain
  situation.

  For example, one group of the ~magit-rebase~ transient uses ~:if
  magit-rebase-in-progress-p~, which contains the suffixes that are
  useful while rebase is already in progress; and another that uses
  ~:if-not magit-rebase-in-progress-p~, which contains the suffixes that
  initiate a rebase.

  These predicates can also be used on individual suffixes and are
  only documented once, see [[*Predicate Slots]].

- The value of ~:hide~, if non-nil, is a predicate that controls
  whether the group is hidden by default.  The key bindings for
  suffixes of a hidden group should all use the same prefix key.
  Pressing that prefix key should temporarily show the group and its
  suffixes, which assumes that a predicate like this is used:

  #+BEGIN_SRC emacs-lisp
    (lambda ()
      (eq (car transient--redisplay-key)
          ?\C-c)) ; the prefix key shared by all bindings
  #+END_SRC

- The value of ~:setup-children~, if non-nil, is a function that takes
  two arguments the group object itself and a list of children.
  The children are given as a, potentially empty, list consisting
  of either group or suffix specifications.  It can make arbitrary
  changes to the children including constructing new children from
  scratch.  Also see ~transient-setup-children~.

- The boolean ~:pad-keys~ argument controls whether keys of all suffixes
  contained in a group are right padded, effectively aligning the
  descriptions.

The ELEMENTs are either all subgroups (vectors), or all suffixes
(lists) and strings.  (At least currently no group type exists that
would allow mixing subgroups with commands at the same level, though
in principle there is nothing that prevents that.)

If the ELEMENTs are not subgroups, then they can be a mixture of lists
that specify commands and strings.  Strings are inserted verbatim.
The empty string can be used to insert gaps between suffixes, which is
particularly useful if the suffixes are outlined as a table.

Variables are supported inside group specifications.  For example in
place of a direct subgroup specification, a variable can be used whose
value is a vector that qualifies as a group specification.  Likewise a
variable can be used where a suffix specification is expected.  Lists
of group or suffix specifications are also supported.  Indirect
specifications are resolved when the transient prefix is being
defined.

The form of suffix specifications is documented in the next node.

*** Suffix Specifications

A transient's suffix and infix commands are bound when the transient
prefix command is defined using ~transient-define-prefix~, see
[[*Defining Transients]].  The commands are organized into groups, see
[[*Group Specifications]].  Here we describe the form used to bind an
individual suffix command.

The same form is also used when later binding additional commands
using functions such as ~transient-insert-suffix~, see [[*Modifying
Existing Transients]].

Note that an infix is a special kind of suffix. Depending on context
"suffixes" means "suffixes (including infixes)" or "non-infix
suffixes".  Here it means the former.

Suffix specifications have this form:

#+BEGIN_SRC emacs-lisp
  ([LEVEL] [KEY] [DESCRIPTION] COMMAND|ARGUMENT [KEYWORD VALUE]...)
#+END_SRC

LEVEL, KEY and DESCRIPTION can also be specified using the KEYWORDs
~:level~, ~:key~ and ~:description~.  If the object that is associated with
COMMAND sets these properties, then they do not have to be specified
here.  You can however specify them here anyway, possibly overriding
the object's values just for the binding inside this transient.

- LEVEL is the suffix level, an integer between 1 and 7.  See
  [[*Enabling and Disabling Suffixes]].

- KEY is the key binding, either a vector or key description string.

- DESCRIPTION is the description, either a string or a function that
  returns a string.  The function should be a lambda expression to
  avoid ambiguity.  In some cases a symbol that is bound as a function
  would also work but to be safe you should use ~:description~ in that
  case.

The next element is either a command or an argument.  This is the only
argument that is mandatory in all cases.

- Usually COMMAND is a symbol that is bound as a function, which has
  to be defined or at least autoloaded as a command by the time the
  containing prefix command is invoked.

  Any command will do; it does not need to have an object associated
  with it (as would be the case if ~transient-define-suffix~ or
  ~transient-define-infix~ were used to define it).

  The command can also be a closure or lambda expression, but that
  should only be used for dynamic transients whose suffixes are
  defined when the prefix command is invoked.  See information about
  the ~:setup-children~ function in [[*Group Specifications]].

  As mentioned above, the object that is associated with a command can
  be used to set the default for certain values that otherwise have to
  be set in the suffix specification.  Therefore if there is no object,
  then you have to make sure to specify the KEY and the DESCRIPTION.

- The mandatory argument can also be a command-line argument, a
  string.  In that case an anonymous command is defined and bound.

  Instead of a string, this can also be a list of two strings, in
  which case the first string is used as the short argument (which can
  also be specified using ~:shortarg~) and the second as the long argument
  (which can also be specified using ~:argument~).

  Only the long argument is displayed in the popup buffer.  See
  ~transient-detect-key-conflicts~ for how the short argument may be
  used.

  Unless the class is specified explicitly, the appropriate class is
  guessed based on the long argument.  If the argument ends with "=​"
  (e.g. "--format=") then ~transient-option~ is used, otherwise
  ~transient-switch~.

Finally, details can be specified using optional KEYWORD-VALUE pairs.
Each keyword has to be a keyword symbol, either ~:class~ or a keyword
argument supported by the constructor of that class.  See [[*Suffix
Slots]].

** Defining Suffix and Infix Commands

Note that an infix is a special kind of suffix. Depending on context
"suffixes" means "suffixes (including infixes)" or "non-infix
suffixes".

- Macro: transient-define-suffix name arglist [docstring] [keyword value]... body...

