Go to the previous, next section.

Commands for Human Languages

The term text has two widespread meanings in our area of the computer field. One is data that is a sequence of characters. Any file that you edit with Emacs is text, in this sense of the word. The other meaning is more restrictive: a sequence of characters in a human language for humans to read (possibly after processing by a text formatter), as opposed to a program or commands for a program.

Human languages have syntactic/stylistic conventions that can be supported or used to advantage by editor commands: conventions involving words, sentences, paragraphs, and capital letters. This chapter describes Emacs commands for all of these things. There are also commands for filling, which means rearranging the lines of a paragraph to be approximately equal in length. The commands for moving over and killing words, sentences and paragraphs, while intended primarily for editing text, are also often useful for editing programs.

Emacs has several major modes for editing human language text. If the file contains text pure and simple, use Text mode, which customizes Emacs in small ways for the syntactic conventions of text. For text which contains embedded commands for text formatters, Emacs has other major modes, each for a particular text formatter. Thus, for input to TeX, you would use TeX mode; for input to nroff, Nroff mode. Outline mode provides special commands for operating on text with an outline structure.

Words

Emacs has commands for moving over or operating on words. By convention, the keys for them are all Meta characters.

M-f
Move forward over a word (forward-word).
M-b
Move backward over a word (backward-word).
M-d
Kill up to the end of a word (kill-word).
M-DEL
Kill back to the beginning of a word (backward-kill-word).
M-@
Mark the end of the next word (mark-word).
M-t
Transpose two words or drag a word across other words (transpose-words).

Notice how these keys form a series that parallels the character-based C-f, C-b, C-d, C-t and DEL. M-@ is cognate to C-@, which is an alias for C-SPC.

The commands M-f (forward-word) and M-b (backward-word) move forward and backward over words. These Meta characters are thus analogous to the corresponding control characters, C-f and C-b, which move over single characters in the text. The analogy extends to numeric arguments, which serve as repeat counts. M-f with a negative argument moves backward, and M-b with a negative argument moves forward. Forward motion stops right after the last letter of the word, while backward motion stops right before the first letter.

M-d (kill-word) kills the word after point. To be precise, it kills everything from point to the place M-f would move to. Thus, if point is in the middle of a word, M-d kills just the part after point. If some punctuation comes between point and the next word, it is killed along with the word. (If you wish to kill only the next word but not the punctuation before it, simply do M-f to get the end, and kill the word backwards with M-DEL.) M-d takes arguments just like M-f.

M-DEL (backward-kill-word) kills the word before point. It kills everything from point back to where M-b would move to. If point is after the space in `FOO, BAR', then `FOO, ' is killed. (If you wish to kill just `FOO', do M-b M-d instead of M-DEL.)

M-t (transpose-words) exchanges the word before or containing point with the following word. The delimiter characters between the words do not move. For example, `FOO, BAR' transposes into `BAR, FOO' rather than `BAR FOO,'. See section Transposing Text, for more on transposition and on arguments to transposition commands.

To operate on the next n words with an operation which applies between point and mark, you can either set the mark at point and then move over the words, or you can use the command M-@ (mark-word) which does not move point, but sets the mark where M-f would move to. M-@ accepts a numeric argument that says how many words to scan for the place to put the mark. In Transient Mark mode, this command activates the mark.

The word commands' understanding of syntax is completely controlled by the syntax table. Any character can, for example, be declared to be a word delimiter. See section The Syntax Table.

Sentences

The Emacs commands for manipulating sentences and paragraphs are mostly on Meta keys, so as to be like the word-handling commands.

M-a
Move back to the beginning of the sentence (backward-sentence).
M-e
Move forward to the end of the sentence (forward-sentence).
M-k
Kill forward to the end of the sentence (kill-sentence).
C-x DEL
Kill back to the beginning of the sentence (backward-kill-sentence).

The commands M-a and M-e (backward-sentence and forward-sentence) move to the beginning and end of the current sentence, respectively. They were chosen to resemble C-a and C-e, which move to the beginning and end of a line. Unlike them, M-a and M-e if repeated or given numeric arguments move over successive sentences.

Moving backward over a sentence places point just before the first character of the sentence; moving forward places point right after the punctuation that ends the sentence. Neither one moves over the whitespace at the sentence boundary.

Just as C-a and C-e have a kill command, C-k, to go with them, so M-a and M-e have a corresponding kill command M-k (kill-sentence) which kills from point to the end of the sentence. With minus one as an argument it kills back to the beginning of the sentence. Larger arguments serve as a repeat count. There is also a command, C-x DEL (backward-kill-sentence), for killing back to the beginning of a sentence. This command is useful when you change your mind in the middle of composing text.

The sentence commands assume that you follow the American typist's convention of putting two spaces at the end of a sentence; they consider a sentence to end wherever there is a `.', `?' or `!' followed by the end of a line or two spaces, with any number of `)', `]', `'', or `"' characters allowed in between. A sentence also begins or ends wherever a paragraph begins or ends.

