E. The Minibuffer

The minibuffer is the facility used by Emacs commands to read arguments more complicated than a single number. Minibuffer arguments can be file names, buffer names, Lisp function names, Emacs command names, Lisp expressions, and many other things, depending on the command reading the argument. You can use the usual Emacs editing commands in the minibuffer to edit the argument text.

When the minibuffer is in use, it appears in the echo area, and the terminal's cursor moves there. The beginning of the minibuffer line displays a prompt which says what kind of input you should supply and how it will be used. Often this prompt is derived from the name of the command that the argument is for. The prompt normally ends with a colon.

Sometimes a default argument appears in parentheses after the colon; it too is part of the prompt. The default will be used as the argument value if you enter an empty argument (that is, just type RET). For example, commands that read buffer names always show a default, which is the name of the buffer that will be used if you type just RET.

The simplest way to enter a minibuffer argument is to type the text you want, terminated by RET which exits the minibuffer. You can cancel the command that wants the argument, and get out of the minibuffer, by typing C-g.

Since the minibuffer uses the screen space of the echo area, it can conflict with other ways Emacs customarily uses the echo area. Here is how Emacs handles such conflicts:

• If a command gets an error while you are in the minibuffer, this does not cancel the minibuffer. However, the echo area is needed for the error message and therefore the minibuffer itself is hidden for a while. It comes back after a few seconds, or as soon as you type anything.

• If in the minibuffer you use a command whose purpose is to display a message in the echo area, such as C-x =, the message hides the minibuffer for a while. The minibuffer contents come back after a few seconds, or as soon as you type anything.

• Echoing of keystrokes does not take place while the minibuffer is in use.

E.1 Minibuffers for File Names

Sometimes the minibuffer starts out with text in it. For example, when you are supposed to give a file name, the minibuffer starts out containing the default directory, which ends with a slash. This is to inform you which directory the file will be found in if you do not specify a directory.

For example, the minibuffer might start out with these contents:

Find File: /u2/emacs/src/

where `Find File: ' is the prompt. Typing buffer.c specifies the file `/u2/emacs/src/buffer.c'. To find files in nearby directories, use ..; thus, if you type ../lisp/simple.el, you will get the file named `/u2/emacs/lisp/simple.el'. Alternatively, you can kill with M-DEL the directory names you don't want (see section T.1 Words).

If you don't want any of the default, you can kill it with C-a C-k. But you don't need to kill the default; you can simply ignore it. Insert an absolute file name, one starting with a slash or a tilde, after the default directory. For example, to specify the file `/etc/termcap', just insert that name, giving these minibuffer contents:

Find File: /u2/emacs/src//etc/termcap

GNU Emacs gives a special meaning to a double slash (which is not normally a useful thing to write): it means, "ignore everything before the second slash in the pair." Thus, `/u2/emacs/src/' is ignored in the example above, and you get the file `/etc/termcap'.

If you set insert-default-directory to nil, the default directory is not inserted in the minibuffer. This way, the minibuffer starts out empty. But the name you type, if relative, is still interpreted with respect to the same default directory.

E.2 Editing in the Minibuffer

The minibuffer is an Emacs buffer (albeit a peculiar one), and the usual Emacs commands are available for editing the text of an argument you are entering.

Since RET in the minibuffer is defined to exit the minibuffer, you can't use it to insert a newline in the minibuffer. To do that, type C-o or C-q C-j. (Recall that a newline is really the character control-J.)

The minibuffer has its own window which always has space on the screen but acts as if it were not there when the minibuffer is not in use. When the minibuffer is in use, its window is just like the others; you can switch to another window with C-x o, edit text in other windows and perhaps even visit more files, before returning to the minibuffer to submit the argument. You can kill text in another window, return to the minibuffer window, and then yank the text to use it in the argument. See section O. Multiple Windows.