  This macro defines NAME as a transient suffix command.

  ARGLIST are the arguments that the command takes.
  DOCSTRING is the documentation string and is optional.

  These arguments can optionally be followed by keyword-value pairs.
  Each keyword has to be a keyword symbol, either ~:class~ or a keyword
  argument supported by the constructor of that class.  The
  ~transient-suffix~ class is used if the class is not specified
  explicitly.

  The BODY must begin with an ~interactive~ form that matches ARGLIST.
  The infix arguments are usually accessed by using ~transient-args~
  inside ~interactive~.

- Macro: transient-define-infix name arglist [docstring] [keyword value]...

  This macro defines NAME as a transient infix command.

  ARGLIST is always ignored (but mandatory never-the-less) and
  reserved for future use.  DOCSTRING is the documentation string and
  is optional.

  The keyword-value pairs are mandatory.  All transient infix commands
  are ~equal~ to each other (but not ~eq~), so it is meaningless to define
  an infix command without also setting at least ~:class~ and one other
  keyword (which it is depends on the used class, usually ~:argument~ or
  ~:variable~).

  Each keyword has to be a keyword symbol, either ~:class~ or a keyword
  argument supported by the constructor of that class.  The
  ~transient-switch~ class is used if the class is not specified
  explicitly.

  The function definition is always:

  #+BEGIN_SRC emacs-lisp
    (lambda ()
      (interactive)
      (let ((obj (transient-suffix-object)))
        (transient-infix-set obj (transient-infix-read obj)))
      (transient--show))
  #+END_SRC

  ~transient-infix-read~ and ~transient-infix-set~ are generic functions.
  Different infix commands behave differently because the concrete
  methods are different for different infix command classes.  In rare
  cases the above command function might not be suitable, even if you
  define your own infix command class.  In that case you have to use
  ~transient-suffix-command~ to define the infix command and use ~t~ as
  the value of the ~:transient~ keyword.

- Macro: transient-define-argument name arglist [docstring] [keyword value]...

  This macro defines NAME as a transient infix command.

  This is an alias for ~transient-define-infix~.  Only use this alias
  to define an infix command that actually sets an infix argument.
  To define an infix command that, for example, sets a variable, use
  ~transient-define-infix~ instead.

** Using Infix Arguments

The function and the variables described below allow suffix commands
to access the value of the transient from which they were invoked;
which is the value of its infix arguments.  These variables are set
when the user invokes a suffix command that exits the transient, but
before actually calling the command.

When returning to the command-loop after calling the suffix command,
the arguments are reset to ~nil~ (which causes the function to return
~nil~ too).

Like for Emacs' prefix arguments it is advisable, but not mandatory,
to access the infix arguments inside the command's ~interactive~ form.
The preferred way of doing that is to call the ~transient-args~
function, which for infix arguments serves about the same purpose as
~prefix-arg~ serves for prefix arguments.

- Function: transient-args prefix

  This function returns the value of the transient prefix command
  PREFIX.

  If the current command was invoked from the transient prefix command
  PREFIX, then it returns the active infix arguments.  If the current
  command was not invoked from PREFIX, then it returns the set, saved
  or default value for PREFIX.

- Function: transient-arg-value arg args

  This function return the value of ARG as it appears in ARGS.

  For a switch a boolean is returned.  For an option the value is
  returned as a string, using the empty string for the empty value,
  or nil if the option does not appear in ARGS.

- Function: transient-suffixes prefix

  This function returns the suffixes of the transient prefix command
  PREFIX.  This is a list of objects.  This function should only be
  used if you need the objects (as opposed to just their values) and
  if the current command is not being invoked from PREFIX.

- Variable: transient-current-suffixes

  The suffixes of the transient from which this suffix command was
  invoked.  This is a list of objects.  Usually it is sufficient to
  instead use the function ~transient-args~, which returns a list of
  values.  In complex cases it might be necessary to use this variable
  instead, i.e. if you need access to information beside the value.

- Variable: transient-current-prefix

  The transient from which this suffix command was invoked.  The
  returned value is a ~transient-prefix~ object, which holds information
  associated with the transient prefix command.

- Variable: transient-current-command

  The transient from which this suffix command was invoked.  The
  returned value is a symbol, the transient prefix command.

** Transient State
*** _ :ignore:

Invoking a transient prefix command "activates" the respective
transient, i.e. it puts a transient keymap into effect, which binds
the transient's infix and suffix commands.

The default behavior while a transient is active is as follows:

- Invoking an infix command does not affect the transient state; the
  transient remains active.

- Invoking a (non-infix) suffix command "deactivates" the transient
  state by removing the transient keymap and performing some
  additional cleanup.

- Invoking a command that is bound in a keymap other than the
  transient keymap is disallowed and trying to do so results in a
  warning.  This does not "deactivate" the transient.

But these are just the defaults.  Whether a certain command
deactivates or "exits" the transient is configurable.  There is more
than one way in which a command can be "transient" or "non-transient";
the exact behavior is implemented by calling a so-called "pre-command"
function.  Whether non-suffix commands are allowed to be called is
configurable per transient.

- The transient-ness of suffix commands (including infix commands) is
  controlled by the value of their ~transient~ slot, which can be set
  either when defining the command or when adding a binding to a
  transient while defining the respective transient prefix command.

  Valid values are booleans and the pre-commands described below.

  - ~t~ is equivalent to ~transient--do-stay~.
  - ~nil~ is equivalent to ~transient--do-exit~.
  - If ~transient~ is unbound (and that is actually the default for
    non-infix suffixes) then the value of the prefix's
    ~transient-suffix~ slot is used instead.  The default value of that
    slot is ~nil~, so the suffix's ~transient~ slot being unbound is
    essentially equivalent to it being ~nil~.

