Go to the previous, next section.
Emacs has many commands designed to understand the syntax of programming languages such as Lisp and C. These commands can
The commands for words, sentences and paragraphs are very useful in editing code even though their canonical application is for editing human language text. Most symbols contain words (see section Words); sentences can be found in strings and comments (see section Sentences). Paragraphs per se are not present in code, but the paragraph commands are useful anyway, because Lisp mode and C mode define paragraphs to begin and end at blank lines (see section Paragraphs). Judicious use of blank lines to make the program clearer will also provide interesting chunks of text for the paragraph commands to work on.
The selective display feature is useful for looking at the overall structure of a function (see section Selective Display). This feature causes only the lines that are indented less than a specified amount to appear on the screen.
Emacs has major modes for the programming languages Lisp, Scheme (a variant of Lisp), C, Fortran and Muddle. Ideally, a major mode should be implemented for each programming language that you might want to edit with Emacs; but often the mode for one language can serve for other syntactically similar languages. The language modes that exist are those that someone decided to take the trouble to write.
There are several forms of Lisp mode, which differ in the way they interface to Lisp execution. See section Major Modes for Lisp.
Each of the programming language modes defines the TAB key to run
an indentation function that knows the indentation conventions of that
language and updates the current line's indentation accordingly. For
example, in C mode TAB is bound to c-indent-line. LFD
is normally defined to do RET followed by TAB; thus, it too
indents in a mode-specific fashion.
In most programming languages, indentation is likely to vary from line to
line. So the major modes for those languages rebind DEL to treat a
tab as if it were the equivalent number of spaces (using the command
backward-delete-char-untabify). This makes it possible to rub out
indentation one column at a time without worrying whether it is made up of
spaces or tabs. Use C-b C-d to delete a tab character before point,
in these modes.
Programming language modes define paragraphs to be separated only by blank lines, so that the paragraph commands remain useful. Auto Fill mode, if enabled in a programming language major mode, indents the new lines which it creates.
Turning on a major mode calls a user-supplied function called the
mode hook, which is the value of a Lisp variable. For example,
turning on C mode calls the value of the variable c-mode-hook if
that value exists and is non-nil. Mode hook variables for other
programming language modes include lisp-mode-hook,
emacs-lisp-mode-hook, lisp-interaction-mode-hook,
scheme-mode-hook and muddle-mode-hook. The mode hook
function receives no arguments.
By convention, Emacs keys for dealing with balanced expressions are usually Control-Meta- characters. They tend to be analogous in function to their Control- and Meta- equivalents. These commands are usually thought of as pertaining to expressions in programming languages, but can be useful with any language in which some sort of parentheses exist (including English).
These commands fall into two classes. Some deal only with lists (parenthetical groupings). They see nothing except parentheses, brackets, braces (whichever ones must balance in the language you are working with), and escape characters that might be used to quote those.
The other commands deal with expressions or sexps. The word `sexp' is derived from s-expression, the ancient term for an expression in Lisp. But in Emacs, the notion of `sexp' is not limited to Lisp. It refers to an expression in whatever language your program is written in. Each programming language has its own major mode, which customizes the syntax tables so that expressions in that language count as sexps.
Sexps typically include symbols, numbers, and string constants, as well as anything contained in parentheses, brackets or braces.
In languages that use prefix and infix operators, such as C, it is not possible for all expressions to be sexps. For example, C mode does not recognize `foo + bar' as a sexp, even though it is a C expression; it recognizes `foo' as one sexp and `bar' as another, with the `+' as punctuation between them. This is a fundamental ambiguity: both `foo + bar' and `foo' are legitimate choices for the sexp to move over if point is at the `f'. Note that `(foo + bar)' is a sexp in C mode.
Some languages have obscure forms of syntax for expressions that nobody has bothered to make Emacs understand properly.
forward-sexp).
backward-sexp).
kill-sexp).
backward-up-list).
down-list).
forward-list).
backward-list).
transpose-sexps).
mark-sexp).
To move forward over a sexp, use C-M-f (forward-sexp). If
the first significant character after point is an opening delimiter
(`(' in Lisp; `(', `[' or `{' in C), C-M-f
moves past the matching closing delimiter. If the character begins a
symbol, string, or number, C-M-f moves over that. If the character
after point is a closing delimiter, C-M-f gets an error.
The command C-M-b (backward-sexp) moves backward over a
sexp. The detailed rules are like those above for C-M-f, but with
directions reversed. If there are any prefix characters (singlequote,
backquote and comma, in Lisp) preceding the sexp, C-M-b moves back
over them as well.
C-M-f or C-M-b with an argument repeats that operation the specified number of times; with a negative argument, it moves in the opposite direction.
The sexp commands move across comments as if they were whitespace, in languages such as C where the comment-terminator can be recognized. In Lisp, and other languages where comments run until the end of a line, it is very difficult to ignore comments when parsing backwards; therefore, in such languages the sexp commands treat the text of comments as if it were code.