The variable sentence-end controls recognition of the end of a sentence. It is a regexp that matches the last few characters of a sentence, together with the whitespace following the sentence. Its normal value is

"[.?!][]\"')]*\\($\\|\t\\|  \\)[ \t\n]*"

This example is explained in the section on regexps. See section Syntax of Regular Expressions.

If you want to use just one space between sentences, you should set sentence-end to this value:

"[.?!][]\"')]*\\($\\|\t\\| \\)[ \t\n]*"

You should also set the variable sentence-end-double-space to nil so that the fill commands expect and leave just one space at the end of a sentence. Note that this makes it impossible to distinguish between periods that end sentences and those that indicate abbreviations.

Paragraphs

The Emacs commands for manipulating paragraphs are also Meta keys.

M-{
Move back to previous paragraph beginning (backward-paragraph).
M-}
Move forward to next paragraph end (forward-paragraph).
M-h
Put point and mark around this or next paragraph (mark-paragraph).

M-{ moves to the beginning of the current or previous paragraph, while M-} moves to the end of the current or next paragraph. Blank lines and text formatter command lines separate paragraphs and are not considered part of any paragraph. Also, an indented line starts a new paragraph.

In major modes for programs (as opposed to Text mode), paragraphs begin and end only at blank lines. This makes the paragraph commands continue to be useful even though there are no paragraphs per se.

When there is a fill prefix, then paragraphs are delimited by all lines which don't start with the fill prefix. See section Filling Text.

When you wish to operate on a paragraph, you can use the command M-h (mark-paragraph) to set the region around it. Thus, for example, M-h C-w kills the paragraph around or after point. The M-h command puts point at the beginning and mark at the end of the paragraph point was in. In Transient Mark mode, it activates the mark. If point is between paragraphs (in a run of blank lines, or at a boundary), the paragraph following point is surrounded by point and mark. If there are blank lines preceding the first line of the paragraph, one of these blank lines is included in the region.

The precise definition of a paragraph boundary is controlled by the variables paragraph-separate and paragraph-start. The value of paragraph-start is a regexp that should match any line that either starts or separates paragraphs. The value of paragraph-separate is another regexp that should match only lines that separate paragraphs without being part of any paragraph. Lines that start a new paragraph and are contained in it must match only paragraph-start, not paragraph-separate. For example, normally paragraph-start is "^[ \t\n\f]" and paragraph-separate is "^[ \t\f]*$".

Normally it is desirable for page boundaries to separate paragraphs. The default values of these variables recognize the usual separator for pages.

Pages

Files are often thought of as divided into pages by the formfeed character (ASCII control-L, octal code 014). When you print hardcopy for a file, this character forces a page break; thus, each page of the file goes on a separate page on paper. Most Emacs commands treat the page-separator character just like any other character: you can insert it with C-q C-l, and delete it with DEL. Thus, you are free to paginate your file or not. However, since pages are often meaningful divisions of the file, Emacs provides commands to move over them and operate on them.

C-x [
Move point to previous page boundary (backward-page).
C-x ]
Move point to next page boundary (forward-page).
C-x C-p
Put point and mark around this page (or another page) (mark-page).
C-x l
Count the lines in this page (count-lines-page).

The C-x [ (backward-page) command moves point to immediately after the previous page delimiter. If point is already right after a page delimiter, it skips that one and stops at the previous one. A numeric argument serves as a repeat count. The C-x ] (forward-page) command moves forward past the next page delimiter.

The C-x C-p command (mark-page) puts point at the beginning of the current page and the mark at the end. The page delimiter at the end is included (the mark follows it). The page delimiter at the front is excluded (point follows it). C-x C-p C-w is a handy way to kill a page to move it elsewhere. If you move to another page delimiter with C-x [ and C-x ], then yank the killed page, all the pages will be properly delimited once again. The reason C-x C-p includes only the following page delimiter in the region is to ensure that.

A numeric argument to C-x C-p is used to specify which page to go to, relative to the current one. Zero means the current page. One means the next page, and -1 means the previous one.

The C-x l command (count-lines-page) is good for deciding where to break a page in two. It prints in the echo area the total number of lines in the current page, and then divides it up into those preceding the current line and those following, as in

Page has 96 (72+25) lines

Notice that the sum is off by one; this is correct if point is not at the beginning of a line.

The variable page-delimiter controls where pages begin. Its value is a regexp that matches the beginning of a line that separates pages. The normal value of this variable is "^\f", which matches a formfeed character at the beginning of a line.

Filling Text

Filling text means breaking it up into lines that fit a specified width. Emacs does filling in two ways. In Auto Fill mode, inserting text with self-inserting characters also automatically fills it. There are also explicit fill commands that you can use when editing text leaves it unfilled.

