|Previous: Icicles - Tripping||Icicles||IciclesIndex||Next: Icicles - Multi M-x|
What on earth for? Have you ever wanted to use one of those myriad
‘C-x’ key sequences, but forgot just what it was? The standard solution to that is to use
‘C-x C-h’, to display all of the
‘C-x’ bindings together with their commands.
OK, but then you have to scroll down the list of bindings, searching for the command you want, and then use its key binding. You can use
‘C-M-s’ to search for a substring of the command name, in case you do not recall the exact name, but why not use Icicles completion for this? Why not match against possible key sequences and commands?
To complete keys in Icicles, start the key sequence as usual, and then hit
‘S-TAB’. For example, use
‘C-x 4’, and then hit
‘S-TAB’ to complete the prefix
‘C-x 4’ (or whatever). You’re then completing against candidates that are composed of two parts, separated by “
So, for example, this is a single completion candidate:
C-j = dired-jump-other-window
You can match your minibuffer input against the key name, the command name, or both.
Suppose, for instance, that you want to use a version-control command, and you remember that all such commands are bound to key sequences that begin with
‘C-x v’. You enter as much of the key sequence as you remember (
‘C-x v’), and then you hit
‘S-TAB’. You can then use completion (either apropos or prefix) against the matching key sequences and command names to invoke the right command. And, as a bonus, you are reminded of its key sequence. You can thus use Icicles key completion to execute a command and, at the same time, learn its key binding.
In Icicle mode, whenever you are not in the minibuffer or buffer
‘S-tab’) initiates key completion. That is, you do not need to first type part of a key sequence to use it – you can start with it. Hit
‘S-TAB’ at any time, and you’re completing a key sequence, even if you have not yet hit any keys. This lets you see all key sequences that are available in a given context. For example, in Dired, keys special to that mode are included (and are highlighted as local bindings – see HighlightLocalBindings, below).
When completing a key sequence, you can type part of a command name, then hit
‘S-TAB’ again to apropos-complete against the command name. In this respect,
‘S-TAB’ acts like
‘M-x’, but the key binding is also part of the completion candidate, so you can also match key names.
So, just how do you complete input against a set of binding-plus-command completion candidates? You can always cycle among the candidates, of course, and then choose one. But what about completion? Just type text to match candidates, then use
‘TAB’ as usual to complete the text. Text? Yes. Completion candidates are always, ultimately, strings.
Suppose that you type
‘C-x S-TAB’ to show all key sequences that begin with
‘C-x’. You might see a candidate that looks like this:
C-q = toggle-read-only
You can then type “
C-q” or “
d-onl” or any other substring, and then use
‘S-TAB’ to complete the candidate. (This second use of
‘S-TAB’ invokes the command
‘icicle-apropos-complete’, which has nothing to do with
‘icicle-key-complete’, which was invoked by the first
‘S-TAB’. The first was invoked outside the minibuffer; the second was invoked from the minibuffer, during completion.)
What happens if the completion candidate is itself a prefix key? For example,
‘C-x S-TAB’ shows some candidates whose commands are shown as “
...”, like this:
4 = ... 5 = ... 6 = ... C-k = ... ESC = ... RET = ...
These represent prefix keys (
‘C-x C-k’, and so on). If you choose such a candidate, then you just continue completing – buffer
‘*Completions*’ is updated to show the completions of the compound prefix:
‘C-x RET’, or whichever you choose. The minibuffer prompt shows the completion so far; if you choose
‘RET’, for instance, then it shows
‘C-x RET’ while prompting you for the rest of the key sequence.
By default, completion candidates are sorted in buffer
‘*Completions*’ with local bindings listed first. You can use `
C-M-,’ at any time during key completion to toggle between this order and sorting with the prefix-key candidates shown first. You can use `
C-,’ at any time to change the sort order among these two orders and sorting by command name.
Here are some screenshots, from
‘S-TAB’ in the Buffer Menu (
‘C-x C-b’), to illustrate this.
Choosing a completion candidate such as `
C-x = ...’ effectively navigates down the key-binding hierachy (prefix-key hierarchy), to complete against all keys with prefix
‘C-x’. Choosing `
5 = ...’ to complete the prefix
‘C-x’ then navigates down another level, to complete keys that have prefix
What about navigating back up the hierarchy, say from the
‘C-x 5’ keys to the
‘C-x’ keys, or from the
‘C-x’ keys to the keys with no prefix? The special completion candidate `
..’ does that. By default, it is always the first candidate in the
‘*Completions*’ list. It is of course not available unless you are completing a prefix; that is, it is not available at the top level.