Killing a sexp at a time can be done with C-M-k (kill-sexp).
C-M-k kills the characters that C-M-f would move over.
The list commands move over lists like the sexp commands but skip
blithely over any number of other kinds of sexps (symbols, strings, etc).
They are C-M-n (forward-list) and C-M-p
(backward-list). The main reason they are useful is that they
usually ignore comments (since the comments usually do not contain any
lists).
C-M-n and C-M-p stay at the same level in parentheses, when
that's possible. To move up one (or n) levels, use C-M-u
(backward-up-list).
C-M-u moves backward up past one unmatched opening delimiter. A
positive argument serves as a repeat count; a negative argument reverses
direction of motion and also requests repetition, so it moves forward and
up one or more levels.
To move down in list structure, use C-M-d (down-list). In Lisp mode,
where `(' is the only opening delimiter, this is nearly the same as
searching for a `('. An argument specifies the number of levels
of parentheses to go down.
A somewhat random-sounding command which is nevertheless easy to use is
C-M-t (transpose-sexps), which drags the previous sexp across
the next one. An argument serves as a repeat count, and a negative
argument drags backwards (thus canceling out the effect of C-M-t with
a positive argument). An argument of zero, rather than doing nothing,
transposes the sexps ending after point and the mark.
To make the region be the next sexp in the buffer, use C-M-@
(mark-sexp) which sets mark at the same place that C-M-f would
move to. C-M-@ takes arguments like C-M-f. In particular, a
negative argument is useful for putting the mark at the beginning of the
previous sexp.
The list and sexp commands' understanding of syntax is completely controlled by the syntax table. Any character can, for example, be declared to be an opening delimiter and act like an open parenthesis. See section The Syntax Table.
In Emacs, a parenthetical grouping at the top level in the buffer is
called a defun. The name derives from the fact that most top-level
lists in a Lisp file are instances of the special form defun, but
any top-level parenthetical grouping counts as a defun in Emacs parlance
regardless of what its contents are, and regardless of the programming
language in use. For example, in C, the body of a function definition is a
defun.
beginning-of-defun).
end-of-defun).
mark-defun).
The commands to move to the beginning and end of the current defun are
C-M-a (beginning-of-defun) and C-M-e (end-of-defun).
If you wish to operate on the current defun, use C-M-h
(mark-defun) which puts point at the beginning and mark at the end
of the current or next defun. For example, this is the easiest way to get
ready to move the defun to a different place in the text. In C mode,
C-M-h runs the function mark-c-function, which is almost the
same as mark-defun; the difference is that it backs up over the
argument declarations, function name and returned data type so that the
entire C function is inside the region.
Emacs assumes that any open-parenthesis found in the leftmost column is the start of a defun. Therefore, never put an open-parenthesis at the left margin in a Lisp file unless it is the start of a top level list. Never put an open-brace or other opening delimiter at the beginning of a line of C code unless it starts the body of a function. The most likely problem case is when you want an opening delimiter at the start of a line inside a string. To avoid trouble, put an escape character (`\', in C and Emacs Lisp, `/' in some other Lisp dialects) before the opening delimiter. It will not affect the contents of the string.
In the remotest past, the original Emacs found defuns by moving upward a level of parentheses until there were no more levels to go up. This always required scanning all the way back to the beginning of the buffer, even for a small function. To speed up the operation, Emacs was changed to assume that any `(' (or other character assigned the syntactic class of opening-delimiter) at the left margin is the start of a defun. This heuristic was nearly always right and avoided the costly scan; however, it mandated the convention described above.
The best way to keep a program properly indented ("ground") is to use Emacs to re-indent it as you change it. Emacs has commands to indent properly either a single line, a specified number of lines, or all of the lines inside a single parenthetical grouping.
newline-and-indent).
The basic indentation command is TAB, which gives the current line
the correct indentation as determined from the previous lines. The
function that TAB runs depends on the major mode; it is lisp-indent-line
in Lisp mode, c-indent-line in C mode, etc. These functions
understand different syntaxes for different languages, but they all do
about the same thing. TAB in any programming language major mode
inserts or deletes whitespace at the beginning of the current line,
independent of where point is in the line. If point is inside the
whitespace at the beginning of the line, TAB leaves it at the end of
that whitespace; otherwise, TAB leaves point fixed with respect to
the characters around it.
Use C-q TAB to insert a tab at point.
When entering a large amount of new code, use LFD (newline-and-indent),
which is equivalent to a RET followed by a TAB. LFD creates
a blank line, and then gives it the appropriate indentation.
TAB indents the second and following lines of the body of a parenthetical grouping each under the preceding one; therefore, if you alter one line's indentation to be nonstandard, the lines below will tend to follow it. This is the right behavior in cases where the standard result of TAB is unaesthetic.