Auto Fill Mode

Auto Fill mode is a minor mode in which lines are broken automatically when they become too wide. Breaking happens only when you type a SPC or RET.

M-x auto-fill-mode
Enable or disable Auto Fill mode.
SPC
RET
In Auto Fill mode, break lines when appropriate.

M-x auto-fill-mode turns Auto Fill mode on if it was off, or off if it was on. With a positive numeric argument it always turns Auto Fill mode on, and with a negative argument always turns it off. You can see when Auto Fill mode is in effect by the presence of the word `Fill' in the mode line, inside the parentheses. Auto Fill mode is a minor mode which is enabled or disabled for each buffer individually. See section Minor Modes.

In Auto Fill mode, lines are broken automatically at spaces when they get longer than the desired width. Line breaking and rearrangement takes place only when you type SPC or RET. If you wish to insert a space or newline without permitting line-breaking, type C-q SPC or C-q LFD (recall that a newline is really a linefeed). Also, C-o inserts a newline without line breaking.

Auto Fill mode works well with Lisp mode, because when it makes a new line in Lisp mode it indents that line with TAB. If a line ending in a comment gets too long, the text of the comment is split into two comment lines. Optionally new comment delimiters are inserted at the end of the first line and the beginning of the second so that each line is a separate comment; the variable comment-multi-line controls the choice (see section Manipulating Comments).

Auto Fill mode does not refill entire paragraphs; it can break lines but cannot merge lines. So editing in the middle of a paragraph can result in a paragraph that is not correctly filled. The easiest way to make the paragraph properly filled again is usually with the explicit fill commands.

Many users like Auto Fill mode and want to use it in all text files. The section on init files says how to arrange this permanently for yourself. See section The Init File, `~/.emacs'.

Explicit Fill Commands

M-q
Fill current paragraph (fill-paragraph).
C-x f
Set the fill column (set-fill-column).
M-x fill-region
Fill each paragraph in the region (fill-region).
M-x fill-region-as-paragraph.
Fill the region, considering it as one paragraph.
M-s
Center a line.

To refill a paragraph, use the command M-q (fill-paragraph). This operates on the paragraph that point is inside, or the one after point if point is between paragraphs. Refilling works by removing all the line-breaks, then inserting new ones where necessary.

To refill many paragraphs, use M-x fill-region, which divides the region into paragraphs and fills each of them.

M-q and fill-region use the same criteria as M-h for finding paragraph boundaries (see section Paragraphs). For more control, you can use M-x fill-region-as-paragraph, which refills everything between point and mark. This command deletes any blank lines within the region, so separate blocks of text end up combined into one block.

A numeric argument to M-q causes it to justify the text as well as filling it. This means that extra spaces are inserted to make the right margin line up exactly at the fill column. To remove the extra spaces, use M-q with no argument. (Likewise for fill-region.)

When adaptive-fill-mode is non-nil (which is normally the case), if you use fill-region-as-paragraph on an indented paragraph and you don't have a fill prefix, it uses the indentation of the second line of the paragraph as the fill prefix. The effect of adaptive filling is not noticeable in Text mode, because an indented line counts as a paragraph starter and thus each line of an indented paragraph is considered a paragraph of its own. But you do notice the effect in Indented Text mode and some other major modes.

The command M-s (center-line) centers the current line within the current fill column. With an argument n, it centers n lines individually and moves past them.

The maximum line width for filling is in the variable fill-column. Altering the value of fill-column makes it local to the current buffer; until that time, the default value is in effect. The default is initially 70. See section Local Variables. The easiest way to set fill-column is to use the command C-x f (set-fill-column). With no argument, it sets fill-column to the current horizontal position of point. With a numeric argument, it uses that as the new fill column.

Emacs commands normally consider a period followed by two spaces or by a newline as the end of a sentence; a period followed by just one space indicates an abbreviation and not the end of a sentence. To preserve the distinction between these two ways of using a period, the fill commands do not break a line after a period followed by just one space.

If the variable sentence-end-double-space is nil, the fill commands expect and leave just one space at the end of a sentence. Ordinarily this variable is t, so the fill commands insist on two spaces for the end of a sentence, as explained above. See section Sentences.

The Fill Prefix

To fill a paragraph in which each line starts with a special marker (which might be a few spaces, giving an indented paragraph), use the fill prefix feature. The fill prefix is a string which Emacs expects every line to start with, and which is not included in filling.

C-x .
Set the fill prefix (set-fill-prefix).
M-q
Fill a paragraph using current fill prefix (fill-paragraph).
M-x fill-individual-paragraphs
Fill the region, considering each change of indentation as starting a new paragraph.
M-x fill-nonuniform-paragraphs
Fill the region, considering only paragraph-separator lines as starting a new paragraph.

To specify a fill prefix, move to a line that starts with the desired prefix, put point at the end of the prefix, and give the command C-x . (set-fill-prefix). That's a period after the C-x. To turn off the fill prefix, specify an empty prefix: type C-x . with point at the beginning of a line.