- A suffix command can be a prefix command itself, i.e. a
  "sub-prefix".  While a sub-prefix is active we nearly always want
  ~C-g~ to take the user back to the "super-prefix".  However in rare
  cases this may not be desirable, and that makes the following
  complication necessary:

  For ~transient-suffix~ objects the ~transient~ slot is unbound.  We can
  ignore that for the most part because, as stated above, ~nil~ and the
  slot being unbound are equivalent, and mean "do exit".  That isn't
  actually true for suffixes that are sub-prefixes though.  For such
  suffixes unbound means "do exit but allow going back", which is the
  default, while ~nil~ means "do exit permanently", which requires that
  slot to be explicitly set to that value.

- The transient-ness of certain built-in suffix commands is specified
  using ~transient-predicate-map~.  This is a special keymap, which
  binds commands to pre-commands (as opposed to keys to commands) and
  takes precedence over the ~transient~ slot.

The available pre-command functions are documented below.  They are
called by ~transient--pre-command~, a function on ~pre-command-hook~ and
the value that they return determines whether the transient is exited.
To do so the value of one of the constants ~transient--exit~ or
~transient--stay~ is used (that way we don't have to remember if ~t~ means
"exit" or "stay").

Additionally these functions may change the value of ~this-command~
(which explains why they have to be called using ~pre-command-hook~),
call ~transient-export~, ~transient--stack-zap~ or ~transient--stack-push~;
and set the values of ~transient--exitp~, ~transient--helpp~ or
~transient--editp~.

*** Pre-commands for Infixes
:PROPERTIES:
:UNNUMBERED: notoc
:END:

The default for infixes is ~transient--do-stay~.  This is also the only
function that makes sense for infixes.

- Function: transient--do-stay

  Call the command without exporting variables and stay transient.

*** Pre-commands for Suffixes
:PROPERTIES:
:UNNUMBERED: notoc
:END:

The default for suffixes is ~transient--do-exit~.

- Function: transient--do-exit

  Call the command after exporting variables and exit the transient.

- Function: transient--do-call

  Call the command after exporting variables and stay transient.

- Function: transient--do-replace

  Call the transient prefix command, replacing the active transient.

  This is used for suffixes that are prefixes themselves, i.e. for
  sub-prefixes.

*** Pre-commands for Non-Suffixes
:PROPERTIES:
:UNNUMBERED: notoc
:END:

The default for non-suffixes, i.e commands that are bound in other
keymaps beside the transient keymap, is ~transient--do-warn~.  Silently
ignoring the user-error is also an option, though probably not a good
one.

If you want to let the user invoke non-suffix commands, then use
~transient--do-stay~ as the value of the prefix's ~transient-non-suffix~
slot.

- Function: transient--do-warn

  Call ~transient-undefined~ and stay transient.

- Function: transient--do-noop

  Call ~transient-noop~ and stay transient.

*** Special Pre-Commands
:PROPERTIES:
:UNNUMBERED: notoc
:END:

- Function: transient--do-quit-one

  If active, quit help or edit mode, else exit the active transient.

  This is used when the user pressed ~C-g~.

- Function: transient--do-quit-all

  Exit all transients without saving the transient stack.

  This is used when the user pressed ~C-q~.

- Function: transient--do-suspend

  Suspend the active transient, saving the transient stack.

  This is used when the user pressed ~C-z~.

* Classes and Methods
** _ :ignore:

Transient uses classes and generic functions to make it possible to
define new types of suffix commands that are similar to existing
types, but behave differently in some aspects.  It does the same for
groups and prefix commands, though at least for prefix commands that
*currently* appears to be less important.

Every prefix, infix and suffix command is associated with an object,
which holds information that controls certain aspects of its behavior.
This happens in two ways.

- Associating a command with a certain class gives the command a type.
  This makes it possible to use generic functions to do certain things
  that have to be done differently depending on what type of command
  it acts on.

  That in turn makes it possible for third-parties to add new types
  without having to convince the maintainer of Transient that that new
  type is important enough to justify adding a special case to a dozen
  or so functions.

- Associating a command with an object makes it possible to easily
  store information that is specific to that particular command.

  Two commands may have the same type, but obviously their key
  bindings and descriptions still have to be different, for example.

  The values of some slots are functions.  The ~reader~ slot for example
  holds a function that is used to read a new value for an infix
  command.  The values of such slots are regular functions.

  Generic functions are used when a function should do something
  different based on the type of the command, i.e. when all commands
  of a certain type should behave the same way but different from the
  behavior for other types.  Object slots that hold a regular function
  as value are used when the task that they perform is likely to
  differ even between different commands of the same type.

** Group Classes

The type of a group can be specified using the ~:class~ property at the
beginning of the class specification, e.g. ~[:class transient-columns
...]~ in a call to ~transient-define-prefix~.

- The abstract ~transient-child~ class is the base class of both
  ~transient-group~ (and therefore all groups) as well as of
  ~transient-suffix~ (and therefore all suffix and infix commands).

  This class exists because the elements (aka "children") of certain
  groups can be other groups instead of suffix and infix commands.

- The abstract ~transient-group~ class is the superclass of all other
  group classes.

- The ~transient-column~ class is the simplest group.

  This is the default "flat" group.  If the class is not specified
  explicitly and the first element is not a vector (i.e. not a group),
  then this class is used.