Remember that an open-parenthesis, open-brace or other opening delimiter at the left margin is assumed by Emacs (including the indentation routines) to be the start of a function. Therefore, you must never have an opening delimiter in column zero that is not the beginning of a function, not even inside a string. This restriction is vital for making the indentation commands fast; you must simply accept it. See section Defuns, for more information on this.
When you wish to re-indent several lines of code which have been altered or moved to a different level in the list structure, you have several commands available.
indent-sexp).
indent-region).
You can re-indent the contents of a single list by positioning point
before the beginning of it and typing C-M-q (indent-sexp in
Lisp mode, indent-c-exp in C mode; also bound to other suitable
functions in other modes). The indentation of the line the sexp starts on
is not changed; therefore, only the relative indentation within the list,
and not its position, is changed. To correct the position as well, type a
TAB before the C-M-q.
If the relative indentation within a list is correct but the indentation of its beginning is not, go to the line the list begins on and type C-u TAB. When TAB is given a numeric argument, it moves all the lines in the grouping starting on the current line sideways the same amount that the current line moves. It is clever, though, and does not move lines that start inside strings, or C preprocessor lines when in C mode.
Another way to specify the range to be re-indented is with point and
mark. The command C-M-\ (indent-region) applies TAB to every line
whose first character is between point and mark.
The indentation pattern for a Lisp expression can depend on the function called by the expression. For each Lisp function, you can choose among several predefined patterns of indentation, or define an arbitrary one with a Lisp program.
The standard pattern of indentation is as follows: the second line of the expression is indented under the first argument, if that is on the same line as the beginning of the expression; otherwise, the second line is indented underneath the function name. Each following line is indented under the previous line whose nesting depth is the same.
If the variable lisp-indent-offset is non-nil, it overrides
the usual indentation pattern for the second line of an expression, so that
such lines are always indented lisp-indent-offset more columns than
the containing list.
The standard pattern is overridden for certain functions. Functions
whose names start with def always indent the second line by
lisp-body-indention extra columns beyond the open-parenthesis
starting the expression.
The standard pattern can be overridden in various ways for individual
functions, according to the lisp-indent-hook property of the
function name. There are four possibilities for this property:
nil
defun
def is used for
this function also.
lisp-body-indent
more columns than the open-parenthesis starting the containing
expression. If the argument is distinguished and is either the first
or second argument, it is indented twice that many extra columns.
If the argument is distinguished and not the first or second argument,
the standard pattern is followed for that line.
parse-partial-sexp (a Lisp primitive for
indentation and nesting computation) when it parses up to the
beginning of this line.
Two variables control which commands perform C indentation and when.
If c-auto-newline is non-nil, newlines are inserted both
before and after braces that you insert, and after colons and semicolons.
Correct C indentation is done on all the lines that are made this way.
If c-tab-always-indent is nil, the TAB command
in C mode does indentation only if point is at the left margin or within
the line's indentation. If there is non-whitespace to the left of point,
then TAB just inserts a tab character in the buffer. Normally,
this variable is t, and TAB always reindents the current line.
C does not have anything analogous to particular function names for which special forms of indentation are desirable. However, it has a different need for customization facilities: many different styles of C indentation are in common use.
There are six variables you can set to control the style that Emacs C mode will use.
c-indent-level
c-continued-statement-offset
c-brace-offset
c-brace-imaginary-offset
c-argdecl-indent
c-label-offset
The variable c-indent-level controls the indentation for C
statements with respect to the surrounding block. In the example
{
foo ();
the difference in indentation between the lines is c-indent-level.
Its standard value is 2.
If the open-brace beginning the compound statement is not at the beginning
of its line, the c-indent-level is added to the indentation of the
line, not the column of the open-brace. For example,
if (losing) {
do_this ();
One popular indentation style is that which results from setting
c-indent-level to 8 and putting open-braces at the end of a line in
this way. I prefer to put the open-brace on a separate line.
In fact, the value of the variable c-brace-imaginary-offset is
also added to the indentation of such a statement. Normally this variable
is zero. Think of this variable as the imaginary position of the open
brace, relative to the first nonblank character on the line. By setting
this variable to 4 and c-indent-level to 0, you can get this style:
if (x == y) {
do_it ();
}
When c-indent-level is zero, the statements inside most braces
will line up right under the open brace. But there is an exception made
for braces in column zero, such as surrounding a function's body. The
statements just inside it do not go at column zero. Instead,
c-brace-offset and c-continued-statement-offset (see below)
are added to produce a typical offset between brace levels, and the
statements are indented that far.
c-continued-statement-offset controls the extra indentation for a
line that starts within a statement (but not within parentheses or
brackets). These lines are usually statements that are within other
statements, such as the then-clauses of if statements and the bodies
of while statements. This parameter is the difference in
indentation between the two lines in
if (x == y) do_it ();
Its standard value is 2. Some popular indentation styles correspond to a
value of zero for c-continued-statement-offset.