When a fill prefix is in effect, the fill commands remove the fill prefix from each line before filling and insert it on each line after filling. Auto Fill mode also inserts the fill prefix automatically when it makes a new line. The C-o command inserts the fill prefix on new lines it creates, when you use it at the beginning of a line (see section Blank Lines). Conversely, the command M-^ deletes the prefix (if it occurs) after the newline that it deletes (see section Indentation).

For example, if fill-column is 40 and you set the fill prefix to `;; ', then M-q in the following text

;; This is an
;; example of a paragraph
;; inside a Lisp-style comment.

produces this:

;; This is an example of a paragraph
;; inside a Lisp-style comment.

Lines that do not start with the fill prefix are considered to start paragraphs, both in M-q and the paragraph commands; this is gives good results for paragraphs with hanging indentation (every line indented except the first one). Lines which are blank or indented once the prefix is removed also separate or start paragraphs; this is what you want if you are writing multi-paragraph comments with a comment delimiter on each line.

You can use M-x fill-individual-paragraphs to set the fill prefix for each paragraph automatically. This command divides the region into paragraphs, treating every change in the amount of indentation as the start of a new paragraph, and fills each of these paragraphs. Thus, all the lines in one "paragraph" have the same amount of indentation. That indentation serves as the fill prefix for that paragraph.

M-x fill-nonuniform-paragraphs is a similar command that divides the region into paragraphs in a different way. It considers only paragraph-separating lines (as defined by paragraph-separate) as starting a new paragraph. Since this means that the lines of one paragraph may have different amounts of indentation, the fill prefix used is the smallest amount of indentation of any of the lines of the paragraph. This gives good results with styles that indent a paragraph's first line more or less that the rest of the paragraph.

The fill prefix is stored in the variable fill-prefix. Its value is a string, or nil when there is no fill prefix. This is a per-buffer variable; altering the variable affects only the current buffer, but there is a default value which you can change as well. See section Local Variables.

Case Conversion Commands

Emacs has commands for converting either a single word or any arbitrary range of text to upper case or to lower case.

M-l
Convert following word to lower case (downcase-word).
M-u
Convert following word to upper case (upcase-word).
M-c
Capitalize the following word (capitalize-word).
C-x C-l
Convert region to lower case (downcase-region).
C-x C-u
Convert region to upper case (upcase-region).

The word conversion commands are the most useful. M-l (downcase-word) converts the word after point to lower case, moving past it. Thus, repeating M-l converts successive words. M-u (upcase-word) converts to all capitals instead, while M-c (capitalize-word) puts the first letter of the word into upper case and the rest into lower case. All these commands convert several words at once if given an argument. They are especially convenient for converting a large amount of text from all upper case to mixed case, because you can move through the text using M-l, M-u or M-c on each word as appropriate, occasionally using M-f instead to skip a word.

When given a negative argument, the word case conversion commands apply to the appropriate number of words before point, but do not move point. This is convenient when you have just typed a word in the wrong case: you can give the case conversion command and continue typing.

If a word case conversion command is given in the middle of a word, it applies only to the part of the word which follows point. This is just like what M-d (kill-word) does. With a negative argument, case conversion applies only to the part of the word before point.

The other case conversion commands are C-x C-u (upcase-region) and C-x C-l (downcase-region), which convert everything between point and mark to the specified case. Point and mark do not move.

The region case conversion commands upcase-region and downcase-region are normally disabled. This means that they ask for confirmation if you try to use them. When you confirm, you may enable the command, which means it will not ask for confirmation again. See section Disabling Commands.

Text Mode

When you edit files of text in a human language, it's more convenient to use Text mode rather than Fundamental mode. Invoke M-x text-mode to enter Text mode. In Text mode, TAB runs the function tab-to-tab-stop, which allows you to use arbitrary tab stops set with M-x edit-tab-stops (see section Tab Stops). Features concerned with comments in programs are turned off in Text mode except when explicitly invoked. The syntax table is changed so that periods are not considered part of a word, while apostrophes, backspaces and underlines are part of words.

A similar variant mode is Indented Text mode, intended for editing text in which most lines are indented. This mode defines TAB to run indent-relative (see section Indentation), and makes Auto Fill indent the lines it creates. The result is that normally a line made by Auto Filling, or by LFD, is indented just like the previous line. In Indented Text mode, only blank lines separate paragraphs--indented lines continue the current paragraph. Use M-x indented-text-mode to select this mode.

Text mode, and all the modes based on it, define M-TAB as the command ispell-complete-word, which performs completion of the partial word in the buffer before point, using the spelling dictionary as the space of possible words. See section Checking and Correcting Spelling.