The screenshots above are all at the top level (in buffer
‘*Buffer List*’). The following screenshot is for the prefix key
‘ESC’ – it is what you see after choosing `
ESC = ...’ at the top level. Notice the special candidate `
..’, which is listed first and highlighted (using face `icicle-multi-command-completion’). (The fact that you see a candidate `
ESC = ...’ here too means that
‘ESC ESC’ is a prefix key at the top level.)
In fact, since menu-bar bindings are also key bindings, you can also use key completion to navigate the menu-bar hierarchy – just complete the prefix key
‘menu-bar’! Start with
‘S-TAB’, choose `
menu-bar = ...’, then choose a menu, and so on. But menu-bar menu completion is handy enough that it has its own Icicle mode key binding, by default:
Icicles key completion thus provides a general browser for key bindings, which you can also use to learn about keys and their associated comands, without necessarily executing them – see Key and Command Help, below.
‘S-TAB’uses apropos completion, by default, so remember that typing `
.’ matches any character (except a newline). To match only the literal string `
..’ (to go up a level), do one of the following:
C-`’ in the minibuffer to toggle this.
\.\.’ (or use `
Sometimes it helps to know which key sequences are local bindings, that is, bindings that are specific to the current mode. For example, DiredMode defines keys specific to Dired buffer, such as `
* %’, `
% g’, and `
To help you distinguish local key bindings from others (global and minor-mode bindings), local bindings are highlighted in buffer
‘*Completions*’ using face `icicle-special-candidate’. Otherwise, the key portion of a candidate (left side of `
KEY = COMMAND’) is highlighted with face `icicle-candidate-part’.
However, to make menu items stand out, they are highlighted with face `icicle-key-complete-menu-local’ for menus specific to the current mode, and face `icicle-key-complete-menu’ for other menu items. By default, these are the same as faces
‘icicle-candidate-part’, respectively, except that they are also boxed.
It may seem odd that you must complete a key sequence by entering the names of keys, rather than just hitting the keys themselves: e.g. typing “
C-f” rather than hitting
‘C-f’. However, if keys themselves were used for completing, then they could not be used normally during key-sequence completion. You could not move the cursor around the minibuffer using
‘right’ (right arrow), because those keys would be treated as input for completion. You could not use
‘down’ to cycle among completion candidates for the same reason. Likewise, you could not use printing (self-inserting) keys, such as
‘a’ and `
$’, to match command names. Having to use key names, instead of keys, for completion is a small price to pay for being able to complete key sequences.
Nevertheless, Icicles also provides a way for you to type key sequences directly, even if it is something of a workaround: precede each key with
‘icicle-insert-key-description’, during key completion) – think of
‘q’ for “quote”. This inserts the key description of whatever key you hit next. This key description (name) can be used to match key-completion candidates. So, for example, instead of typing “
C-f”, you can hit
‘M-q’ and then hit
‘C-f’. The key description “
C-f” is inserted in the minibuffer. If you use
‘M-q C-M-right’, then “
C-M-right” is inserted. Try it:
‘S-TAB M-q C-M-right’ → “
C-M-right”. Then hit
‘S-TAB’ to complete the candidate all the way to this:
C-M-right = enlarge-frame-horizontally
‘icicle-key-descriptions-use-angle-brackets-flag’). By default, this is
‘nil’, so angle brackets are not used, which I think improves readability. If you set this to non-
‘nil’, then you will see “
<C-M-right>” instead of “
C-M-right”, both as a completion candidate and as what is inserted when you use
‘M-q’. You can also provide a prefix argument to
‘M-q’to flip the behavior of `
icicle-key-descriptions-use-<>-flag’ for that occurrence only. See also my library naked.el, which lets you use the no-angle-brackets style also outside of Icicles.
That points out another use of key completion, opposite to learning the bindings of commands: learning the commands bound to given keys. In other words,
‘S-TAB M-q’ does both what
‘C-h w’ (
‘where-is’) does and what
‘C-h c’ (
‘describe-key-briefly’) does. It also does what
‘C-h b’ (
The point here is not that
‘S-TAB M-q’ is quicker than
‘C-h w’ or
‘C-h c’ or
‘C-h b’ – it’s not. The point is that key completion can be handy in several ways, and it can teach you various things about keys and commands as you use it.
In addition to this key-completion help about bindings, you can display help on the commands that are the right sides of the
‘S-TAB’ completion-candidate equations, by using the multi-command help keys (see Icicles - Help on Candidates). That is, while completing, you can use
‘C-M-next’, and so on to describe the command named in the current completion candidate.