  This class displays each element on a separate line.

- The ~transient-row~ class displays all elements on a single line.

- The ~transient-columns~ class displays commands organized in columns.

  Direct elements have to be groups whose elements have to be commands
  or strings.  Each subgroup represents a column.  This class takes
  care of inserting the subgroups' elements.

  This is the default "nested" group.  If the class is not specified
  explicitly and the first element is a vector (i.e. a group), then
  this class is used.

- The ~transient-subgroups~ class wraps other groups.

  Direct elements have to be groups whose elements have to be commands
  or strings.  This group inserts an empty line between subgroups.
  The subgroups themselves are responsible for displaying their
  elements.

** Group Methods

- Function: transient-setup-children group children

  This generic function can be used to setup the children or a group.

  The default implementation usually just returns the children
  unchanged, but if the ~setup-children~ slot of GROUP is non-nil, then
  it calls that function with CHILDREN as the only argument and
  returns the value.

  The children are given as a, potentially empty, list consisting of
  either group or suffix specifications.  These functions can make
  arbitrary changes to the children including constructing new
  children from scratch.

- Function: transient--insert-group group

  This generic function formats the group and its elements and inserts
  the result into the current buffer, which is a temporary buffer.
  The contents of that buffer are later inserted into the popup buffer.

  Functions that are called by this function may need to operate in
  the buffer from which the transient was called.  To do so they can
  temporarily make the ~transient--source-buffer~ the current buffer.

** Prefix Classes

Currently the ~transient-prefix~ class is being used for all prefix
commands and there is only a single generic function that can be
specialized based on the class of a prefix command.

- Function: transient--history-init obj

  This generic function is called while setting up the transient and
  is responsible for initializing the ~history~ slot.  This is the
  transient-wide history; many individual infixes also have a history
  of their own.

  The default (and currently only) method extracts the value from the
  global variable ~transient-history~.

A transient prefix command's object is stored in the ~transient--prefix~
property of the command symbol.  While a transient is active, a clone
of that object is stored in the variable ~transient--prefix~.  A clone
is used because some changes that are made to the active transient's
object should not affect later invocations.

** Suffix Classes

- All suffix and infix classes derive from ~transient-suffix~, which in
  turn derives from ~transient-child~, from which ~transient-group~ also
  derives (see [[*Group Classes]]).

- All infix classes derive from the abstract ~transient-infix~ class,
  which in turn derives from the ~transient-suffix~ class.

  Infixes are a special type of suffixes.  The primary difference is
  that infixes always use the ~transient--do-stay~ pre-command, while
  non-infix suffixes use a variety of pre-commands (see [[*Transient
  State]]).  Doing that is most easily achieved by using this class,
  though theoretically it would be possible to define an infix class
  that does not do so.  If you do that then you get to implement many
  methods.

  Also, infixes and non-infix suffixes are usually defined using
  different macros (see [[*Defining Suffix and Infix Commands]]).

- Classes used for infix commands that represent arguments should
  be derived from the abstract ~transient-argument~ class.

- The ~transient-switch~ class (or a derived class) is used for infix
  arguments that represent command-line switches (arguments that do
  not take a value).

- The ~transient-option~ class (or a derived class) is used for infix
  arguments that represent command-line options (arguments that do
  take a value).

- The ~transient-switches~ class can be used for a set of mutually
  exclusive command-line switches.

- The ~transient-files~ class can be used for a "--" argument that
  indicates that all remaining arguments are files.

- Classes used for infix commands that represent variables should
  derived from the abstract ~transient-variables~ class.

Magit defines additional classes, which can serve as examples for the
fancy things you can do without modifying Transient.  Some of these
classes will likely get generalized and added to Transient.  For now
they are very much subject to change and not documented.

** Suffix Methods
*** _ :ignore:

To get information about the methods implementing these generic
functions use ~describe-function~.

*** Suffix Value Methods

- Function: transient-init-value obj

  This generic function sets the initial value of the object OBJ.

  This function is called for all suffix commands, but unless a
  concrete method is implemented this falls through to the default
  implementation, which is a noop.  In other words this usually
  only does something for infix commands, but note that this is
  not implemented for the abstract class ~transient-infix~, so if
  your class derives from that directly, then you must implement
  a method.

- Function: transient-infix-read obj

  This generic function determines the new value of the infix object
  OBJ.

  This function merely determines the value; ~transient-infix-set~ is
  used to actually store the new value in the object.

  For most infix classes this is done by reading a value from the
  user using the reader specified by the ~reader~ slot (using the
  ~transient-infix-value~ method described below).

  For some infix classes the value is changed without reading
  anything in the minibuffer, i.e. the mere act of invoking the
  infix command determines what the new value should be, based
  on the previous value.

- Function: transient-prompt obj

  This generic function returns the prompt to be used to read infix
  object OBJ's value.

- Function: transient-infix-set obj value

  This generic function sets the value of infix object OBJ to VALUE.

- Function: transient-infix-value obj

  This generic function returns the value of the suffix object OBJ.

  This function is called by ~transient-args~ (which see), meaning this
  function is how the value of a transient is determined so that the
  invoked suffix command can use it.

  Currently most values are strings, but that is not set in stone.
  ~nil~ is not a value, it means "no value".

  Usually only infixes have a value, but see the method for
  ~transient-suffix~.

- Function: transient-init-scope obj

  This generic function sets the scope of the suffix object OBJ.

  The scope is actually a property of the transient prefix, not of
  individual suffixes.  However it is possible to invoke a suffix
  command directly instead of from a transient.  In that case, if
  the suffix expects a scope, then it has to determine that itself
  and store it in its ~scope~ slot.