Entering Text mode or Indented Text mode runs the hook text-mode-hook. Other major modes related to Text mode also run this hook, followed by hooks of their own; this includes Nroff mode, TeX mode, Outline mode and Mail mode. Hook functions on text-mode-hook can look at the value of major-mode to see which of these modes is actually being entered. See section Hooks.

Outline Mode

Outline mode is a major mode much like Text mode but intended for editing outlines. It allows you to make parts of the text temporarily invisible so that you can see the outline structure. Type M-x outline-mode to switch to Outline mode as the major mode of the current buffer. Type M-x outline-minor-mode to enable Outline mode as a minor mode in the current buffer. Outline minor mode provides the same commands as the major mode, Outline mode, but you can use it in conjunction with other major modes.

The major mode, Outline mode, provides special key bindings on the C-c prefix. Outline minor mode provides similar bindings with C-c C-o as the prefix; this is to reduce the conflicts with the major mode's special commands. (The variable outline-minor-mode-prefix controls the prefix used.)

When Outline mode makes a line invisible, the line does not appear on the screen. The screen appears exactly as if the invisible line were deleted, except that an ellipsis (three periods in a row) appears at the end of the previous visible line (only one ellipsis no matter how many invisible lines follow).

All editing commands treat the text of the invisible line as part of the previous visible line. For example, C-n moves onto the next visible line. Killing an entire visible line, including its terminating newline, really kills all the following invisible lines along with it; yanking it all back yanks the invisible lines and they remain invisible.

Entering Outline mode runs the hook text-mode-hook followed by the hook outline-mode-hook (see section Hooks).

Format of Outlines

Outline mode assumes that the lines in the buffer are of two types: heading lines and body lines. A heading line represents a topic in the outline. Heading lines start with one or more stars; the number of stars determines the depth of the heading in the outline structure. Thus, a heading line with one star is a major topic; all the heading lines with two stars between it and the next one-star heading are its subtopics; and so on. Any line that is not a heading line is a body line. Body lines belong with the preceding heading line. Here is an example:

* Food

This is the body,
which says something about the topic of food.

** Delicious Food

This is the body of the second-level header.

** Distasteful Food

This could have
a body too, with
several lines.

*** Dormitory Food

* Shelter

Another first-level topic with its header line.

A heading line together with all following body lines is called collectively an entry. A heading line together with all following deeper heading lines and their body lines is called a subtree.

You can customize the criterion for distinguishing heading lines by setting the variable outline-regexp. Any line whose beginning has a match for this regexp is considered a heading line. Matches that start within a line (not at the left margin) do not count. The length of the matching text determines the level of the heading; longer matches make a more deeply nested level. Thus, for example, if a text formatter has commands `@chapter', `@section' and `@subsection' to divide the document into chapters and sections, you could make those lines count as heading lines by setting outline-regexp to `"@chap\\|@\\(sub\\)*section"'. Note the trick: the two words `chapter' and `section' are equally long, but by defining the regexp to match only `chap' we ensure that the length of the text matched on a chapter heading is shorter, so that Outline mode will know that sections are contained in chapters. This works as long as no other command starts with `@chap'.

It is possible to change the rule for calculating the level of a heading line by setting the variable outline-level. The value of outline-level should be a function that takes no arguments and returns the level of the current heading. Some major modes such as C, Nroff, and Emacs Lisp mode set this variable in order to work with Outline minor mode.

Outline mode makes a line invisible by changing the newline before it into an ASCII control-M (code 015). Most editing commands that work on lines treat an invisible line as part of the previous line because, strictly speaking, it is part of that line, since there is no longer a newline in between. When you save the file in Outline mode, control-M characters are saved as newlines, so the invisible lines become ordinary lines in the file. But saving does not change the visibility status of a line inside Emacs.

Outline Motion Commands

There are some special motion commands in Outline mode that move backward and forward to heading lines.

C-c C-n
Move point to the next visible heading line (outline-next-visible-heading).
C-c C-p
Move point to the previous visible heading line (outline-previous-visible-heading).
C-c C-f
Move point to the next visible heading line at the same level as the one point is on (outline-forward-same-level).
C-c C-b
Move point to the previous visible heading line at the same level (outline-backward-same-level).
C-c C-u
Move point up to a lower-level (more inclusive) visible heading line (outline-up-heading).

C-c C-n (next-visible-heading) moves down to the next heading line. C-c C-p (previous-visible-heading) moves similarly backward. Both accept numeric arguments as repeat counts. The names emphasize that invisible headings are skipped, but this is not really a special feature. All editing commands that look for lines ignore the invisible lines automatically.

More powerful motion commands understand the level structure of headings. C-c C-f (outline-forward-same-level) and C-c C-b (outline-backward-same-level) move from one heading line to another visible heading at the same depth in the outline. C-c C-u (outline-up-heading) moves backward to another heading that is less deeply nested.

Outline Visibility Commands

The other special commands of outline mode are used to make lines visible or invisible. Their names all start with hide or show. Most of them fall into pairs of opposites. They are not undoable; instead, you can undo right past them. Making lines visible or invisible is simply not recorded by the undo mechanism.