‘icicle-complete-keys’ is a multi-command. This means that you can, within the same execution of
‘S-TAB’, invoke any number of keys by clicking (
‘C-mouse-2’) their names in buffer
‘*Completions*’ or choosing them any other way (
‘C-next’, and so on).
Since you can navigate up and down the key-binding hierarchy, you could even stay within a single
‘S-TAB’ invocation to do nearly everything you want in Emacs (see ThreeKeyEmacs, below)!
Keep in mind that
‘S-TAB’ has two different uses in Icicles when you are providing input in the minibuffer:
‘S-TAB’performs apropos completion (it is, in effect, bound to
‘S-TAB’performs key completion (it is, in effect, bound to
In addition, in buffer
‘S-TAB’ moves backward among the candidate completions.
This is by design; it takes advantage of the fact that these contexts are mutually exclusive. However, this economy comes at a risk of potential confusion. It’s important that you know whether or not completion is available when you are inputting text. If input completion is not available, but you think it is, then hitting
‘S-TAB’ might give you a surprise by key completing. That behavior is normal – you can use key-completion to input special characters, for instance. But if you think that you are instead completing the original input requested, then you can become confused.
Icicles provides completion status indicators so that you can easily tell when completion is available for minibuffer input. There are two indicators: (1) at the beginning of the minibuffer prompt and (2) in the
‘Icy’ minor-mode lighter in the mode line. See Completion Status Indicators. If completion is not indicated when you are prompted for input, it means that
‘S-TAB’ is available, not for input completion, but for key completion. Another clue can be found in the prompt text. For key completion, it says “
Complete keys: ”.
If you nevertheless find the overloaded use of
‘S-TAB’ confusing, you can change the bindings of the key
‘S-TAB’ in these different contexts. To do that, you can customize options `icicle-apropos-complete-keys’, `icicle-key-complete-keys’, and `icicle-completion-list-key-bindings’.
In the minibuffer,
‘S-TAB’ performs apropos completion of your typed input. So it is not available for key completion. But you can still complete keys in the minibuffer. You just use
‘M-S-TAB’ instead of
‘S-TAB’. (If your window manager steals
‘M-S-TAB’ then try
And just as the actual keys that perform key completion outside the minibuffer are customizable using option `icicle-key-complete-keys’, so too for the key-completion keys in the minibuffer: It is actually the keys defined by option `icicle-key-complete-keys-for-minibuffer’ that are used --
‘M-S-TAB’ is just the key used by default.
Icicles key completion piles a lot of stuff into
‘S-TAB’. Just as
‘M-x’ lets you execute any Emacs command, so does
‘S-TAB’ also lets you insert characters. You might say that it gives you all of Emacs in one key binding.
Of course, you need a couple other keys, as well. How many? Suppose you had limited accessibility in terms of input devices. Maybe you use Emacs on a cell phone, without voice recognition – or whatever. How many keys, buttons, or whatnot do you need to use Emacs?
‘C-g’, to interrupt commands.
(#2 and #3 might be combined somehow.)
What does vanilla Emacs offer out of the box in this regard?
‘mouse-1’and the menu bar menus, but they do not cover all of Emacs. You cannot use them to enter text, for instance. Of course, you could add more menus, to be able to do more.
‘RET’to execute any command. But how would you insert text?
Icicles key completion lets you do almost anything in Emacs with three or four keys, buttons, or whatever:
‘S-TAB’– Offers every key sequence as a possible choice to execute.
‘next’– Cycles among candidates, for choosing.
‘RET’– Chooses the current candidate.
‘C-g’, to cancel the current operation.
‘S-TAB’ includes key
‘M-x’, which offers all commands (even those not bound) as possible choices. It also includes key `M-:’, which lets you execute any EmacsLisp expression. That’s almost all of Emacs! (See also M-x via `mouse-2'.)
You could even perhaps get away with only three mouse buttons, and no keyboard:
‘mouse-1’– Choose candidates, scroll, and so on (direct access, no cycling).
‘mouse-2’– Do what
‘S-TAB’does (bind it to
‘mouse-3’– Do what
‘C-g’does (bind it to
‘mouse-3’ are not even used as mouse (pointer) functions; any keys or buttons would do. You could use just
‘mouse-1’ plus a Shift key and a Control key.
Would you want to use Emacs only this way? Of course not, if you had a choice. Typing the character
‘a’ by cycling through every possible key binding/command combination and hitting
‘RET’ when you get to `
a = self-insert-command’ would be the epitome of tedium. Likewise, doing everything with a single pointer-device button. Using only three or four keys or buttons is definitely not the ideal way to take advantage of Emacs.