There are some restrictions on the use of the minibuffer window, however. You cannot switch buffers in it--the minibuffer and its window are permanently attached. Also, you cannot split or kill the minibuffer window. But you can make it taller in the normal fashion with C-x ^.

The minibuffer window expands vertically as necessary to hold the text that you put in the minibuffer, if resize-mini-windows is non-nil. If resize-mini-windows is t, the window is always resized to fit the size of the text it displays. If resize-mini-windows is the symbol grow-only, the window grows when the size of displayed text increases, but shrinks (back to the normal size) only when the minibuffer becomes inactive.

The variable max-mini-window-height controls the maximum height for resizing the minibuffer window: a floating-point number specifies a fraction of the frame's height; an integer specifies the maximum number of lines; nil means do not resize the minibuffer window automatically. The default value is 0.25.

If while in the minibuffer you issue a command that displays help text of any sort in another window, you can use the C-M-v command while in the minibuffer to scroll the help text. This lasts until you exit the minibuffer. This feature is especially useful when you display a buffer listing possible completions. See section O.3 Using Other Windows.

Emacs normally disallows most commands that use the minibuffer while the minibuffer is active. This rule is to prevent recursive minibuffers from confusing novice users. If you want to be able to use such commands in the minibuffer, set the variable enable-recursive-minibuffers to a non-nil value.

E.3 Completion

For certain kinds of arguments, you can use completion to enter the argument value. Completion means that you type part of the argument, then Emacs visibly fills in the rest, or as much as can be determined from the part you have typed.

When completion is available, certain keys---TAB, RET, and SPC---are rebound to complete the text present in the minibuffer into a longer string that it stands for, by matching it against a set of completion alternatives provided by the command reading the argument. ? is defined to display a list of possible completions of what you have inserted.

For example, when M-x uses the minibuffer to read the name of a command, it provides a list of all available Emacs command names to complete against. The completion keys match the text in the minibuffer against all the command names, find any additional name characters implied by the ones already present in the minibuffer, and add those characters to the ones you have given. This is what makes it possible to type M-x ins SPC b RET instead of M-x insert-buffer RET (for example).

Case is normally significant in completion, because it is significant in most of the names that you can complete (buffer names, file names and command names). Thus, `fo' does not complete to `Foo'. Completion does ignore case distinctions for certain arguments in which case does not matter.

E.3.1 Completion Example

A concrete example may help here. If you type M-x au TAB, the TAB looks for alternatives (in this case, command names) that start with `au'. There are several, including auto-fill-mode and auto-save-mode---but they are all the same as far as auto-, so the `au' in the minibuffer changes to `auto-'.

If you type TAB again immediately, there are multiple possibilities for the very next character--it could be any of `cfilrs'---so no more characters are added; instead, TAB displays a list of all possible completions in another window.

If you go on to type f TAB, this TAB sees `auto-f'. The only command name starting this way is auto-fill-mode, so completion fills in the rest of that. You now have `auto-fill-mode' in the minibuffer after typing just au TAB f TAB. Note that TAB has this effect because in the minibuffer it is bound to the command minibuffer-complete when completion is available.

E.3.2 Completion Commands

Here is a list of the completion commands defined in the minibuffer when completion is available.

TAB

Complete the text in the minibuffer as much as possible (minibuffer-complete).

SPC

Complete the minibuffer text, but don't go beyond one word (minibuffer-complete-word).

RET

Submit the text in the minibuffer as the argument, possibly completing first as described below (minibuffer-complete-and-exit).

?

Display a list of all possible completions of the text in the minibuffer (minibuffer-list-completions).

SPC completes much like TAB, but never goes beyond the next hyphen or space. If you have `auto-f' in the minibuffer and type SPC, it finds that the completion is `auto-fill-mode', but it stops completing after `fill-'. This gives `auto-fill-'. Another SPC at this point completes all the way to `auto-fill-mode'. The command that implements this behavior is called minibuffer-complete-word.