C-c C-t
Make all body lines in the buffer invisible (hide-body).
C-c C-a
Make all lines in the buffer visible (show-all).
C-c C-d
Make everything under this heading invisible, not including this heading itself
(hide-subtree).
C-c C-s
Make everything under this heading visible, including body, subheadings, and their bodies (show-subtree).
C-c C-l
Make the body of this heading line, and of all its subheadings, invisible (hide-leaves).
C-c C-k
Make all subheadings of this heading line, at all levels, visible (show-branches).
C-c C-i
Make immediate subheadings (one level down) of this heading line visible (show-children).
C-c C-c
Make this heading line's body invisible (hide-entry).
C-c C-e
Make this heading line's body visible (show-entry).
C-c C-q
Hide everything except the top n levels of heading lines (hide-sublevels).
C-c C-o
Hide everything except for the heading or body that point is in, plus the headings leading up from there to the top level of the outline (hide-other).

Two commands that are exact opposites are C-c C-c (hide-entry) and C-c C-e (show-entry). They are used with point on a heading line, and apply only to the body lines of that heading. Subheadings and their bodies are not affected.

Two more powerful opposites are C-c C-d (hide-subtree) and C-c C-s (show-subtree). Both expect to be used when point is on a heading line, and both apply to all the lines of that heading's subtree: its body, all its subheadings, both direct and indirect, and all of their bodies. In other words, the subtree contains everything following this heading line, up to and not including the next heading of the same or higher rank.

Intermediate between a visible subtree and an invisible one is having all the subheadings visible but none of the body. There are two commands for doing this, depending on whether you want to hide the bodies or make the subheadings visible. They are C-c C-l (hide-leaves) and C-c C-k (show-branches).

A little weaker than show-branches is C-c C-i (show-children). It makes just the direct subheadings visible--those one level down. Deeper subheadings remain invisible, if they were invisible.

Two commands have a blanket effect on the whole file. C-c C-t (hide-body) makes all body lines invisible, so that you see just the outline structure. C-c C-a (show-all) makes all lines visible. These commands can be thought of as a pair of opposites even though C-c C-a applies to more than just body lines.

The command C-c C-q (hide-sublevels) hides all but the top level headings. With a numeric argument n, it hides everything except the top n levels of heading lines.

The command C-c C-o (hide-other) hides everything except the heading or body text that point is in, plus its parents (the headers leading up from there to top level in the outline).

You can turn off the use of ellipses at the ends of visible lines by setting selective-display-ellipses to nil. Then there is no visible indication of the presence of invisible lines.

TeX Mode

TeX is a powerful text formatter written by Donald Knuth; it is also free, like GNU Emacs. LaTeX is a simplified input format for TeX, implemented by TeX macros; it comes with TeX. SliTeX is a special form of LaTeX.

Emacs has a special TeX mode for editing TeX input files. It provides facilities for checking the balance of delimiters and for invoking TeX on all or part of the file.

TeX mode has three variants, Plain TeX mode, LaTeX mode, and SliTeX mode (these three distinct major modes differ only slightly). They are designed for editing the three different formats. The command M-x tex-mode looks at the contents of the buffer to determine whether the contents appear to be either LaTeX input or SliTeX input; if so, it selects the appropriate mode. If the file contents do not appear to be LaTeX or SliTeX, it selects Plain TeX mode. If the contents are insufficient to determine this, the variable tex-default-mode controls which mode is used.

When M-x tex-mode does not guess right, you can use the commands M-x plain-tex-mode, M-x latex-mode, and M-x slitex-mode to select explicitly the particular variants of TeX mode.

TeX Editing Commands

Here are the special commands provided in TeX mode for editing the text of the file.

"
Insert, according to context, either `"' or `"' or `"' (tex-insert-quote).
LFD
Insert a paragraph break (two newlines) and check the previous paragraph for unbalanced braces or dollar signs (tex-terminate-paragraph).
M-x validate-tex-region
Check each paragraph in the region for unbalanced braces or dollar signs.
C-c {
Insert `{}' and position point between them (tex-insert-braces).
C-c }
Move forward past the next unmatched close brace (up-list).

In TeX, the character `"' is not normally used; we use `"' to start a quotation and `"' to end one. To make editing easier under this formatting convention, TeX mode overrides the normal meaning of the key " with a command that inserts a pair of single-quotes or backquotes (tex-insert-quote). To be precise, this command inserts `"' after whitespace or an open brace, `"' after a backslash, and `"' after any other character.

If you need the character `"' itself in unusual contexts, use C-q to insert it. Also, " with a numeric argument always inserts that number of `"' characters.