c-brace-offset is the extra indentation given to a line that
starts with an open-brace. Its standard value is zero;
compare
if (x == y)
{
with
if (x == y) do_it ();
if c-brace-offset were set to 4, the first example would become
if (x == y)
{
c-argdecl-indent controls the indentation of declarations of the
arguments of a C function. It is absolute: argument declarations receive
exactly c-argdecl-indent spaces. The standard value is 5, resulting
in code like this:
char *
index (string, c)
char *string;
int c;
c-label-offset is the extra indentation given to a line that
contains a label, a case statement, or a default: statement. Its
standard value is -2, resulting in code like this
switch (c)
{
case 'x':
If c-label-offset were zero, the same code would be indented as
switch (c)
{
case 'x':
This example assumes that the other variables above also have their standard values.
I strongly recommend that you try out the indentation style produced by the standard settings of these variables, together with putting open braces on separate lines. You can see how it looks in all the C source files of GNU Emacs.
The Emacs parenthesis-matching feature is designed to show automatically how parentheses match in the text. Whenever a self-inserting character that is a closing delimiter is typed, the cursor moves momentarily to the location of the matching opening delimiter, provided that is on the screen. If it is not on the screen, some text starting with that opening delimiter is displayed in the echo area. Either way, you can tell what grouping is being closed off.
In Lisp, automatic matching applies only to parentheses. In C, it applies to braces and brackets too. Emacs knows which characters to regard as matching delimiters based on the syntax table, which is set by the major mode. See section The Syntax Table.
If the opening delimiter and closing delimiter are mismatched--such as in `[x)'---a warning message is displayed in the echo area. The correct matches are specified in the syntax table.
Two variables control parenthesis match display. blink-matching-paren
turns the feature on or off; nil turns it off, but the default is
t to turn match display on. blink-matching-paren-distance
specifies how many characters back to search to find the matching opening
delimiter. If the match is not found in that far, scanning stops, and
nothing is displayed. This is to prevent scanning for the matching
delimiter from wasting lots of time when there is no match. The default
is 4000.
The comment commands insert, kill and align comments.
indent-for-comment).
set-comment-column).
kill-comment).
indent-new-comment-line).
The command that creates a comment is Meta-; (indent-for-comment).
If there is no comment already on the line, a new comment is created,
aligned at a specific column called the comment column. The comment
is created by inserting the string Emacs thinks comments should start with
(the value of comment-start; see below). Point is left after that
string. If the text of the line extends past the comment column, then the
indentation is done to a suitable boundary (usually, at least one space is
inserted). If the major mode has specified a string to terminate comments,
that is inserted after point, to keep the syntax valid.
Meta-; can also be used to align an existing comment. If a line already contains the string that starts comments, then M-; just moves point after it and re-indents it to the conventional place. Exception: comments starting in column 0 are not moved.
Some major modes have special rules for indenting certain kinds of comments in certain contexts. For example, in Lisp code, comments which start with two semicolons are indented as if they were lines of code, instead of at the comment column. Comments which start with three semicolons are supposed to start at the left margin. Emacs understands these conventions by indenting a double-semicolon comment using TAB, and by not changing the indentation of a triple-semicolon comment at all.
;; This function is just an example ;;; Here either two or three semicolons are appropriate. (defun foo (x) ;;; And now, the first part of the function: ;; The following line adds one. (1+ x)) ; This line adds one.
In C code, a comment preceded on its line by nothing but whitespace is indented like a line of code.
Even when an existing comment is properly aligned, M-; is still useful for moving directly to the start of the comment.
C-u - C-x ; (kill-comment) kills the comment on the current line,
if there is one. The indentation before the start of the comment is killed
as well. If there does not appear to be a comment in the line, nothing is
done. To reinsert the comment on another line, move to the end of that
line, do C-y, and then do M-; to realign it. Note that
C-u - C-x ; is not a distinct key; it is C-x ; (set-comment-column)
with a negative argument. That command is programmed so that when it
receives a negative argument it calls kill-comment. However,
kill-comment is a valid command which you could bind directly to a
key if you wanted to.
If you are typing a comment and find that you wish to continue it on
another line, you can use the command Meta-LFD (indent-new-comment-line),
which terminates the comment you are typing, creates a new blank line
afterward, and begins a new comment indented under the old one. When Auto
Fill mode is on, going past the fill column while typing a comment causes
the comment to be continued in just this fashion. If point is not at the
end of the line when M-LFD is typed, the text on the rest of
the line becomes part of the new comment line.
The comment column is stored in the variable comment-column. You
can set it to a number explicitly. Alternatively, the command C-x ;
(set-comment-column) sets the comment column to the column point is
at. C-u C-x ; sets the comment column to match the last comment
before point in the buffer, and then does a Meta-; to align the
current line's comment under the previous one. Note that C-u - C-x ;
runs the function kill-comment as described above.
comment-column 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. Many major modes initialize this variable
for the current buffer.