Here are some commands you can use to choose a completion from a window that displays a list of completions:

Mouse-2

Clicking mouse button 2 on a completion in the list of possible completions chooses that completion (mouse-choose-completion). You normally use this command while point is in the minibuffer, but you must click in the list of completions, not in the minibuffer itself.

PRIOR

M-v

Typing PRIOR or PAGE-UP, or M-v, while in the minibuffer, selects the window showing the completion list buffer (switch-to-completions). This paves the way for using the commands below. (Selecting that window in the usual ways has the same effect, but this way is more convenient.)

RET

Typing RET in the completion list buffer chooses the completion that point is in or next to (choose-completion). To use this command, you must first switch windows to the window that shows the list of completions.

RIGHT

Typing the right-arrow key RIGHT in the completion list buffer moves point to the following completion (next-completion).

LEFT

Typing the left-arrow key LEFT in the completion list buffer moves point toward the beginning of the buffer, to the previous completion (previous-completion).

E.3.3 Strict Completion

There are three different ways that RET can work in completing minibuffers, depending on how the argument will be used.

• Strict completion is used when it is meaningless to give any argument except one of the known alternatives. For example, when C-x k reads the name of a buffer to kill, it is meaningless to give anything but the name of an existing buffer. In strict completion, RET refuses to exit if the text in the minibuffer does not complete to an exact match.

• Cautious completion is similar to strict completion, except that RET exits only if the text was an exact match already, not needing completion. If the text is not an exact match, RET does not exit, but it does complete the text. If it completes to an exact match, a second RET will exit.

  Cautious completion is used for reading file names for files that must already exist.

• Permissive completion is used when any string whatever is meaningful, and the list of completion alternatives is just a guide. For example, when C-x C-f reads the name of a file to visit, any file name is allowed, in case you want to create a file. In permissive completion, RET takes the text in the minibuffer exactly as given, without completing it.

The completion commands display a list of all possible completions in a window whenever there is more than one possibility for the very next character. Also, typing ? explicitly requests such a list. If the list of completions is long, you can scroll it with C-M-v (see section O.3 Using Other Windows).

E.3.4 Completion Options

When completion is done on file names, certain file names are usually ignored. The variable completion-ignored-extensions contains a list of strings; a file whose name ends in any of those strings is ignored as a possible completion. The standard value of this variable has several elements including ".o", ".elc", ".dvi" and "~". The effect is that, for example, `foo' can complete to `foo.c' even though `foo.o' exists as well. However, if all the possible completions end in "ignored" strings, then they are not ignored. Ignored extensions do not apply to lists of completions--those always mention all possible completions.

Normally, a completion command that cannot determine even one additional character automatically displays a list of all possible completions. If the variable completion-auto-help is set to nil, this automatic display is disabled, so you must type ? to display the list of completions.

Partial Completion mode implements a more powerful kind of completion that can complete multiple words in parallel. For example, it can complete the command name abbreviation p-b into print-buffer, because no other command starts with two words whose initials are `p' and `b'.

Partial completion of directories in file names uses `*' to indicate the places for completion; thus, `/u*/b*/f*' might complete to `/usr/bin/foo'.

To enable this mode, use the command M-x partial-completion-mode, or customize the option partial-completion-mode. This binds the partial completion commands to TAB, SPC, RET, and ?. The usual completion commands are available on M-TAB, M-SPC, M-RET and M-?.

Another feature of Partial Completion mode is to extend find-file so that the `<include>' stands for the file named include in some directory in the path PC-include-file-path. If you set PC-disable-includes to non-nil, this feature is disabled.

Icomplete mode presents a constantly-updated display that tells you what completions are available for the text you've entered so far. The command to enable or disable this minor mode is M-x icomplete-mode.

E.4 Minibuffer History

Every argument that you enter with the minibuffer is saved on a minibuffer history list so that you can use it again later in another argument. Special commands load the text of an earlier argument in the minibuffer. They discard the old minibuffer contents, so you can think of them as moving through the history of previous arguments.