In TeX mode, `$' has a special syntax code which attempts to understand the way TeX math mode delimiters match. When you insert a `$' that is meant to exit math mode, the position of the matching `$' that entered math mode is displayed for a second. This is the same feature that displays the open brace that matches a close brace that is inserted. However, there is no way to tell whether a `$' enters math mode or leaves it; so when you insert a `$' that enters math mode, the previous `$' position is shown as if it were a match, even though they are actually unrelated.

TeX uses braces as delimiters that must match. Some users prefer to keep braces balanced at all times, rather than inserting them singly. Use C-c { (tex-insert-braces) to insert a pair of braces. It leaves point between the two braces so you can insert the text that belongs inside. Afterward, use the command C-c } (up-list) to move forward past the close brace.

There are two commands for checking the matching of braces. LFD (tex-terminate-paragraph) checks the paragraph before point, and inserts two newlines to start a new paragraph. It prints a message in the echo area if any mismatch is found. M-x validate-tex-region checks a region, paragraph by paragraph. When it finds a paragraph that contains a mismatch, it displays point at the beginning of the paragraph for a few seconds and sets the mark at that spot. Scanning continues until the whole buffer has been checked or until you type another key. Afterward, you can use the mark ring to find the last several paragraphs that had mismatches (see section The Mark Ring).

Note that Emacs commands count square brackets and parentheses in TeX mode, not just braces. This is not strictly correct for the purpose of checking TeX syntax. However, parentheses and square brackets are likely to be used in text as matching delimiters and it is useful for the various motion commands and automatic match display to work with them.

LaTeX Editing Commands

LaTeX mode, and its variant, SliTeX mode, provide a few extra features not applicable to plain TeX.

C-c C-o
Insert `\begin' and `\end' for LaTeX block and position point on a line between them. (tex-latex-block).
C-c C-e
Close the last unended block for LaTeX (tex-close-latex-block).

In LaTeX input, `\begin' and `\end' commands are used to group blocks of text. To insert a `\begin' and a matching `\end' (on a new line following the `\begin'), use C-c C-o (tex-latex-block). A blank line is inserted between the two, and point is left there. You can use completion when you enter the block type; to specify additional block type names beyond the standard list, set the variable latex-block-names. For example, here's how to add `theorem', `corollary', and `proof':

(setq latex-block-names '("theorem" "corollary" "proof"))

In LaTeX input, `\begin' and `\end' commands must balance. You can use C-c C-e (tex-close-latex-block) to insert automatically a matching `\end' to match the last unmatched `\begin'. It indents the `\end' to match the corresponding `\begin'. It inserts a newline after `\end' if point is at the beginning of a line.

TeX Printing Commands

You can invoke TeX as an inferior of Emacs on either the entire contents of the buffer or just a region at a time. Running TeX in this way on just one chapter is a good way to see what your changes look like without taking the time to format the entire file.

C-c C-r
Invoke TeX on the current region, together with the buffer's header (tex-region).
C-c C-b
Invoke TeX on the entire current buffer (tex-buffer).
C-c TAB
Invoke BibTeX on the current file (tex-bibtex-file).
C-c C-f
Invoke TeX on the current file (tex-file).
C-c C-l
Recenter the window showing output from the inferior TeX so that the last line can be seen (tex-recenter-output-buffer).
C-c C-k
Kill the TeX subprocess (tex-kill-job).
C-c C-p
Print the output from the last C-c C-r, C-c C-b, or C-c C-f command (tex-print).
C-c C-v
Preview the output from the last C-c C-r, C-c C-b, or C-c C-f command (tex-view).
C-c C-q
Show the printer queue (tex-show-print-queue).

You can pass the current buffer through an inferior TeX by means of C-c C-b (tex-buffer). The formatted output appears in a temporary file; to print it, type C-c C-p (tex-print). Afterward, you can use C-c C-q (tex-show-print-queue) to view the progress of your output towards being printed. If your terminal has the ability to display TeX output files, you can preview the output on the terminal with C-c C-v (tex-view).

You can specify the directory to use for running TeX by setting the variable tex-directory. "." is the default value. If your environment variable TEXINPUTS contains relative directory names, or if your files contains `\input' commands with relative file names, then tex-directory must be "." or you will get the wrong results. Otherwise, it is safe to specify some other directory, such as "/tmp".

If you want to specify which shell commands are used in the inferior TeX, you can do so by setting the values of the variables tex-run-command, latex-run-command, slitex-run-command, tex-dvi-print-command, tex-dvi-view-command, and tex-show-queue-command. You must set the value of tex-dvi-view-command for your particular terminal; this variable has no default value. The other variables have default values that may (or may not) be appropriate for your system.

Normally, the file name given to these commands comes at the end of the command string; for example, `latex filename'. In some cases, however, the file name needs to be embedded in the command; an example is when you need to provide the file name as an argument to one command whose output is piped to another. You can specify where to put the file name with `*' in the command string. For example,

(setq tex-dvi-print-command "dvips -f * | lpr")

The terminal output from TeX, including any error messages, appears in a buffer called `*tex-shell*'. If TeX gets an error, you can switch to this buffer and feed it input (this works as in Shell mode; see section Interactive Inferior Shell). Without switching to this buffer you can scroll it so that its last line is visible by typing C-c C-l.