The comment commands recognize comments based on the regular expression
that is the value of the variable comment-start-skip. This regexp
should not match the null string. It may match more than the comment
starting delimiter in the strictest sense of the word; for example, in C
mode the value of the variable is "/\\*+ *", which matches extra
stars and spaces after the `/*' itself. (Note that `\\' is
needed in Lisp syntax to include a `\' in the string, which is needed
to deny the first star its special meaning in regexp syntax. See section Syntax of Regular Expressions.)
When a comment command makes a new comment, it inserts the value of
comment-start to begin it. The value of comment-end is
inserted after point, so that it will follow the text that you will insert
into the comment. In C mode, comment-start has the value
"/* " and comment-end has the value " */".
comment-multi-line controls how M-LFD (indent-new-comment-line)
behaves when used inside a comment. If comment-multi-line is
nil, as it normally is, then the comment on the starting line is
terminated and a new comment is started on the new following line. If
comment-multi-line is not nil, then the new following line is
set up as part of the same comment that was found on the starting line.
This is done by not inserting a terminator on the old line, and not
inserting a starter on the new line. In languages where multi-line comments
work, the choice of value for this variable is a matter of taste.
The variable comment-indent-hook should contain a function that
will be called to compute the indentation for a newly inserted comment or
for aligning an existing comment. It is set differently by various major
modes. The function is called with no arguments, but with point at the
beginning of the comment, or at the end of a line if a new comment is to be
inserted. It should return the column in which the comment ought to start.
For example, in Lisp mode, the indent hook function bases its decision
on how many semicolons begin an existing comment, and on the code in the
preceding lines.
When you are debugging C code that uses macros, sometimes it is hard to figure out precisely how the macros expand. The command M-x c-macro-expand. It runs the C preprocessor and shows you what expansion results from the region. The portion of the buffer before the region is also included in preprocessing, for the sake of macros defined there, but the output from this part isn't shown.
insert-parentheses).
move-over-close-and-reindent).
The two commands, M-( (insert-parentheses) and M-)
(move-over-close-and-reindent), are designed to facilitate a style of
editing which keeps parentheses balanced at all times. M-( inserts a
pair of parentheses, either together as in `()', or, if given an
argument, around the next several sexps, and leaves point after the open
parenthesis. Instead of typing ( F O O ), you can type M-( F O
O, which has the same effect except for leaving the cursor before the
close parenthesis. Then you would type M-), which moves past the
close parenthesis, deleting any indentation preceding it (in this example
there is none), and indenting with LFD after it.
Usually completion happens in the minibuffer. But one kind of completion is available in all buffers: completion for Lisp symbol names.
The command M-TAB (lisp-complete-symbol) takes the
partial Lisp symbol before point to be an abbreviation, and compares it
against all nontrivial Lisp symbols currently known to Emacs. Any
additional characters that they all have in common are inserted at point.
Nontrivial symbols are those that have function definitions, values or
properties.
If there is an open-parenthesis immediately before the beginning of the partial symbol, only symbols with function definitions are considered as completions.
If the partial name in the buffer has more than one possible completion and they have no additional characters in common, a list of all possible completions is displayed in another window.
As you edit Lisp code to be run in Emacs, the commands C-h f
(describe-function) and C-h v (describe-variable) can
be used to print documentation of functions and variables that you want to
call. These commands use the minibuffer to read the name of a function or
variable to document, and display the documentation in a window.
For extra convenience, these commands provide default arguments based on the code in the neighborhood of point. C-h f sets the default to the function called in the innermost list containing point. C-h v uses the symbol name around or adjacent to point as its default.
Documentation on Unix commands, system calls and libraries can be
obtained with the M-x manual-entry command. This reads a topic as an
argument, and displays the text on that topic from the Unix manual.
manual-entry always searches all 8 sections of the manual, and
concatenates all the entries that are found. For example, the topic
`termcap' finds the description of the termcap library from section 3,
followed by the description of the termcap data base from section 5.
The Emacs command M-x add-change-log-entry helps you keep a record of when and why you have changed a program. It assumes that you have a file in which you write a chronological sequence of entries describing individual changes. The default is to store the change entries in a file called `ChangeLog' in the same directory as the file you are editing. The same `ChangeLog' file therefore records changes for all the files in the directory.
A change log entry starts with a header line that contains your name and the current date. Aside from these header lines, every line in the change log starts with a tab. One entry can describe several changes; each change starts with a line starting with a tab and a star. M-x add-change-log-entry visits the change log file and creates a new entry unless the most recent entry is for today's date and your name. In either case, it adds a new line to start the description of another change just after the header line of the entry. When M-x add-change-log-entry is finished, all is prepared for you to edit in the description of what you changed and how. You must then save the change log file yourself.
The change log file is always visited in Indented Text mode, which means that LFD and auto-filling indent each new line like the previous line. This is convenient for entering the contents of an entry, which must all be indented. See section Text Mode.