  This function is called for all suffix commands, but unless a
  concrete method is implemented this falls through to the default
  implementation, which is a noop.

*** Suffix Format Methods

- Function: transient-format obj

  This generic function formats and returns OBJ for display.

  When this function is called, then the current buffer is some
  temporary buffer.  If you need the buffer from which the prefix
  command was invoked to be current, then do so by temporarily
  making ~transient--source-buffer~ current.

- Function: transient-format-key obj

  This generic function formats OBJ's ~key~ for display and returns the
  result.

- Function: transient-format-description obj

  This generic function formats OBJ's ~description~ for display and
  returns the result.

- Function: transient-format-value obj

  This generic function formats OBJ's value for display and returns
  the result.

- Function: transient-show-help obj

  Show help for the prefix, infix or suffix command represented by
  OBJ.

  For prefixes, show the info manual, if that is specified using the
  ~info-manual~ slot.  Otherwise show the manpage if that is specified
  using the ~man-page~ slot.  Otherwise show the command's doc string.

  For suffixes, show the command's doc string.

  For infixes, show the manpage if that is specified.  Otherwise show
  the command's doc string.

** Prefix Slots

- ~man-page~ or ~info-manual~ can be used to specify the documentation for
  the prefix and its suffixes.  The command ~transient-help~ uses the
  method ~transient-show-help~ (which see) to lookup and use these
  values.

- ~history-key~ If multiple prefix commands should share a single value,
  then this slot has to be set to the same value for all of them.  You
  probably don't want that.

- ~transient-suffix~ and ~transient-non-suffix~ play a part when
  determining whether the currently active transient prefix command
  remains active/transient when a suffix or abitrary non-suffix
  command is invoked.  See [[*Transient State]].

- ~incompatible~ A list of lists.  Each sub-list specifies a set of
  mutually exclusive arguments.  Enabling one of these arguments
  causes the others to be disabled.

- ~scope~ For some transients it might be necessary to have a sort of
  secondary value, called a "scope".  See ~transient-define-prefix~.

*** Internal Prefix Slots
:PROPERTIES:
:UNNUMBERED: notoc
:END:

These slots are mostly intended for internal use.  They should not be
set in calls to ~transient-define-prefix~.

- ~prototype~ When a transient prefix command is invoked, then a clone
  of that object is stored in the global variable ~transient--prefix~
  and the prototype is stored in the clone's ~prototype~ slot.

- ~command~ The command, a symbol.  Each transient prefix command
  consists of a command, which is stored in a symbol's function slot
  and an object, which is stored in the ~transient--prefix~ property
  of the same symbol.

- ~level~ The level of the prefix commands.  The suffix commands whose
  layer is equal or lower are displayed.  See [[*Enabling and Disabling
  Suffixes]].

- ~value~ The likely outdated value of the prefix.  Instead of accessing
  this slot directly you should use the function ~transient-get-value~,
  which is guaranteed to return the up-to-date value.

- ~history~ and ~history-pos~ are used to keep track of historic values.
  Unless you implement your own ~transient-infix-read~ method you should
  not have to deal with these slots.

** Suffix Slots

Here we document most of the slots that are only available for suffix
objects.  Some slots are shared by suffix and group objects, they are
documented in [[*Predicate Slots]].

Also see [[*Suffix Classes]].

*** Slots of ~transient-suffix~
:PROPERTIES:
:UNNUMBERED: notoc
:END:

- ~key~ The key, a key vector or a key description string.

- ~command~ The command, a symbol.

- ~transient~ Whether to stay transient.  See [[*Transient State]].

- ~format~ The format used to display the suffix in the popup buffer.
  It must contain the following %-placeholders:

  - ~%k~ For the key.
  - ~%d~ For the description.
  - ~%v~ For the infix value.  Non-infix suffixes don't have a value.

- ~description~ The description, either a string or a function that is
  called with no argument and returns a string.

*** Slots of ~transient-infix~
:PROPERTIES:
:UNNUMBERED: notoc
:END:

Some of these slots are only meaningful for some of the subclasses.
They are defined here anyway to allow sharing certain methods.

- ~argument~ The long argument, e.g. ~--verbose~.

- ~shortarg~ The short argument, e.g. ~-v~.

- ~value~ The value.  Should not be accessed directly.

- ~init-value~ Function that is responsable for setting the object's
  value.  If bound, then this is called with the object as the only
  argument.  Usually this is not bound, in which case the object's
  primary ~transient-init-value~ method is called instead.

- ~unsavable~ Whether the value of the suffix is not saved as part of
  the prefixes.

- ~multi-value~ For options, whether the option can have multiple
  values.  If non-nil, then default to use ~completing-read-multiple~.

- ~always-read~ For options, whether to read a value on every invocation.
  If this is nil, then options that have a value are simply unset and
  have to be invoked a second time to set a new value.

- ~allow-empty~ For options, whether the empty string is a valid value.

- ~history-key~ The key used to store the history.  This defaults to the
  command name.  This is useful when multiple infixes should share the
  same history because their values are of the same kind.

- ~reader~ The function used to read the value of an infix.  Not used
  for switches.  The function takes three arguments, PROMPT,
  INITIAL-INPUT and HISTORY, and must return a string.

- ~prompt~ The prompt used when reading the value, either a string or a
  function that takes the object as the only argument and which
  returns a prompt string.

- ~choices~ A list of valid values.  How exactly that is used depends on
  the class of the object.