UP

M-p

Move to the next earlier argument string saved in the minibuffer history (previous-history-element).

DOWN

M-n

Move to the next later argument string saved in the minibuffer history (next-history-element).

M-r regexp RET

Move to an earlier saved argument in the minibuffer history that has a match for regexp (previous-matching-history-element).

M-s regexp RET

Move to a later saved argument in the minibuffer history that has a match for regexp (next-matching-history-element).

The simplest way to reuse the saved arguments in the history list is to move through the history list one element at a time. While in the minibuffer, use M-p or up-arrow (previous-history-element) to "move to" the next earlier minibuffer input, and use M-n or down-arrow (next-history-element) to "move to" the next later input.

The previous input that you fetch from the history entirely replaces the contents of the minibuffer. To use it as the argument, exit the minibuffer as usual with RET. You can also edit the text before you reuse it; this does not change the history element that you "moved" to, but your new argument does go at the end of the history list in its own right.

For many minibuffer arguments there is a "default" value. In some cases, the minibuffer history commands know the default value. Then you can insert the default value into the minibuffer as text by using M-n to move "into the future" in the history. Eventually we hope to make this feature available whenever the minibuffer has a default value.

There are also commands to search forward or backward through the history; they search for history elements that match a regular expression that you specify with the minibuffer. M-r (previous-matching-history-element) searches older elements in the history, while M-s (next-matching-history-element) searches newer elements. By special dispensation, these commands can use the minibuffer to read their arguments even though you are already in the minibuffer when you issue them. As with incremental searching, an upper-case letter in the regular expression makes the search case-sensitive (see section K.6 Searching and Case).

All uses of the minibuffer record your input on a history list, but there are separate history lists for different kinds of arguments. For example, there is a list for file names, used by all the commands that read file names. (As a special feature, this history list records the absolute file name, no more and no less, even if that is not how you entered the file name.)

There are several other very specific history lists, including one for command names read by M-x, one for buffer names, one for arguments of commands like query-replace, and one for compilation commands read by compile. Finally, there is one "miscellaneous" history list that most minibuffer arguments use.

The variable history-length specifies the maximum length of a minibuffer history list; once a list gets that long, the oldest element is deleted each time an element is added. If the value of history-length is t, though, there is no maximum length and elements are never deleted.

E.5 Repeating Minibuffer Commands

Every command that uses the minibuffer at least once is recorded on a special history list, together with the values of its arguments, so that you can repeat the entire command. In particular, every use of M-x is recorded there, since M-x uses the minibuffer to read the command name.

C-x ESC ESC

Re-execute a recent minibuffer command (repeat-complex-command).

M-x list-command-history

Display the entire command history, showing all the commands C-x ESC ESC can repeat, most recent first.

C-x ESC ESC is used to re-execute a recent minibuffer-using command. With no argument, it repeats the last such command. A numeric argument specifies which command to repeat; one means the last one, and larger numbers specify earlier ones.

C-x ESC ESC works by turning the previous command into a Lisp expression and then entering a minibuffer initialized with the text for that expression. If you type just RET, the command is repeated as before. You can also change the command by editing the Lisp expression. Whatever expression you finally submit is what will be executed. The repeated command is added to the front of the command history unless it is identical to the most recently executed command already there.

Even if you don't understand Lisp syntax, it will probably be obvious which command is displayed for repetition. If you do not change the text, it will repeat exactly as before.

Once inside the minibuffer for C-x ESC ESC, you can use the minibuffer history commands (M-p, M-n, M-r, M-s; see section E.4 Minibuffer History) to move through the history list of saved entire commands. After finding the desired previous command, you can edit its expression as usual and then resubmit it by typing RET as usual.

The list of previous minibuffer-using commands is stored as a Lisp list in the variable command-history. Each element is a Lisp expression which describes one command and its arguments. Lisp programs can re-execute a command by calling eval with the command-history element.