An alternative convenient command for starting a change log entry is
C-x 4 a (add-change-log-entry-other-window). It resembles
add-change-log-entry except that it visits the change log in
another window, and always uses the file `./ChangeLog'---it does
not ask you for the file name.
Here is an example of the formatting conventions used in the change log for Emacs:
Wed Jun 26 19:29:32 1985 Richard M. Stallman (rms at mit-prep)
* xdisp.c (try_window_id):
If C-k is done at end of next-to-last line,
this fn updates window_end_vpos and cannot leave
window_end_pos nonnegative (it is zero, in fact).
If display is preempted before lines are output,
this is inconsistent. Fix by setting
blank_end_of_window to nonzero.
Tue Jun 25 05:25:33 1985 Richard M. Stallman (rms at mit-prep)
* cmds.c (Fnewline):
Call the auto fill hook if appropriate.
* xdisp.c (try_window_id):
If point is found by compute_motion after xp, record that
permanently. If display_text_line sets point position wrong
(case where line is killed, point is at eob and that line is
not displayed), set it again in final compute_motion.
A tag table is a description of how a multi-file program is broken up into files. It lists the names of the component files and the names and positions of the functions in each file. Grouping the related files makes it possible to search or replace through all the files with one command. Recording the function names and positions makes possible the Meta-. command which you can use to find the definition of a function without having to know which of the files it is in.
Tag tables are stored in files called tag table files. The conventional name for a tag table file is `TAGS'.
Each entry in the tag table records the name of one tag, the name of the file that the tag is defined in (implicitly), and the position in that file of the tag's definition.
Just what names from the described files are recorded in the tag table depends on the programming language of the described file. They normally include all functions and subroutines, and may also include global variables, data types, and anything else convenient. In any case, each name recorded is called a tag.
In Lisp code, any function defined with defun, any variable
defined with defvar or defconst, and in general the first
argument of any expression that starts with `(def' in column zero, is
a tag.
In C code, any C function is a tag, and so is any typedef if -t is
specified when the tag table is constructed.
In Fortran code, functions and subroutines are tags.
In LaTeX text, the argument of any of the commands \chapter,
\section, \subsection, \subsubsection, \eqno,
\label, \ref, \cite, \bibitem and
\typeout is a tag.
The etags program is used to create a tag table file. It knows
the syntax of C, Fortran, LaTeX, Scheme and Emacs Lisp/Common Lisp. To
use etags, type
etags inputfiles...
as a shell command. It reads the specified files and writes a tag table
named `TAGS' in the current working directory. etags
recognizes the language used in an input file based on its file name and
contents; there are no switches for specifying the language. The -t
switch tells etags to record typedefs in C code as tags.
If the tag table data become outdated due to changes in the files described in the table, the way to update the tag table is the same way it was made in the first place. It is not necessary to do this often.
If the tag table fails to record a tag, or records it for the wrong file, then Emacs cannot possibly find its definition. However, if the position recorded in the tag table becomes a little bit wrong (due to some editing in the file that the tag definition is in), the only consequence is to slow down finding the tag slightly. Even if the stored position is very wrong, Emacs will still find the tag, but it must search the entire file for it.
So you should update a tag table when you define new tags that you want to have listed, or when you move tag definitions from one file to another, or when changes become substantial. Normally there is no need to update the tag table after each edit, or even every day.
Emacs has at any time one selected tag table, and all the commands for working with tag tables use the selected one. To select a tag table, type M-x visit-tags-table, which reads the tag table file name as an argument. The name `TAGS' in the default directory is used as the default file name.
All this command does is store the file name in the variable
tags-file-name. Emacs does not actually read in the tag table
contents until you try to use them. Setting this variable yourself is just
as good as using visit-tags-table. The variable's initial value is
nil; this value tells all the commands for working with tag tables
that they must ask for a tag table file name to use.
The most important thing that a tag table enables you to do is to find the definition of a specific tag.
find-tag).
find-tag-other-window).
M-. (find-tag) is the command to find the definition of a
specified tag. It searches through the tag table for that tag, as a
string, and then uses the tag table info to determine the file that the
definition is in and the approximate character position in the file of the
definition. Then find-tag visits that file, moves point to the
approximate character position, and starts searching ever-increasing
distances away for the the text that should appear at the beginning of the
definition.
If an empty argument is given (just type RET), the sexp in the buffer before or around point is used as the name of the tag to find. See section Lists and Sexps, for info on sexps.
The argument to find-tag need not be the whole tag name; it can be
a substring of a tag name. However, there can be many tag names containing
the substring you specify. Since find-tag works by searching the
text of the tag table, it finds the first tag in the table that the
specified substring appears in.
The way to find other tags that match the substring is to give
find-tag a numeric argument, as in C-u M-.; this does not
read a tag name, but continues searching the tag table's text for
another tag containing the same substring last used. If you have a real
META key, M-0 M-. is an easier alternative to C-u
M-.. (That is a zero in M-0.)