*** Slots of ~transient-variable~
:PROPERTIES:
:UNNUMBERED: notoc
:END:

- ~variable~ The variable.

*** Slots of ~transient-switches~
:PROPERTIES:
:UNNUMBERED: notoc
:END:

- ~argument-format~ The display format.  Must contain ~%s~, one of the
  ~choices~ is substituted for that.  E.g. ~--%s-order~.

- ~argument-regexp~ The regexp used to match any one of the switches.
  E.g. ~\\(--\\(topo\\|author-date\\|date\\)-order\\)~.

** Predicate Slots

Suffix and group objects share some predicate slots that control
whether a group or suffix should be available depending on some state.
Only one of these slots can be used at the same time.  It is undefined
what happens if you use more than one.

- ~if~ Enable if predicate returns non-nil.
- ~if-not~ Enable if predicate returns nil.
- ~if-non-nil~ Enable if variable's value is non-nil.
- ~if-nil~ Enable if variable's value is nil.
- ~if-mode~ Enable if major-mode matches value.
- ~if-not-mode~ Enable if major-mode does not match value.
- ~if-derived~ Enable if major-mode derives from value.
- ~if-not-derived~ Enable if major-mode does not derive from value.

One more slot is shared between group and suffix classes, ~level~.  Like
the slots documented above, it is a predicate, but it is used for a
different purpose.  The value has to be an integer between 1
and 7.  ~level~ controls whether a suffix or a group should be
available depending on user preference.
See [[*Enabling and Disabling Suffixes]].

* Related Abstractions and Packages
** Comparison With Prefix Keys and Prefix Arguments

While transient commands were inspired by regular prefix keys and
prefix arguments, they are also quite different and much more complex.

The following diagrams illustrate some of the differences.

- ~(c)~ represents a return to the command loop.
- ~(+)~ represents the user's choice to press one key or another.
- ~{WORD}~ are possible behaviors.
- ~{NUMBER}~ is a footnote.

*** Regular Prefix Commands
:PROPERTIES:
:UNNUMBERED: notoc
:END:

See [[info:elisp#Prefix Keys]].

#+BEGIN_EXAMPLE
                                    ,--> command1 --> (c)
                                    |
  (c)-(+)-> prefix command or key --+--> command2 --> (c)
                                    |
                                    `--> command3 --> (c)
#+END_EXAMPLE

*** Regular Prefix Arguments
:PROPERTIES:
:UNNUMBERED: notoc
:END:

See [[info:elisp#Prefix Command Arguments]].

#+BEGIN_EXAMPLE
          ,----------------------------------,
          |                                  |
          v                                  |
  (c)-(+)---> prefix argument command --(c)-(+)-> any command --> (c)
                 |                                        ^        |
                 |                                        |        |
                 `-- sets or changes --, ,-- maybe used --'        |
                                       | |                         |
                                       v |                         |
                            prefix argument state                  |
                                        ^                          |
                                        |                          |
                                        `-------- discards --------'
#+END_EXAMPLE

*** Transients
:PROPERTIES:
:UNNUMBERED: notoc
:END:

(∩｀-´)⊃━☆ﾟ.*･｡ﾟ

This diagram ignores the infix value and external state:

#+BEGIN_EXAMPLE
  (c)
   |        ,- {stay} ------<-,-<------------<-,-<---,
  (+)       |                 |                |     |
   |        |                 |                |     |
   |        |   ,--> infix1 --|                |     |
   |        |   |             |                |     |
   |        |   |--> infix2 --|                |     |
   v        v   |             |                |     |
   prefix -(c)-(+)-> infix3 --'                ^     |
                |                              |     |
                |---------------> suffix1 -->--|     |
                |                              |     |
                |---------------> suffix2 ----{1}------> {exit} --> (c)
                |                                    |
                |---------------> suffix3 -------------> {exit} --> (c)
                |                                    |
                `--> any command --{2}-> {warn} -->--|
                                    |                |
                                    |--> {noop} -->--|
                                    |                |
                                    |--> {call} -->--'
                                    |
                                    `------------------> {exit} --> (c)
#+END_EXAMPLE

This diagram takes the infix value into account to an extend, while
still ignoring external state:

#+BEGIN_EXAMPLE
  (c)
   |        ,- {stay} ------<-,-<------------<-,-<---,
  (+)       |                 |                |     |
   |        |                 |                |     |
   |        |   ,--> infix1 --|                |     |
   |        |   |    |        |                |     |
   |        |   ,--> infix2 --|                |     |
   v        v   |    |        |                |     |
   prefix -(c)-(+)-> infix3 --'                |     |
                |    |                         ^     |
                |    |                         |     |
                |---------------> suffix1 -->--|     |
                |    |             ^           |     |
                |    |             |           |     |
                |---------------> suffix2 ----{1}------> {exit} --> (c)
                |    |             ^                 |     |
                |    |             |                 |     v
                |    |             |                 |     |
                |---------------> suffix3 -------------> {exit} --> (c)
                |    |             ^                 |     |
                | sets             |                 |     v
                |    |             maybe             |     |
                |    |             used              |     |
                |    |             |                 |     |
                |    |     infix --'                 |     |
                |    `---> value                     |     |
                |           ^                        |     |
                |           |                        |     |
                |       hides                        |     |
                |           |                        |     |
                |           `--------------------------<---|
                |                                    |     |
                `--> any command --{2}-> {warn} -->--|     |
                                    |                |     |
                                    |--> {noop} -->--|     |
                                    |                |     |
                                    |--> {call} -->--'     ^
                                    |                      |
                                    `------------------> {exit} --> (c)
#+END_EXAMPLE

This diagram provides more information about the infix value
and also takes external state into account.