But you are probably not limited to just 3 or 4 keys or buttons. The real point here is that Icicles
‘S-TAB’ opens the door to almost everything in Emacs. And if you do have a normal keyboard, then you can type letters and such to match command names and key sequences. Key
‘next’ matches substrings (regexps, actually), which makes choice even quicker.
‘self-insert-command’ is bound to each key that is associated with a character that can be inserted in text. It is the binding of the key
‘a’ and the key `
$’, for example. It is also the binding of keys that your keyboard might not even have – keys that correspond to special or odd characters and characters in other languages.
To Icicles key completion, these keys are like other keys. However, because there are many, many keys bound to
‘self-insert-command’, it can be distracting and slow to allow such keys as completion candidates. If option `icicle-complete-keys-self-insert-ranges’ is
‘nil’ (the default value), then such keys are excluded as candidates. This is probably what you want.
If the option is non-
‘nil’, then you can use key completion to insert the characters whose codes are in the range(s) defined by the option value. This lets you see the candidate characters in
‘*Completions*’ (WYSIWYG), but it is not a terribly convenient or quick way to insert characters.
Starting with Emacs 23, vanilla Emacs has Unicode support, and you can insert any Unicode characters using either an input method or
‘C-x 8 RET’.
‘C-x 8 RET’ is bound by default in Emacs to command
‘ucs-insert’ prior to Emacs 24). If you use my library ucs-cmds.el then you might want to remap that command to command
‘ucsc-insert’, which is an enhancement.
Icicles enhances this by (a) showing the character to be inserted after its name in
‘*Completions*’ (but you complete against only the name), and (b) showing the current character and its code point in the mode line when you cycle.
I recommend that you use
‘insert-char’ (or better,
‘ucsc-insert’), not key completion, to insert Unicode characters.
There are thousands of Unicode characters. So if you do use a non-
‘nil’ value for
‘icicle-complete-keys-self-insert-ranges’ then use only small ranges for better performance, e.g., `
((0 . 687))’ covers Latin characters. For Emacs 22, the option is effectively Boolean: any non-
‘nil’ value allows all self-inserting keys as candidates (there are far fewer available characters in Emacs 22).
For each Unicode character, the completion candidate is `
CHAR = UNICODE-NAME’, where
UNICODE-NAME is the name of the Unicode character. This is so that you can complete against the name. But again, I recommend that you use
‘C-x 8 RET’, not key completion, to insert a Unicode character.
‘S-TAB’ is not bound to
‘icicle-complete-keys’ in every keymap. That would be inconvenient, in general. By default, it is so bound in each keymap that is accessible from the global keymap, as determined by function
You’ve seen, above, how you can navigate through prefix keys, starting with the global map. In Dired, for instance, you can use
‘S-TAB’ at the top level, then choose the prefix key
‘*Completions*’, then choose a key, such as
‘/’ (to mark directories), in the
However, the act of binding of
‘S-TAB’ in keymaps that are accessible from the global map does not bind it in the
‘*’ prefix keymap itself. To handle this case, Icicles explicitly does for
‘dired-mode-map’ what it does for the global map: it binds
‘S-TAB’ in each keymap that is accessible from
‘dired-mode-map’. Because of this, you can use
‘* S-TAB’ to show all key completions of
This treatment of
‘dired-mode-map’ is done by default. Similarly for a few other keymaps. But you might have other keymaps that you would like to treat similarly – keymaps that Icicles might be unaware of. You do this by including them in the list value of user option `icicle-keymaps-for-key-completion’, along with
‘dired-mode-map’ and the others provided in the default value. The list entries are EmacsLisp symbols that are bound to keymaps, each of which should define at least one prefix key. If you add a keymap variable to this list, then
‘S-TAB’ will be bound so that you can use it to complete the prefix keys defined by that map.
Notice that there is no keymap variable that corresponds to prefix key
‘*’ in DiredMode. You need only provide a keymap (variable
‘dired-mode-map’) from which the prefix key is accessible; it is not necessary to also provide a variable that corresponds to the prefix keymap itself.
If a keymap listed in
‘icicle-keymaps-for-key-completion’ is not defined when Icicle mode is entered, then it is ignored. If you later define that keymap, then just exit and reenter Icicle mode for the
‘S-TAB’ binding to take effect. For example, use
‘M-x icy-mode’ twice after entering CalendarMode, to be able to complete
‘calendar-mode’ prefix keys such as
|Previous: Icicles - Tripping||Icicles||IciclesIndex||Next: Icicles - Multi M-x|