Like most commands that can switch buffers, find-tag has another
similar command that displays the new buffer in another window. C-x 4
. invokes the function find-tag-other-window. (This key sequence
ends with a period.)
Emacs comes with a tag table file `TAGS', in the `src'
subdirectory, which includes all the Lisp libraries and all the C
sources of Emacs. By specifying this file with visit-tags-table
and then using M-. you can quickly look at the source of any
Emacs function.
The commands in this section visit and search all the files listed in the selected tag table, one by one. For these commands, the tag table serves only to specify a sequence of files to search. A related command is M-x grep (see section Running `make', or Compilers Generally).
query-replace on each file in the selected tag table.
tags-loop-continue).
M-x tags-search reads a regexp using the minibuffer, then visits
the files of the selected tag table one by one, and searches through each
one for that regexp. It displays the name of the file being searched so
you can follow its progress. As soon as an occurrence is found,
tags-search returns.
Having found one match, you probably want to find all the rest. To find
one more match, type M-, (tags-loop-continue) to resume the
tags-search. This searches the rest of the current buffer, followed
by the remaining files of the tag table.
M-x tags-query-replace performs a single query-replace
through all the files in the tag table. It reads a string to search for
and a string to replace with, just like ordinary M-x query-replace.
It searches much like M-x tags-search but repeatedly, processing
matches according to your input. See section Replacement Commands, for more information on
query-replace.
It is possible to get through all the files in the tag table with a single invocation of M-x tags-query-replace. But since any unrecognized character causes the command to exit, you may need to continue where you left off. M-, can be used for this. It resumes the last tags search or replace command that you did.
It may have struck you that tags-search is a lot like grep.
You can also run grep itself as an inferior of Emacs and have Emacs
show you the matching lines one by one. This works mostly the same as
running a compilation and having Emacs show you where the errors were.
See section Running `make', or Compilers Generally.
If you wish to process all the files in the selected tag table, but M-x tags-search and M-x tags-query-replace in particular are not what you want, you can use M-x next-file.
M-x list-tags reads the name of one of the files described by the selected tag table, and displays a list of all the tags defined in that file. The "file name" argument is really just a string to compare against the names recorded in the tag table; it is read as a string rather than as a file name. Therefore, completion and defaulting are not available, and you must enter the string the same way it appears in the tag table. Do not include a directory as part of the file name unless the file name recorded in the tag table includes a directory.
M-x tags-apropos is like apropos for tags. It reads a regexp,
then finds all the tags in the selected tag table whose entries match that
regexp, and displays the tag names found.
Fortran mode provides special motion commands for Fortran statements and subprograms, and indentation commands that understand Fortran conventions of nesting, line numbers and continuation statements.
Special commands for comments are provided because Fortran comments are unlike those of other languages.
Built-in abbrevs optionally save typing when you insert Fortran keywords.
Use M-x fortran-mode to switch to this major mode. Doing so calls
the value of fortran-mode-hook as a function of no arguments if
that variable has a value that is not nil.
Fortran mode was contributed by Michael Prange.
Fortran mode provides special commands to move by subprograms (functions and subroutines) and by statements. There is also a command to put the region around one subprogram, convenient for killing it or moving it.
beginning-of-fortran-subprogram).
end-of-fortran-subprogram).
mark-fortran-subprogram).
fortran-next-statement).
fortran-previous-statement).
Special commands and features are needed for indenting Fortran code in order to make sure various syntactic entities (line numbers, comment line indicators and continuation line flags) appear in the columns that are required for standard Fortran.
fortran-indent-line).
fortran-indent-subprogram).
TAB is redefined by Fortran mode to reindent the current line for
Fortran (fortran-indent-line). Line numbers and continuation
markers are indented to their required columns, and the body of the
statement is independently indented based on its nesting in the program.
The key C-M-q is redefined as fortran-indent-subprogram, a
command to reindent all the lines of the Fortran subprogram (function or
subroutine) containing point.
The key M-LFD is redefined as fortran-split-line, a
command to split a line in the appropriate fashion for Fortran. In a
non-comment line, the second half becomes a continuation line and is
indented accordingly. In a comment line, both halves become separate
comment lines.
If a number is the first non-whitespace in the line, it is assumed to be a line number and is moved to columns 0 through 4. (Columns are always counted from 0 in GNU Emacs.) If the text on the line starts with the conventional Fortran continuation marker `$', it is moved to column 5. If the text begins with any non whitespace character in column 5, it is assumed to be an unconventional continuation marker and remains in column 5.
Line numbers of four digits or less are normally indented one space.
This amount is controlled by the variable fortran-line-number-indent
which is the maximum indentation a line number can have. Line numbers
are indented to right-justify them to end in column 4 unless that would
require more than this maximum indentation. The default value of the
variable is 1.