#+BEGIN_EXAMPLE
                                         ,----sets--- "anything"
                                         |
                                         v
                        ,---------> external
                        |           state
                        |            | |
                        |  initialized |                      ☉‿⚆
                     sets         from |
                        |            | maybe
                        | ,----------' used
                        | |            |
  (c)                   | |            v
   |        ,- {stay} --|---<-,-<------|-----<-,-<---,
  (+)       |           | |   |        |       |     |
   |        |           | v   |        |       |     |
   |        |   ,--> infix1 --|        |       |     |
   |        |   |       | |   |        |       |     |
   |        |   |       | v   |        |       |     |
   |        |   ,--> infix2 --|        |       |     |
   |        |   |    | ^      |        |       |     |
   v        v   |    | |      |        |       |     |
   prefix -(c)-(+)-> infix3 --'        |       |     |
                |    | ^               |       ^     |
                |    | |               v       |     |
                |---------------> suffix1 -->--|     |
                |    | |            ^  |       |     |
                |    | |            |  v       |     |
                |---------------> suffix2 ----{1}------> {exit} --> (c)
                |    | |            ^  |             |     |
                |    | |            |  |             |     v
                |    | |            |  v             |     |
                |---------------> suffix3 -------------> {exit} --> (c)
                |    | |            ^                |     |
                | sets |            |                |     v
                |    | initialized  maybe            |     |
                |    | from         used             |     |
                |    | |            |                |     |
                |    | `-- infix ---'                |     |
                |    `---> value -----------------------------> persistent
                |           ^ ^                      |     |    across
                |           | |                      |     |    invocations -,
                |       hides |                      |     |                 |
                |           | `----------------------------------------------'
                |           |                        |     |
                |           `--------------------------<---|
                |                                    |     |
                `--> any command --{2}-> {warn} -->--|     |
                                    |                |     |
                                    |--> {noop} -->--|     |
                                    |                |     |
                                    |--> {call} -->--'     ^
                                    |                      |
                                    `------------------> {exit} --> (c)
#+END_EXAMPLE

- ~{1}~ Transients can be configured to be exited when a suffix command
  is invoked.  The default is to do so for all suffixes except for
  those that are common to all transients and which are used to
  perform tasks such as providing help and saving the value of the
  infix arguments for future invocations.  The behavior can also be
  specified for individual suffix commands and may even depend on
  state.

- ~{2}~ Transients can be configured to allow the user to invoke
  non-suffix commands.  The default is to not allow that and instead
  warn the user.

Despite already being rather complex, even the last diagram leaves out
many details.  Most importantly it implies that the decision whether
to remain transient is made later than it actually is made (for the
most part a function on ~pre-command-hook~ is responsible).  But such
implementation details are of little relevance to users and are
covered elsewhere.

** Comparison With Other Packages
*** Magit-Popup
:PROPERTIES:
:UNNUMBERED: notoc
:END:

Transient is the successor to Magit-Popup (see [[info:magit-popup]]).

One major difference between these two implementations of the same
ideas is that while Transient uses transient keymaps and embraces the
command-loop, Magit-Popup implemented an inferior mechanism that does
not use transient keymaps and that instead of using the command-loop
implements a naive alternative based on ~read-char~.

Magit-Popup does not use classes and generic functions and defining a
new command type is near impossible as it involves adding hard-coded
special-cases to many functions.  Because of that only a single new
type was added, which was not already part of Magit-Popup's initial
release.

A lot of things are hard-coded in Magit-Popup.  One random example is
that the key bindings for switches must begin with "-" and those for
options must begin with "=".

*** Hydra
:PROPERTIES:
:UNNUMBERED: notoc
:END:

Hydra (see https://github.com/abo-abo/hydra) is another package that
provides features similar to those of Transient.

Both packages use transient keymaps to make a set of commands
temporarily available and show the available commands in a popup
buffer.

A Hydra "body" is equivalent to a Transient "prefix" and a Hydra
"head" is equivalent to a Transient "suffix".  Hydra has no equivalent
of a Transient "infix".

Both hydras and transients can be used as simple command dispatchers.
Used like this they are similar to regular prefix commands and prefix
keys, except that the available commands are shown in the popup buffer.

(Another package that does this is ~which-key~. It does so automatically
for any incomplete key sequence.  The advantage of that approach is
that no additional work is necessary; the disadvantage is that the
available commands are not organized semantically.)

Both Hydra and Transient provide features that go beyond simple
command dispatchers:

- Invoking a command from a hydra does not necessarily exit the hydra.
  That makes it possible to invoke the same command again, but using a
  shorter key sequence (i.e. the key that was used to enter the hydra
  does not have to be pressed again).

  Transient supports that too, but for now this feature is not a focus
  and the interface is a bit more complicated.  A very basic example
  using the current interface:

  #+BEGIN_SRC emacs-lisp
    (transient-define-prefix outline-navigate ()
      :transient-suffix     'transient--do-stay
      :transient-non-suffix 'transient--do-warn
      [("p" "previous visible heading" outline-previous-visible-heading)
       ("n" "next visible heading" outline-next-visible-heading)])
  #+END_SRC

- Transient supports infix arguments; values that are set by infix
  commands and then consumed by the invoked suffix command(s).

  To my knowledge, Hydra does not support that.

Both packages make it possible to specify how exactly the available
commands are outlined:

- With Hydra this is often done using an explicit format string, which
  gives authors a lot of flexibility and makes it possible to do fancy
  things.