Type C-c C-k (tex-kill-job) to kill the TeX process if you see that its output is no longer useful. Using C-c C-b or C-c C-r also kills any TeX process still running.

You can also pass an arbitrary region through an inferior TeX by typing C-c C-r (tex-region). This is tricky, however, because most files of TeX input contain commands at the beginning to set parameters and define macros, without which no later part of the file will format correctly. To solve this problem, C-c C-r allows you to designate a part of the file as containing essential commands; it is included before the specified region as part of the input to TeX. The designated part of the file is called the header.

To indicate the bounds of the header in Plain TeX mode, you insert two special strings in the file. Insert `%**start of header' before the header, and `%**end of header' after it. Each string must appear entirely on one line, but there may be other text on the line before or after. The lines containing the two strings are included in the header. If `%**start of header' does not appear within the first 100 lines of the buffer, C-c C-r assumes that there is no header.

In LaTeX mode, the header begins with `\documentstyle' and ends with `\begin{document}'. These are commands that LaTeX requires you to use in any case, so nothing special needs to be done to identify the header.

The commands (tex-buffer) and (tex-region) do all of their work in a temporary directory, and do not have available any of the auxiliary files needed by TeX for cross-references; these commands are generally not suitable for running the final copy in which all of the cross-references need to be correct. When you want the auxiliary files, use C-c C-f (tex-file) which runs TeX on the current buffer's file, in that file's directory. Before TeX runs, you will be asked about saving any modified buffers. Generally, you need to use (tex-file) twice to get cross-references correct.

For LaTeX files, you can use BibTeX to process the auxiliary file for the current buffer's file. BibTeX looks up bibliographic citations in a data base and prepares the cited references for the bibliography section. The command C-c TAB (tex-bibtex-file) runs the shell command (tex-bibtex-command) to produce a `.bbl' file for the current buffer's file. Generally, you need to do C-c C-f (tex-file) once to generate the `.aux' file, then do C-c TAB (tex-bibtex-file), and then repeat C-c C-f (tex-file) twice more to get the cross-references correct.

Entering any kind of TeX mode runs the hooks text-mode-hook and tex-mode-hook. Then it runs either plain-tex-mode-hook or latex-mode-hook, whichever is appropriate. For SliTeX files, it calls slitex-mode-hook. Starting the TeX shell runs the hook tex-shell-hook. See section Hooks.

Unix TeX Distribution

TeX for Unix systems can be obtained from the University of Washington for a distribution fee.

To order a full distribution, specify whether you prefer 1/4 inch QIC-24 or 4mm DAT tape (9-track reel-to-reel is no longer available) and send $210.00 for a (tar or cpio) cartridge, payable to the University of Washington to:

Pierre MacKay
Department of Classics
Denny Hall, Mail Stop DH-10
University of Washington
Seattle, Washington 98195

Purchase orders are acceptable, but there is an extra charge of $10.00, to pay for processing charges.

For overseas orders please add $20.00 to the base cost for shipment via air parcel post, or $30.00 for shipment via courier.

The normal distribution is a tar tape, blocked 20, 1600 bpi, on an industry standard 2400 foot half-inch reel. The physical format for the 1/4 inch streamer cartridges is QIC-24. System V tapes can be written in cpio format, blocked 5120 bytes, with ASCII headers.

Nroff Mode

Nroff mode is a mode like Text mode but modified to handle nroff commands present in the text. Invoke M-x nroff-mode to enter this mode. It differs from Text mode in only a few ways. All nroff command lines are considered paragraph separators, so that filling will never garble the nroff commands. Pages are separated by `.bp' commands. Comments start with backslash-doublequote. Also, three special commands are provided that are not in Text mode:

M-n
Move to the beginning of the next line that isn't an nroff command (forward-text-line). An argument is a repeat count.
M-p
Like M-n but move up (backward-text-line).
M-?
Prints in the echo area the number of text lines (lines that are not nroff commands) in the region (count-text-lines).

The other feature of Nroff mode is that you can turn on Electric Nroff mode. This is a minor mode that you can turn on or off with M-x electric-nroff-mode (see section Minor Modes). When the mode is on, each time you use RET to end a line that contains an nroff command that opens a kind of grouping, the matching nroff command to close that grouping is automatically inserted on the following line. For example, if you are at the beginning of a line and type . ( b RET, this inserts the matching command `.)b' on a new line following point.

If you use Outline minor mode with Nroff mode (see section Outline Mode), heading lines are lines of the form `.H' followed by a number (the header level).

Entering Nroff mode runs the hook text-mode-hook, followed by the hook nroff-mode-hook (see section Hooks).

Go to the previous, next section.