Simply inserting a line number is enough to indent it according to these
rules. As each digit is inserted, the indentation is recomputed. To turn
off this feature, set the variable fortran-electric-line-number to
nil. Then inserting line numbers is like inserting anything else.
Fortran mode assumes that you follow certain conventions that simplify the task of understanding a Fortran program well enough to indent it properly:
fortran-continuation-char.
By default, this character is `$'.
If you fail to follow these conventions, the indentation commands may indent some lines unaesthetically. However, a correct Fortran program will retain its meaning when reindented even if the conventions are not followed.
Several additional variables control how Fortran indentation works.
fortran-do-indent
fortran-if-indent
fortran-continuation-indent
fortran-check-all-num-for-matching-do
nil, indentation assumes that each `do'
statement ends on a `continue' statement. Therefore, when
computing indentation for a statement other than `continue', it
can save time by not checking for a `do' statement ending there.
If this is non-nil, indenting any numbered statement must check
for a `do' that ends there. The default is nil.
fortran-minimum-statement-indent
The usual Emacs comment commands assume that a comment can follow a line of code. In Fortran, the standard comment syntax requires an entire line to be just a comment. Therefore, Fortran mode replaces the standard Emacs comment commands and defines some new variables.
Fortran mode can also handle a nonstandard comment syntax where comments
start with `!' and can follow other text. Because only some Fortran
compilers accept this syntax, Fortran mode will not insert such comments
unless you have said in advance to do so. To do this, set the variable
comment-start to `"!"' (see section Variables).
fortran-comment-indent).
fortran-comment-region).
M-; in Fortran mode is redefined as the command
fortran-comment-indent. Like the usual M-; command, this
recognizes any kind of existing comment and aligns its text appropriately;
if there is no existing comment, a comment is inserted and aligned. But
inserting and aligning comments are not the same in Fortran mode as in
other modes.
When a new comment must be inserted, if the current line is blank, a full-line comment is inserted. On a non-blank line, a nonstandard `!' comment is inserted if you have said you want to use them. Otherwise a full-line comment is inserted on a new line before the current line.
Nonstandard `!' comments are aligned like comments in other
languages, but full-line comments are different. In a standard full-line
comment, the comment delimiter itself must always appear in column zero.
What can be aligned is the text within the comment. You can choose from
three styles of alignment by setting the variable
fortran-comment-indent-style to one of these values:
fixed
fortran-comment-line-column. This is the default.
relative
fortran-comment-line-column columns of indentation.
nil
In addition, you can specify the character to be used to indent within
full-line comments by setting the variable fortran-comment-indent-char
to the character you want to use.
Fortran mode introduces the two variables, comment-line-start and
comment-line-start-skip, which play for full-line comments the same
roles played by comment-start and comment-start-skip for
ordinary text-following comments. Normally these are set properly by
Fortran mode so you do not need to change them.
The normal Emacs comment command C-x ; has not been redefined. If you use `!' comments, this command can be used with them. Otherwise it is useless in Fortran mode.
The command C-c ; (fortran-comment-region) turns all the
lines of the region into comments by inserting the string `C$$$' at
the front of each one. With a numeric arg, the region is turned back into
live code by deleting `C$$$' from the front of each line in it. The
string used for these comments can be controlled by setting the variable
fortran-comment-region. Note that here we have an example of a
command and a variable with the same name; these two uses of the name never
conflict because in Lisp and in Emacs it is always clear from the context
which one is meant.
fortran-column-ruler).
fortran-window-create).
The command C-c C-r (fortran-column-ruler) shows a column
ruler momentarily above the current line. The comment ruler is two lines
of text that show you the locations of columns with special significance
in Fortran programs. Square brackets show the limits of the columns for
line numbers, and curly brackets show the limits of the columns for the
statement body. Column numbers appear above them.
Note that the column numbers count from zero, as always in GNU Emacs. As a result, the numbers may not be those you are familiar with; but the actual positions in the line are standard Fortran.
The text used to display the column ruler is the value of the variable
fortran-comment-ruler. By changing this variable, you can change
the display.
For even more help, use C-c C-w (fortran-window-create), a
command which splits the current window horizontally, making a window 72
columns wide. By editing in this window you can immediately see when you
make a line too wide to be correct Fortran.
Fortran mode provides many built-in abbrevs for common keywords and declarations. These are the same sort of abbrev that you can define yourself. To use them, you must turn on Abbrev mode (see section Abbrevs).
The built-in abbrevs are unusual in one way: they all start with a semicolon. You cannot normally use semicolons in an abbrev, but Fortran mode makes this possible by changing the syntax of semicolon to "word constituent".
For example, one built-in Fortran abbrev is `;c' for `continue'. If you insert `;c' and then insert a punctuation character such as a space or a newline, the `;c' will change automatically to `continue', provided Abbrev mode is enabled.
Type `;?' or `;C-h' to display a list of all the built-in Fortran abbrevs and what they stand for.
Go to the previous, next section.