  The downside of this is that it becomes harder for a user to add
  additional commands to an existing hydra and to change key bindings.

- Transient allows the author of a transient to organize the commands
  into groups and the use of generic functions allows authors of
  transients to control exactly how a certain command type is
  displayed.

  However while Transient supports giving sections a heading it does
  not currently support giving the displayed information more
  structure by, for example, using box-drawing characters.

  That could be implemented by defining a new group class, which lets
  the author specify a format string.  It should be possible to
  implement that without modifying any existing code, but it does not
  currently exist.

* FAQ
:PROPERTIES:
:APPENDIX:   t
:END:

** Can I control how the popup buffer is displayed?
:PROPERTIES:
:UNNUMBERED: notoc
:END:

Yes, see ~transient-display-buffer-action~ in [[*Other Options]].

** Why did some of the key bindings change?
:PROPERTIES:
:UNNUMBERED: notoc
:END:

You may have noticed that the bindings for some of the common commands
do *not* have the prefix ~C-x~ and that furthermore some of these commands
are grayed out while others are not.  That unfortunately is a bit
confusing if the section of common commands is not shown permanently,
making the following explanation necessary.

The purpose of usually hiding that section but showing it after the
user pressed the respective prefix key is to conserve space and not
overwhelm users with too much noise, while allowing the user to
quickly list common bindings on demand.

That however should not keep us from using the best possible key
bindings.  The bindings that do use a prefix do so to avoid wasting
too many non-prefix bindings, keeping them available for use in
individual transients.  The bindings that do not use a prefix and that
are *not* grayed out are very important bindings that are *always*
available, even when invoking the "common command key prefix" or *any
other* transient-specific prefix.  The non-prefix keys that *are* grayed
out however, are not available when any incomplete prefix key sequence
is active.  They do not use the "common command key prefix" because it
is likely that users want to invoke them several times in a row and
e.g. ~M-p M-p M-p~ is much more convenient than ~C-x M-p C-x M-p C-x M-p~.

You may also have noticed that the "Set" command is bound to ~C-x s~,
while Magit-Popup used to bind ~C-c C-c~ instead.  I have seen several
users praise the latter binding (sic), so I did not change it
willy-nilly.  The reason that I changed it is that using different
prefix keys for different common commands, would have made the
temporary display of the common commands even more confusing,
i.e. after pressing ~C-c~ all the ~C-x ...~ bindings would be grayed out.

Using a single prefix for common commands key means that all other
potential prefix keys can be used for transient-specific commands
*without* the section of common commands also popping up.  ~C-c~ in
particular is a prefix that I want to (and already do) use for Magit, and
also using that for a common command would prevent me from doing so.

(Also see the next question.)

** Why does ~q~ not quit popups anymore?
:PROPERTIES:
:UNNUMBERED: notoc
:END:

I agree that ~q~ is a good binding for commands that quit something.
This includes quitting whatever transient is currently active, but it
also includes quitting whatever it is that some specific transient is
controlling.  The transient ~magit-blame~ for example binds ~q~ to the
command that turns ~magit-blame-mode~ off.

So I had to decide if ~q~ should quit the active transient (like
Magit-Popup used to) or whether ~C-g~ should do that instead, so that ~q~
could be bound in individual transient to whatever commands make sense
for them.  Because all other letters are already reserved for use by
individual transients, I have decided to no longer make an exception
for ~q~.

If you want to get ~q~'s old binding back then you can do so.  Doing
that is a bit more complicated than changing a single key binding, so
I have implemented a function, ~transient-bind-q-to-quit~ that makes the
necessary changes.  See its doc string for more information.

* Keystroke Index
:PROPERTIES:
:APPENDIX:   t
:INDEX:      ky
:COOKIE_DATA: recursive
:END:
* Command Index
:PROPERTIES:
:APPENDIX:   t
:INDEX:      cp
:END:
* Function Index
:PROPERTIES:
:APPENDIX:   t
:INDEX:      fn
:END:
* Variable Index
:PROPERTIES:
:APPENDIX:   t
:INDEX:      vr
:END:

* _ Copying
:PROPERTIES:
:COPYING:    t
:END:

#+BEGIN_QUOTE
Copyright (C) 2018-2021 Jonas Bernoulli <jonas@bernoul.li>

You can redistribute this document and/or modify it under the terms
of the GNU General Public License as published by the Free Software
Foundation, either version 3 of the License, or (at your option) any
later version.

This document is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
General Public License for more details.
#+END_QUOTE

* _ :ignore:

#  LocalWords:  ARGLIST ARGS DOCSTRING ELEMENTs EVAL GROUPs Infixes
#  LocalWords:  Infixes KEYWORDs LOC LocalWords MERCHANTABILITY Magit
#  LocalWords:  Magit's Makefile OBJ OBJ's Pre arglist
#  LocalWords:  args boolean booleans customizable docstring eval
#  LocalWords:  featurep infixes init keymap keymaps loc magit manpage
#  LocalWords:  minibuffer ness nilly noop plist pre prev
#  LocalWords:  rebase src subclass subclasses subprocess superclass
#  LocalWords:  utils

# IMPORTANT: Also update ORG_ARGS and ORG_EVAL in the Makefile.
# Local Variables:
# eval: (require 'ox-extra    nil t)
# eval: (require 'ox-texinfo+ nil t)
# eval: (and (featurep 'ox-extra) (ox-extras-activate '(ignore-headlines)))
# indent-tabs-mode: nil
# org-src-preserve-indentation: nil
# End: