@macro ccmode CC Mode
Welcome to @ccmode{}. This is a GNU Emacs mode for editing files
containing C, C++, Objective-C, Java, and CORBA IDL code. This
incarnation of the mode is descendant from `c-mode.el' (also called
"Boring Old C Mode" or BOCM :-), and `c++-mode.el' version
2, which I have been maintaining since 1992. @ccmode{} represents a
significant milestone in the mode's life. It has been fully merged back
with Emacs 19's `c-mode.el'. Also a new, more intuitive and
flexible mechanism for controlling indentation has been developed.
@ccmode{} supports the editing of K&R and ANSI C, ARM (1) C++, Objective-C, Java and CORBA's Interface Definition Language files. In this way, you can easily set up consistent coding styles for use in editing all C, C++, Objective-C, Java and IDL programs. @ccmode{} does not handle font-locking (a.k.a. syntax coloring, keyword highlighting) or anything of that nature, for any of these modes. Font-locking is handled by other Emacs packages.
This manual will describe the following:
Note that the name of this package is "@ccmode{}", but there is no top
level cc-mode entry point. All of the variables, commands, and
functions in @ccmode{} are prefixed with c-<thing>, and
c-mode, c++-mode, objc-mode, java-mode, and
idl-mode entry points are provided. This file is intended to be
a replacement for `c-mode.el' and `c++-mode.el'.
This distribution also contains a file called `cc-compat.el' which should ease your transition from BOCM to @ccmode{}. If you have a BOCM configuration you are really happy with, and want to postpone learning how to configure @ccmode{}, take a look at that file. It maps BOCM configuration variables to @ccmode{}'s new indentation model. It is not actively supported so for the long run, you should learn how to customize @ccmode{} to support your coding style.
A special word of thanks goes to Krishna Padmasola for his work in converting the original `README' file to Texinfo format. I'd also like to thank all the @ccmode{} victims who help enormously during the early beta stages of @ccmode{}'s development.
If you got this version of @ccmode{} with Emacs or XEmacs, it should work just fine right out of the box. Note however that you may not have the latest @ccmode{} release and may want to upgrade your copy.
If you are upgrading an existing @ccmode{} installation, please see the `README' file for installation details. @ccmode{} may not work with older versions of Emacs or XEmacs. See the @ccmode{} release notes Web pages for the latest information on Emacs version and package compatibility (see section Getting the latest CC Mode release).
Note that @ccmode{ no longer works with Emacs 18!} The `cc-mode-18.el' file is no longer distributed with @ccmode{}. If you haven't upgraded from Emacs 18 by now, you are out of luck.
You can find out what version of @ccmode{} you are using by visiting a C file and entering M-x c-version RET. You should see this message in the echo area:
Using CC Mode version 5.XX
where `XX' is the minor release number.
@ccmode{} has a new indentation engine, providing a simplified, yet flexible and general mechanism for customizing indentation. It separates indentation calculation into two steps: first, @ccmode{} analyzes the line of code being indented to determine the kind of language construct it's looking at, then it applies user defined offsets to the current line based on this analysis.
This section will briefly cover how indentation is calculated in @ccmode{}. It is important to understand the indentation model being used so that you will know how to customize @ccmode{} for your personal coding style.
The first thing @ccmode{} does when indenting a line of code, is to
analyze the line, determining the syntactic component list of the
construct on that line. A syntactic component consists of a pair
of information (in lisp parlance, a cons cell), where the first
part is a syntactic symbol, and the second part is a relative
buffer position. Syntactic symbols describe elements of C code
(2), e.g. statement,
substatement, class-open, class-close, etc.
See section Syntactic Symbols, for a complete list of currently recognized
syntactic symbols and their semantics. The variable
c-offsets-alist also contains the list of currently supported
syntactic symbols.
Conceptually, a line of C code is always indented relative to the indentation of some line higher up in the buffer. This is represented by the relative buffer position in the syntactic component.
Here is an example. Suppose we had the following code as the only thing
in a c++-mode buffer (3):
1: void swap( int& a, int& b )
2: {
3: int tmp = a;
4: a = b;
5: b = tmp;
6: }
We can use the command C-c C-s
(c-show-syntactic-information) to simply report what the
syntactic analysis is for the current line. Running this command on
line 4 of this example, we'd see in the echo area(4):
((statement . 35))
This tells us that the line is a statement and it is indented relative
to buffer position 35, which happens to be the `i' in int on
line 3. If you were to move point to line 3 and hit C-c C-s, you
would see:
((defun-block-intro . 29))
This indicates that the `int' line is the first statement in a top level function block, and is indented relative to buffer position 29, which is the brace just after the function header.
Here's another example:
1: int add( int val, int incr, int doit )
2: {
3: if( doit )
4: {
5: return( val + incr );
6: }
7: return( val );
8: }
Hitting C-c C-s on line 4 gives us:
((substatement-open . 46))
which tells us that this is a brace that opens a substatement block. (5)
Syntactic component lists can contain more than one component, and individual syntactic components need not have relative buffer positions. The most common example of this is a line that contains a comment only line.
1: void draw_list( List<Drawables>& drawables )
2: {
3: // call the virtual draw() method on each element in list
4: for( int i=0; i < drawables.count(), ++i )
5: {
6: drawables[i].draw();
7: }
8: }
Hitting C-c C-s on line 3 of this example gives:
((comment-intro) (defun-block-intro . 46))
and you can see that the syntactic component list contains two syntactic components. Also notice that the first component, `(comment-intro)' has no relative buffer position.
Indentation for a line is calculated using the syntactic component list derived in step 1 above (see section Syntactic Analysis). Each component contributes to the final total indentation of the line in two ways.
First, the syntactic symbols are looked up in the c-offsets-alist
variable, which is an association list of syntactic symbols and the
offsets to apply for those symbols. These offsets are added to a
running total.
Second, if the component has a relative buffer position, @ccmode{} adds the column number of that position to the running total. By adding up the offsets and columns for every syntactic component on the list, the final total indentation for the current line is computed.
Let's use our two code examples above to see how this works. Here is our first example again:
1: void swap( int& a, int& b )
2: {
3: int tmp = a;
4: a = b;
5: b = tmp;
6: }
Let's say point is on line 3 and we hit the TAB key to re-indent the line. Remember that the syntactic component list for that line is:
((defun-block-intro . 29))
@ccmode{} looks up defun-block-intro in the
c-offsets-alist variable. Let's say it finds the value `4';
it adds this to the running total (initialized to zero), yielding a
running total indentation of 4 spaces.
Next @ccmode{} goes to buffer position 29 and asks for the current column. This brace is in column zero, so @ccmode{} adds `0' to the running total. Since there is only one syntactic component on the list for this line, indentation calculation is complete, and the total indentation for the line is 4 spaces.
Here's another example:
1: int add( int val, int incr, int doit )
2: {
3: if( doit )
4: {
5: return( val + incr );
6: }
7: return( val );
8: }
If we were to hit TAB on line 4 in the above example, the same basic process is performed, despite the differences in the syntactic component list. Remember that the list for this line is:
((substatement-open . 46))
Here, @ccmode{} first looks up the substatement-open symbol
in c-offsets-alist. Let's say it finds the value `4'. This
yields a running total of 4. @ccmode{} then goes to
buffer position 46, which is the `i' in if on line 3. This
character is in the fourth column on that line so adding this to the
running total yields an indentation for the line of 8 spaces.
Simple, huh?
Actually, the mode usually just does The Right Thing without you having to think about it in this much detail. But when customizing indentation, it's helpful to understand the general indentation model being used.
As you configure @ccmode{}, you might want to set the variable
c-echo-syntactic-information-p to non-nil so that the
syntactic component list and calculated offset will always be echoed in
the minibuffer when you hit TAB.
@ccmode{} contains two minor-mode-like features that you should find useful while you enter new C code. The first is called auto-newline mode, and the second is called hungry-delete mode. These minor modes can be toggled on and off independently, and @ccmode{} can be configured so that it starts up with any combination of these minor modes. By default, both of these minor modes are turned off.
The state of the minor modes is always reflected in the minor mode list on the modeline of the @ccmode{} buffer. When auto-newline mode is enabled, you will see `C/a' on the mode line (6). When hungry delete mode is enabled you would see `C/h' and when both modes are enabled, you'd see `C/ah'.
@ccmode{} provides keybindings which allow you to toggle the minor
modes on the fly while editing code. To toggle just the auto-newline
state, hit C-c C-a (c-toggle-auto-state). When you do
this, you should see the `a' indicator either appear or disappear
on the modeline. Similarly, to toggle just the hungry-delete state, use
C-c C-d (c-toggle-hungry-state), and to toggle both states,
use C-c C-t (c-toggle-auto-hungry-state).
To set up the auto-newline and hungry-delete states to your preferred
values, you would need to add some lisp to your `.emacs' file that
called one of the c-toggle-*-state functions directly. When
called programmatically, each function takes a numeric value, where
a positive number enables the minor mode, a negative number disables the
mode, and zero toggles the current state of the mode.
So for example, if you wanted to enable both auto-newline and hungry-delete for all your C file editing, you could add the following to your `.emacs' file:
(add-hook 'c-mode-common-hook '(lambda () (c-toggle-auto-hungry-state 1)))
Auto-newline minor mode works by enabling certain electric commands. Electric commands are typically bound to special characters such as the left and right braces, colons, semi-colons, etc., which when typed, perform some magic formatting in addition to inserting the typed character. As a general rule, electric commands are only electric when the following conditions apply:
When you type either an open or close brace (i.e. { or }),
the electric command c-electric-brace gets run. This command has
two electric formatting behaviors. First, it will perform some
re-indentation of the line the brace was typed on, and second, it will
add various newlines before and/or after the typed brace.
Re-indentation occurs automatically whenever the electric behavior is
enabled. If the brace ends up on a line other than the one it was typed
on, then that line is also re-indented.
The insertion of newlines is controlled by the
c-hanging-braces-alist variable. This variable contains a
mapping between syntactic symbols related to braces, and a list of
places to insert a newline. The syntactic symbols that are useful for
this list are: class-open, class-close, defun-open,
defun-close, inline-open, inline-close,
brace-list-open, brace-list-close,
brace-list-intro, brace-list-entry, block-open,
block-close, substatement-open,
statement-case-open,
extern-lang-open, extern-lang-close,
namespace-open, and namespace-close.
See section Syntactic Symbols, for a more
detailed description of these syntactic symbols.
The value associated with each syntactic symbol in this association list is called an ACTION which can be either a function or a list. See section Custom Brace and Colon Hanging, for a more detailed discussion of using a function as a brace hanging ACTION.
When the ACTION is a list, it can contain any combination of the
symbols before and after, directing @ccmode{} where to
put newlines in relationship to the brace being inserted. Thus, if the
list contains only the symbol after, then the brace is said to
hang on the right side of the line, as in:
// here, open braces always `hang'
void spam( int i ) {
if( i == 7 ) {
dosomething(i);
}
}
When the list contains both after and before, the braces
will appear on a line by themselves, as shown by the close braces in the
above example. The list can also be empty, in which case no newlines
are added either before or after the brace.
For example, the default value of c-hanging-braces-alist is:
(defvar c-hanging-braces-alist '((brace-list-open)
(substatement-open after)
(block-close . c-snug-do-while)
(extern-lang-open after)))
which says that brace-list-open braces should both hang on the
right side, and allow subsequent text to follow on the same line as the
brace. Also, substatement-open and extern-lang-open
braces should hang on the right side, but subsequent text should follow
on the next line. Here, in the block-close entry, you also see
an example of using a function as an ACTION.
A word of caution: it is not a good idea to hang top-level construct
introducing braces, such as class-open or defun-open.
Emacs makes an assumption that such braces will always appear in column
zero, hanging such braces can introduce performance problems.
See section Performance Issues, for more information.
Using a mechanism similar to brace hanging (see section Hanging Braces),
colons can also be made to hang using the variable
c-hanging-colons-alist. The syntactic symbols appropriate for
this assocation list are: case-label, label,
access-label, member-init-intro, and inher-intro.
Note however that for c-hanging-colons-alist, ACTIONs as
functions are not supported. See also section Custom Brace and Colon Hanging for details.
In C++, double-colons are used as a scope operator but because these colons always appear right next to each other, newlines before and after them are controlled by a different mechanism, called clean-ups in @ccmode{}. See section Clean-ups, for details.
Semicolons and commas are also electric in @ccmode{}, but since these characters do not correspond directly to syntactic symbols, a different mechanism is used to determine whether newlines should be automatically inserted after these characters. See section Customizing Semi-colons and Commas, for details.
A few other keys also provide electric behavior. For example
# (c-electric-pound) is electric when typed as
the first non-whitespace character on a line. In this case, the
variable c-electric-pound-behavior is consulted for the electric
behavior. This variable takes a list value, although the only element
currently defined is alignleft, which tells this command to force
the `#' character into column zero. This is useful for entering
C preprocessor macro definitions.
Stars and slashes (i.e. * and /, c-electric-star and
c-electric-slash respectively) are also electric under
certain circumstances. If a star is inserted as the second character of
a C style block comment on a comment-only line, then the comment
delimiter is indented as defined by c-offsets-alist. A
comment-only line is defined as a line which contains only a comment, as
in:
void spam( int i )
{
// this is a comment-only line...
if( i == 7 ) // but this is not
{
dosomething(i);
}
}
Likewise, if a slash is inserted as the second slash in a C++ style line
comment (also only on a comment-only line), then the line is indented as
defined by c-offsets-alist.
Less-than and greater-than signs (c-electric-lt-gt) are also
electric, but only in C++ mode. Hitting the second of two < or
> keys re-indents the line if it is a C++ style stream operator.
Clean-ups are a mechanism complementary to colon and brace
hanging. On the surface, it would seem that clean-ups overlap the
functionality provided by the c-hanging-*-alist variables, and
similarly, clean-ups are only enabled when auto-newline minor mode is
enabled. Clean-ups are used however to adjust code "after-the-fact",
i.e. to eliminate some whitespace that is inserted by electric
commands, or whitespace that contains intervening constructs.
You can configure @ccmode{}'s clean-ups by setting the variable
c-cleanup-list, which is a list of clean-up symbols. By default,
@ccmode{} cleans up only the scope-operator construct, which
is necessary for proper C++ support. Note that clean-ups are only
performed when the construct does not occur within a literal (see
section Auto-newline insertion), and when there is nothing but whitespace
appearing between the individual components of the construct.
There are currently only five specific constructs that @ccmode{} can clean up, as indicated by these symbols:
brace-else-brace -- cleans up `} else {' constructs by
placing the entire construct on a single line. Clean-up occurs when the
open brace after the `else' is typed. So for example, this:
void spam(int i)
{
if( i==7 )
{
dosomething();
}
else
{
appears like this after the open brace is typed:
void spam(int i)
{
if( i==7 ) {
dosomething();
} else {
brace-elseif-brace -- similar to the brace-else-brace
clean-up, but this cleans up `} else if (...) {' constructs. For
example:
void spam(int i)
{
if( i==7 )
{
dosomething();
}
else if( i==3 ) {
appears like this after the open brace is typed:
void spam(int i)
{
if( i==7 ) {
dosomething();
} else if( i==3 ) {
empty-defun-braces -- cleans up braces following a top-level
function or class definition that contains no body. Clean up occurs
when the closing brace is typed. Thus the following:
class Spam
{
}
is transformed into this when the close brace is typed:
class Spam
{}
defun-close-semi -- cleans up the terminating semi-colon on
top-level function or class definitions when they follow a close
brace. Clean up occurs when the semi-colon is typed.
So for example, the following:
class Spam
{
}
;
is transformed into this when the semi-colon is typed:
class Spam
{
};
list-close-comma -- cleans up commas following braces in array
and aggregate initializers. Clean up occurs when the comma is typed.
scope-operator -- cleans up double colons which may designate a
C++ scope operator split across multiple lines(8). Clean up occurs when
the second colon is typed. You will always want scope-operator
in the c-cleanup-list when you are editing C++ code.
Hungry deletion of whitespace, or as it more commonly called, hungry-delete mode, is a simple feature that some people find extremely useful. In fact, you might find yourself wanting hungry-delete in all your editing modes!
In a nutshell, when hungry-delete mode is enabled, hitting the Backspace key(9) will consume all preceding whitespace, including newlines and tabs. This can really cut down on the number of Backspace's you have to type if, for example you made a mistake on the preceding line.
By default, when you hit the Backspace key
@ccmode{} runs the command c-electric-backspace, which deletes
text in the backwards direction. When deleting a single character, or
when Backspace is hit in a literal
(see section Auto-newline insertion),
or when hungry-delete mode is disabled, the function
contained in the c-backspace-function variable is called with one
argument (the number of characters to delete). This variable is set to
backward-delete-char-untabify by default.
Similarly, hitting the Delete key runs the command
c-electric-delete. When deleting a single character, or when
Delete is hit in a literal, or when hungry-delete mode is
disabled, the function contained in the c-delete-function
variable is called with one argument (the number of characters to
delete). This variable is set to delete-char by default.
However, if delete-key-deletes-forward is nil, or your
Emacs does not support separation of Backspace and DEL, then
c-electric-delete simply calls c-electric-backspace.
One other note about minor modes is worth mentioning here. CC Mode now
works much better with auto-fill mode (a standard Emacs minor mode) by
correctly auto-filling both line (e.g. C++ style) and block (e.g. C
style) oriented comments. When auto-fill-mode is enabled, line
oriented comments will also be auto-filled by inserting a newline at the
line break, and inserting `//' at the start of the next line.
When auto-filling block oriented comments, the behavior is dependent on
the value of the variable c-comment-continuation-stars. When
this variable is nil, the old behavior for auto-filling C
comments is in effect. In this case, the line is broken by closing the
comment and starting a new comment on the next line.
If you set c-comment-continuation-stars to a string, then a long
C block comment line is broken by inserting a newline at the line break
position, and inserting this string at the beginning of the next comment
line. The default value for c-comment-continuation-stars is
`* ' (a star followed by a single space)(10).
Various commands are provided which allow you to conveniently re-indent C constructs. There are several things to note about these indentation commands. First, when you change your programming style, either interactively or through some other means, your file does not automatically get re-indented. When you change style parameters, you will typically need to reformat the line, expression, or buffer to see the effects of your changes.
Second, changing some variables have no effect on existing code, even
when you do re-indent. For example, the c-hanging-* variables
and c-cleanup-list only affect new code as it is typed in
on-the-fly, so changing c-hanging-braces-alist and re-indenting
the buffer will not adjust placement of braces already in the file.
Third, re-indenting large portions of code is currently rather
inefficient. Improvements have been made since previous releases of
@ccmode{}, and much more radical improvements are planned, but for now
you need to be aware of this (11).
Some provision has been made to at least inform you as to the progress
of the re-indentation. The variable c-progress-interval controls
how often a progress message is displayed. Set this variable to
nil to inhibit progress messages, including messages normally
printed when indentation is started and completed.
Also, except as noted below, re-indentation is always driven by the same mechanisms that control on-the-fly indentation of code. See section New Indentation Engine, for details.
To indent a single line of code, use TAB
(c-indent-command). The behavior of this command is controlled
by the variable c-tab-always-indent. When this variable is
t, TAB always just indents the current line. When
nil, the line is indented only if point is at the left margin, or
on or before the first non-whitespace character on the line, otherwise
something else happens(12).
If the value of c-tab-always-indent is something other than
t or nil (e.g. 'other), then a real tab
character(13) is inserted only when point is
inside a literal (see section Auto-newline insertion), otherwise the line
is indented.
To indent an entire balanced brace or parenthesis expression, use
M-C-q (c-indent-exp). Note that point should be on
the opening brace or parenthesis of the expression you want to indent.
Another very convenient keystroke is C-c C-q
(c-indent-defun) when re-indents the entire top-level function or
class definition that encompasses point. It leaves point at the
same position within the buffer.
To indent any arbitrary region of code, use M-C-\
(indent-region). This is a standard Emacs command, specially
tailored for C code in a @ccmode{} buffer. Note that of course,
point and mark must delineate the region you
want to indent.
While not strictly an indentation function, M-C-h
(c-mark-function) is useful for marking the current top-level
function or class definition as the current region.
@ccmode{} contains other useful command for moving around in C code.
M-x c-beginning-of-defun
beginning-of-defun,
except it eliminates the constraint that the top-level opening brace
must be in column zero. See beginning-of-defun for more
information.
Depending on the coding style being used, you might prefer
c-beginning-of-defun to beginning-of-defun. If so,
consider binding C-M-a to the former instead. For backwards
compatibility reasons, the default binding remains in effect.
M-x c-end-of-defun
end-of-defun,
except it eliminates the constraint that the top-level opening brace of
the defun must be in column zero. See beginning-of-defun for more
information.
Depending on the coding style being used, you might prefer
c-end-of-defun to end-of-defun. If so,
consider binding C-M-e to the former instead. For backwards
compatibility reasons, the default binding remains in effect.
C-c C-u (c-up-conditional)
#elif is treated
like #else followed by #if. When going forwards,
#elif is ignored.
C-c C-p (c-backward-conditional)
C-c C-n (c-forward-conditional)
M-a (c-beginning-of-statement)
nil) or sentence (if non-nil).
M-e (c-end-of-statement)
nil) or sentence (if non-nil).
M-x c-forward-into-nomenclature
M-x c-backward-into-nomenclature
C-c : (c-scope-operator)
M-q (fill-paragraph)
c-hanging-comment-starter-p controls whether comment
start delimiters which appear on a line by themselves, end up on a line
by themselves after the fill. When the value is nil, the comment
starter will remain on its own line(15). Otherwise,
text on the next line will be put on the same line as the comment
starter. This is called hanging because the following text hangs
on the line with the comment starter(16)
The variable c-hanging-comment-ender-p controls the analogous
behavior for the block comment end delimiter. When the value is
nil, the comment ender will remain on its own line after the
file(17). Otherwise, the
comment end delimiter will be placed at the end of the previous line.
The variable c-offsets-alist contains the mappings between
syntactic symbols and the offsets to apply for those symbols. You
should never modify this variable directly though. Use the function
c-set-offset instead (see below for details).
The c-offsets-alist variable is where you customize all your
indentations. You simply need to decide what additional offset you want
to add for every syntactic symbol. You can use the command C-c
C-o (c-set-offset) as the way to set offsets, both interactively
and from your mode hook. Also, you can set up styles of
indentatio. Most likely, you'll
find one of the pre-defined styles will suit your needs, but if not,
this section will describe how to set up basic editing configurations.
See section Styles, for an explanation of how to set up named styles.
As mentioned previously, the variable c-offsets-alist is an
association list of syntactic symbols and the offsets to be applied for
those symbols. In fact, these offset values can be any of an integer, a
function or lambda expression, a variable name, or one of the following
symbols: +, -, ++, --, *, or
/. These symbols describe offset in multiples of the value of
the variable c-basic-offset. By defining a style's indentation
in terms of this fundamental variable, you can change the amount of
whitespace given to an indentation level while leaving the same
relationship between levels. Here are the values that the special
symbols correspond to:
+
c-basic-offset times 1
-
c-basic-offset times -1
++
c-basic-offset times 2
--
c-basic-offset times -2
*
c-basic-offset times 0.5
/
c-basic-offset times -0.5
So, for example, because most of the default offsets are defined in
terms of +, -, and 0, if you like the general
indentation style, but you use 4 spaces instead of 2 spaces per level,
you can probably achieve your style just by changing
c-basic-offset like so (in your `.emacs' file):
(setq c-basic-offset 4)
This would change
int add( int val, int incr, int doit )
{
if( doit )
{
return( val + incr );
}
return( val );
}
to
int add( int val, int incr, int doit )
{
if( doit )
{
return( val + incr );
}
return( val );
}
To change indentation styles more radically, you will want to change the
value associated with the syntactic symbols in the
c-offsets-alist variable. First, I'll show you how to do that
interactively, then I'll describe how to make changes to your
`.emacs' file so that your changes are more permanent.
As an example of how to customize indentation, let's change the style of this example(18):
1: int add( int val, int incr, int doit )
2: {
3: if( doit )
4: {
5: return( val + incr );
6: }
7: return( val );
8: }
to:
1: int add( int val, int incr, int doit )
2: {
3: if( doit )
4: {
5: return( val + incr );
6: }
7: return( val );
8: }
In other words, we want to change the indentation of braces that open a block following a condition so that the braces line up under the conditional, instead of being indented. Notice that the construct we want to change starts on line 4. To change the indentation of a line, we need to see which syntactic components affect the offset calculations for that line. Hitting C-c C-s on line 4 yields:
((substatement-open . 44))
so we know that to change the offset of the open brace, we need to
change the indentation for the substatement-open syntactic
symbol. To do this interactively, just hit C-c C-o
(c-set-offset). This prompts you for the syntactic symbol to
change, providing a reasonable default. In this case, the default is
substatement-open, which is just the syntactic symbol we want to
change!
After you hit return, @ccmode{} will then prompt you for the new
offset value, with the old value as the default. The default in this
case is `+', but we want no extra indentation so enter
`0' and RET. This will associate the offset 0 with the
syntactic symbol substatement-open in the c-offsets-alist
variable.
To check your changes quickly, just hit C-c C-q
(c-indent-defun) to reindent the entire function. The example
should now look like:
1: int add( int val, int incr, int doit )
2: {
3: if( doit )
4: {
5: return( val + incr );
6: }
7: return( val );
8: }
Notice how just changing the open brace offset on line 4 is all we needed to do. Since the other affected lines are indented relative to line 4, they are automatically indented the way you'd expect. For more complicated examples, this may not always work. The general approach to take is to always start adjusting offsets for lines higher up in the file, then re-indent and see if any following lines need further adjustments.
To make your changes permanent, you need to add some lisp code to your `.emacs' file, but first you need to decide whether your styles should be global in every buffer, or local to each specific buffer.
If you edit primarily one style of code, you may want to make the @ccmode{} style variables have global values so that every buffer will share the style settings. This will allow you to set the @ccmode{} variables at the top level of your `.emacs' file, and is the way @ccmode{} works by default.
If you edit many different styles of code at
the same time, you might want to make the @ccmode{} style variables
have buffer local values. If you do this, then you will need to set any
@ccmode{} style variables in a hook function (e.g. off of
c-mode-common-hook instead of at the top level of your
`.emacs' file). The recommended way to do this is to set the
variable c-style-variables-are-local-p to t
before @ccmode{} is loaded into your Emacs session.
@ccmode{} provides several hooks that you can use to customize the mode according to your coding style. Each language mode has its own hook, adhering to standard Emacs major mode conventions. There is also one general hook and one package initialization hook:
c-mode-hook -- for C buffers only
c++-mode-hook -- for C++ buffers only
objc-mode-hook -- for Objective-C buffers only
java-mode-hook -- for Java buffers only
idl-mode-hook -- for IDL buffers only
c-mode-common-hook -- common across all languages
c-initialization-hook -- hook run only once per Emacs session,
when @ccmode{} is initialized.
The language hooks get run as the last thing when you enter that
language mode. The c-mode-common-hook is run by all
supported modes before the language specific hook, and thus can
contain customizations that are common across all languages. Most of
the examples in this section will assume you are using the common
hook(19).
Here's a simplified example of what you can add to your `.emacs' file to make the changes described in the previous section (section Interactive Customization) more permanent. See the Emacs manuals for more information on customizing Emacs via hooks. See section Sample .emacs file, for a more complete sample `.emacs' file.
(defun my-c-mode-common-hook () ;; my customizations for all of c-mode and related modes (c-set-offset 'substatement-open 0) ;; other customizations can go here ) (add-hook 'c-mode-common-hook 'my-c-mode-common-hook)
For complex customizations, you will probably want to set up a style that groups all your customizations under a single name.
Most people only need to edit code formatted in just a few well-defined and consistent styles. For example, their organization might impose a "blessed" style that all its programmers must conform to. Similarly, people who work on GNU software will have to use the GNU coding style on C code. Some shops are more lenient, allowing a variety of coding styles, and as programmers come and go, there could be a number of styles in use. For this reason, @ccmode{} makes it convenient for you to set up logical groupings of customizations called styles, associate a single name for any particular style, and pretty easily start editing new or existing code using these styles.
If you're lucky, one of @ccmode{}'s built-in styles might be just what you're looking for. These include:
gnu -- coding style blessed by the Free Software Foundation
for C code in GNU programs. This is the default style for all newly
created buffers, but you can change this by setting the variable
c-default-style.
k&r -- The classic Kernighan and Ritchie style for C code.
bsd -- Also known as "Allman style" after Eric Allman.
whitesmith -- Popularized by the examples that came with
Whitesmiths C, an early commercial C compiler.
stroustrup -- The classic Stroustrup style for C++ code.
ellemtel -- Popular C++ coding standards as defined by
"Programming in C++, Rules and Recommendations", Erik Nyquist and Mats
Henricson, Ellemtel (20).
linux -- C coding standard for Linux development.
python -- C coding standard for Python extension
modules(21).
java -- The style for editing Java code. Note that this style is
automatically installed when you enter java-mode.
user -- This is a special style for several reasons. First, if
you customize @ccmode{} by using either the new Custom interface or by
doing setq's at the top level of your `.emacs' file, these
settings will be captured in the user style. Also, all other
styles implicitly inherit their settings from user style. This
means that for any styles you add via c-add-style (see section Adding Styles) you need only define the differences between your new style and
user style.
Note however that user style is not the default style.
gnu is the default style for all newly created buffers, but you
can change this by setting variable c-default-style. Be careful
if you customize @ccmode{} as described above; since your changes will
be captured in the user style, you will also have to change
c-default-style to "user" to see the effect of your
customizations.
If you'd like to experiment with these built-in styles you can simply type the following in a @ccmode{} buffer:
C-c . STYLE-NAME RET
C-c . runs the command c-set-style. Note that all style
names are case insensitive, even the ones you define.
Setting a style in this way does not automatically re-indent your file. For commands that you can use to view the effect of your changes, see section Commands.
Once you find a built-in style you like, you can make the change permanent by adding some lisp to your `.emacs' file. Let's say for example that you want to use the `ellemtel' style in all your files. You would add this:
(defun my-c-mode-common-hook () ;; use Ellemtel style for all C like languages (c-set-style "ellemtel") ;; other customizations can go here ) (add-hook 'c-mode-common-hook 'my-c-mode-common-hook)
Note that for BOCM compatibility, `gnu' is the default
style, and any non-style based customizations you make (i.e. in
c-mode-common-hook in your
`.emacs' file) will be based on `gnu' style unless you do
a c-set-style as the first thing in your hook. The variable
c-indentation-style always contains the buffer's current style name,
as a string.
If none of the built-in styles is appropriate, you'll probably want to
add a new style definition. Styles are kept in the
c-style-alist variable, but you should never modify this variable
directly. Instead, @ccmode{} provides the function
c-add-style that you can use to easily add new styles or change
existing styles. This function takes two arguments, a stylename
string, and an association list description of style
customizations. If stylename is not already in
c-style-alist, the new style is added, otherwise the style is
changed to the new description.
This function also takes an optional third argument, which if
non-nil, automatically applies the new style to the current
buffer.
The sample `.emacs' file provides a concrete example of how a new style can be added and automatically set. See section Sample .emacs file.
The Emacs manual describes how you can customize certain variables on a
per-file basis by including a Local Variable block at the end of
the file. So far, you've only seen a functional interface to @ccmode{}
customization, which is highly inconvenient for use in a Local Variable
block. @ccmode{} provides two variables that make it easier for you to
customize your style on a per-file basis.
It works via the standard Emacs hook variable
hack-local-variables-hook.
The variable c-file-style can be set to a style name string.
When the file is visited, @ccmode{} will automatically set the
file's style to this style using c-set-style.
Another variable, c-file-offsets, takes an association list
similar to what is allowed in c-offsets-alist. When the file is
visited, @ccmode{} will automatically institute these offets using
c-set-offset.
Note that file style settings (i.e. c-file-style) are applied
before file offset settings (i.e. c-file-offsets). Also, if
either of these are set in a file's local variable section, all the
style variable values are made local to that buffer.
For most users, @ccmode{} will support their coding styles with
very little need for more advanced customizations. Usually, one of the
standard styles defined in c-style-alist will do the trick. At
most, perhaps one of the syntactic symbol offsets will need to be
tweaked slightly, or maybe c-basic-offset will need to be
changed. However, some styles require a more flexible framework for
customization, and one of the real strengths of @ccmode{} is that
the syntactic analysis model provides just such a framework. This allows
you to implement custom indentation calculations for situations not
handled by the mode directly.
Note that the style controlling variables can either have global values,
or can be buffer local (e.g. different in every buffer). If all the C
files you edit tend to have the same style, you might want to keep the
variables global. If you tend to edit files with many different styles,
you will have to make the variables buffer local. The variable
c-style-variables-are-local-p controls this.
When c-style-variables-are-local-p is non-nil, then the style
variables will have a different settable value for each buffer,
otherwise all buffers will share the same values. By default, its value
is nil (i.e. global values). You must set this variable
before @ccmode{} is loaded into your Emacs session, and once the
variables are made buffer local, they cannot be made global again
(unless you restart Emacs of course!)
The most flexible way to customize @ccmode{} is by writing @dfn{custom indentation functions} and associating them with specific syntactic symbols (see section Syntactic Symbols). @ccmode{} itself uses custom indentation functions to provide more sophisticated indentation, for example when lining up C++ stream operator blocks:
1: void main(int argc, char**)
2: {
3: cout << "There were "
4: << argc
5: << "arguments passed to the program"
6: << endl;
7: }
In this example, lines 4 through 6 are assigned the stream-op
syntactic symbol. Here, stream-op has an offset of +, and
with a c-basic-offset of 2, you can see that lines 4 through 6
are simply indented two spaces to the right of line 3. But perhaps we'd
like @ccmode{} to be a little more intelligent so that it aligns
all the `<<' symbols in lines 3 through 6. To do this, we have
to write a custom indentation function which finds the column of first
stream operator on the first line of the statement. Here is sample
lisp code implementing this:
(defun c-lineup-streamop (langelem)
;; lineup stream operators
(save-excursion
(let* ((relpos (cdr langelem))
(curcol (progn (goto-char relpos)
(current-column))))
(re-search-forward "<<\\|>>" (c-point 'eol) 'move)
(goto-char (match-beginning 0))
(- (current-column) curcol))))
Custom indent functions take a single argument, which is a syntactic component cons cell (see section Syntactic Analysis). The function returns an integer offset value that will be added to the running total indentation for the line. Note that what actually gets returned is the difference between the column that the first stream operator is on, and the column of the buffer relative position passed in the function's argument. Remember that @ccmode{} automatically adds in the column of the component's relative buffer position and we don't the column offset added in twice.
Now, to associate the function c-lineup-streamop with the
stream-op syntactic symbol, we can add something like the
following to our c++-mode-hook(22):
(c-set-offset 'stream-op 'c-lineup-streamop)
Now the function looks like this after re-indenting (using C-c C-q):
1: void main(int argc, char**)
2: {
3: cout << "There were "
4: << argc
5: << "arguments passed to the program"
6: << endl;
7: }
Custom indentation functions can be as simple or as complex as you like,
and any syntactic symbol that appears in c-offsets-alist can have
a custom indentation function associated with it. @ccmode{} comes
with several standard custom indentation functions, not all of which are
used by the default styles.
c-lineup-arglist -- lines up function argument lines under the
argument on the previous line.
c-lineup-arglist-intro-after-paren -- similar to
c-lineup-arglist, but works for argument lists that begin with an
open parenthesis followed by a newline.
c-lineup-arglist-close-under-paren -- set your
arglist-close syntactic symbol to this line-up function so that
parentheses that close argument lists will line up under the parenthesis
that opened the argument list.
c-lineup-close-paren -- lines up the closing parenthesis under
its corresponding open parenthesis if that one is followed by code.
Otherwise, if the open parenthesis ends its line, no indentation is
added. Works with any ...-close symbol.
c-lineup-streamop -- lines up C++ stream operators
(e.g. `<<' and `>>').
c-lineup-multi-inher -- lines up multiple inheritance lines.
c-indent-one-line-block -- adds c-basic-offset to the
indentation if the line is a one line block, otherwise 0. Intended to
be used with any opening brace symbol, e.g. substatement-open.
c-lineup-C-comments -- lines up C block comment continuation
lines.
c-lineup-comment -- lines up comment only lines according to
the variable c-comment-only-line-offset.
c-lineup-runin-statements -- lines up statements for coding
standards which place the first statement in a block on the same line as
the block opening brace(23).
c-lineup-math -- lines up math statement-cont lines under
the previous line after the equals sign.
c-lineup-ObjC-method-call -- for Objective-C code, lines up
selector arguments just after the message receiver.
c-lineup-ObjC-method-args -- for Objective-C code, lines up the
colons that separate arguments by aligning colons vertically.
c-lineup-ObjC-method-args-2 -- similar to
c-lineup-ObjC-method-args but lines up the colon on the current
line with the colon on the previous line.
c-lineup-dont-change -- this lineup function returns the
indentation of the current line. Think of it as an identity function
for lineups; it is used for cpp-macro-cont lines.
Syntactic symbols aren't the only place where you can customize
@ccmode{} with the lisp equivalent of callback functions. Brace
"hanginess" can also be determined by custom functions associated with
syntactic symbols on the c-hanging-braces-alist variable.
Remember that ACTION's are typically a list containing some
combination of the symbols before and after (see
section Hanging Braces). However, an ACTION can also be a function
which gets called when a brace matching that syntactic symbol is
entered.
These ACTION functions are called with two arguments: the
syntactic symbol for the brace, and the buffer position at which the
brace was inserted. The ACTION function is expected to return a
list containing some combination of before and after. The
function can also return nil. This return value has the normal
brace hanging semantics.
As an example, @ccmode{} itself uses this feature to dynamically determine the hanginess of braces which close "do-while" constructs:
void do_list( int count, char** atleast_one_string )
{
int i=0;
do {
handle_string( atleast_one_string[i] );
i++;
} while( i < count );
}
@ccmode{} assigns the block-close syntactic symbol to the
brace that closes the do construct, and normally we'd like the
line that follows a block-close brace to begin on a separate
line. However, with "do-while" constructs, we want the
while clause to follow the closing brace. To do this, we
associate the block-close symbol with the ACTION function
c-snug-do-while:
(defun c-snug-do-while (syntax pos)
"Dynamically calculate brace hanginess for do-while statements.
Using this function, `while' clauses that end a `do-while' block will
remain on the same line as the brace that closes that block.
See `c-hanging-braces-alist' for how to utilize this function as an
ACTION associated with `block-close' syntax."
(save-excursion
(let (langelem)
(if (and (eq syntax 'block-close)
(setq langelem (assq 'block-close c-syntactic-context))
(progn (goto-char (cdr langelem))
(if (= (following-char) ?{)
(forward-sexp -1))
(looking-at "\\<do\\>[^_]")))
'(before)
'(before after)))))
This function simply looks to see if the brace closes a "do-while" clause and if so, returns the list `(before)' indicating that a newline should be inserted before the brace, but not after it. In all other cases, it returns the list `(before after)' so that the brace appears on a line by itself.
During the call to the brace hanging ACTION function, the variable
c-syntactic-context is bound to the full syntactic analysis list.
Note that for symmetry, colon hanginess should be customizable by
allowing function symbols as ACTIONs on the
c-hanging-colon-alist variable. Since no use has actually been
found for this feature, it isn't currently implemented!
You can also customize the insertion of newlines after semi-colons and
commas, when the auto-newline minor mode is enabled (see section Minor Modes). This is controlled by the variable
c-hanging-semi&comma-criteria, which contains a list of functions
that are called in the order they appear. Each function is called with
zero arguments, and is expected to return one of the following values:
nil -- A newline is inserted, and no more functions from the
list are called.
stop -- No more functions from the list are called, but no
newline is inserted.
nil -- No determination is made, and the next function in the
list is called.
If every function in the list is called without a determination being
made, then no newline is added. The default value for this variable is a
list containing a single function which inserts newlines only after
semi-colons which do not appear inside parenthesis lists (i.e. those
that separate for-clause statements).
Here's an example of a criteria function, provided by @ccmode{}, that
will prevent newlines from being inserted after semicolons when there is
a non-blank following line. Otherwise, it makes no determination. To
use, add this to the front of the c-hanging-semi&comma-criteria
list.
(defun c-semi&comma-no-newlines-before-nonblanks ()
(save-excursion
(if (and (eq last-command-char ?\;)
(zerop (forward-line 1))
(not (looking-at "^[ \t]*$")))
'stop
nil)))
The default value of c-hanging-semi&comma-criteria is a list
containing just the function c-semi&comma-inside-parenlist, which
suppresses newlines after semicolons inside parenthesis lists
(e.g. for-loops). In addition to
c-semi&comma-no-newlines-before-nonblanks described above,
@ccmode{} also comes with the criteria function
c-semi&comma-no-newlines-for-oneline-inliners, which suppresses
newlines after semicolons inside one-line inline method definitions
(i.e. in C++ or Java).
In `gnu' style (see section Built-in Styles), a minimum indentation
is imposed on lines inside top-level constructs. This minimum
indentation is controlled by the variable
c-label-minimum-indentation. The default value for this variable
is 1.
One other customization variable is available in @ccmode{}:
c-special-indent-hook. This is a standard hook variable that is
called after every line is indented by @ccmode{}. You can use it
to do any special indentation or line adjustments your style dictates,
such as adding extra indentation to constructors or destructor
declarations in a class definition, etc. Note however, that you should
not change point or mark inside your c-special-indent-hook
functions (i.e. you'll probably want to wrap your function in a
save-excursion).
Setting c-special-indent-hook in your style definition is handled
slightly differently than other variables. In your style definition,
you should set the value for
c-special-indent-hook to a function or list of functions, which
will be appended to c-special-indent-hook using add-hook.
That way, the current setting for the buffer local value of
c-special-indent-hook won't be overridden.
Normally, the standard Emacs command M-;
(indent-for-comment) will indent comment only lines to
comment-column. Some users however, prefer that M-; act
just like TAB for purposes of indenting comment-only lines;
i.e. they want the comments to always indent as they would for normal
code, regardless of whether TAB or M-; were used. This
behavior is controlled by the variable
c-indent-comments-syntactically-p. When nil (the
default), M-; indents comment-only lines to comment-column,
otherwise, they are indented just as they would be if TAB were
typed.
Here is a complete list of the recognized syntactic symbols as described
in the c-offsets-alist variable, along with a brief description.
More detailed descriptions follow below.
string -- inside multi-line string
c -- inside a multi-line C style block comment
defun-open -- brace that opens a function definition
defun-close -- brace that closes a function definition
defun-block-intro -- the first line in a top-level defun
class-open -- brace that opens a class definition
class-close -- brace that closes a class definition
inline-open -- brace that opens an in-class inline method
inline-close -- brace that closes an in-class inline method
func-decl-cont -- the region between a function definition's
argument list and the function opening brace (excluding K&R argument
declarations). In C, you cannot put anything but whitespace and comments
between them; in C++ and Java, throws declarations and other
things can appear in this context.
knr-argdecl-intro -- first line of a K&R C argument declaration
knr-argdecl -- subsequent lines in a K&R C argument declaration
topmost-intro -- the first line in a topmost definition
topmost-intro-cont -- topmost definition continuation lines
member-init-intro -- first line in a member initialization list
member-init-cont -- subsequent member initialization list lines
inher-intro -- first line of a multiple inheritance list
inher-cont -- subsequent multiple inheritance lines
block-open -- statement block open brace
block-close -- statement block close brace
brace-list-open -- open brace of an enum or static array list
brace-list-close -- close brace of an enum or static array list
brace-list-intro -- first line in an enum or static array list
brace-list-entry -- subsequent lines in an enum or static array list
statement -- a C statement
statement-cont -- a continuation of a C statement
statement-block-intro -- the first line in a new statement block
statement-case-intro -- the first line in a case `block'
statement-case-open -- the first line in a case block starting
with brace
substatement -- the first line after a conditional
substatement-open -- the brace that opens a substatement block
case-label -- a case or default label
access-label -- C++ access control label
label -- any non-special C label
do-while-closure -- the `while' that ends a
do-while construct
else-clause -- the `else' of an if-else construct
comment-intro -- a line containing only a comment introduction
arglist-intro -- the first line in an argument list
arglist-cont -- subsequent argument list lines when no arguments
follow on the same line as the the arglist opening paren
arglist-cont-nonempty -- subsequent argument list lines when at
least one argument follows on the same line as the arglist opening paren
arglist-close -- the solo close paren of an argument list
stream-op -- lines continuing a stream operator
inclass -- the line is nested inside a class definition
cpp-macro -- the start of a C preprocessor macro definition
cpp-macro-cont -- subsequent lines of a multi-line C
preprocessor macro definition
friend -- a C++ friend declaration
objc-method-intro -- the first line of an Objective-C method definition
objc-method-args-cont -- lines continuing an Objective-C method
definition
objc-method-call-cont -- lines continuing an Objective-C method call
extern-lang-open -- brace that opens an external language block
extern-lang-close -- brace that closes an external language block
inextern-lang -- analogous to `inclass' syntactic symbol, but
used inside external language blocks (e.g. extern "C" {).
namespace-open -- brace that opens a C++ namespace block.
namespace-close -- brace that closes a C++ namespace block.
innamespace -- analogous to `inextern-lang' syntactic symbol,
but used inside C++ namespace blocks.
template-args-cont -- C++ template argument list continuations
Most syntactic symbol names follow a general naming convention. When a
line begins with an open or close brace, the syntactic symbol will
contain the suffix -open or -close respectively.
Usually, a distinction is made between the first line that introduces a
construct and lines that continue a construct, and the syntactic symbols
that represent these lines will contain the suffix -intro or
-cont respectively. As a sub-classification of this scheme, a
line which is the first of a particular brace block construct will
contain the suffix -block-intro.
Let's look at some examples to understand how this works. Remember that you can check the syntax of any line by using C-c C-s.
1: void
2: swap( int& a, int& b )
3: {
4: int tmp = a;
5: a = b;
6: b = tmp;
7: int ignored =
8: a + b;
9: }
Line 1 shows a topmost-intro since it is the first line that
introduces a top-level construct. Line 2 is a continuation of the
top-level construct introduction so it has the syntax
topmost-intro-cont. Line 3 shows a defun-open since it is
the brace that opens a top-level function definition. Line 9 is a
defun-close since it contains the brace that closes the top-level
function definition. Line 4 is a defun-block-intro, i.e. it is
the first line of a brace-block, enclosed in a
top-level function definition.
Lines 5, 6, and 7 are all given statement syntax since there
isn't much special about them. Note however that line 8 is given
statement-cont syntax since it continues the statement begun
on the previous line.
Here's another example, which illustrates some C++ class syntactic symbols:
1: class Bass
2: : public Guitar,
3: public Amplifiable
4: {
5: public:
6: Bass()
7: : eString( new BassString( 0.105 )),
8: aString( new BassString( 0.085 )),
9: dString( new BassString( 0.065 )),
10: gString( new BassString( 0.045 ))
11: {
12: eString.tune( 'E' );
13: aString.tune( 'A' );
14: dString.tune( 'D' );
15: gString.tune( 'G' );
16: }
17: friend class Luthier;
18: }
As in the previous example, line 1 has the topmost-intro syntax.
Here however, the brace that opens a C++ class definition on line 4 is
assigned the class-open syntax. Note that in C++, classes,
structs, and unions are essentially equivalent syntactically (and are
very similar semantically), so replacing the class keyword in the
example above with struct or union would still result in a
syntax of class-open for line 4 (24).
Similarly, line 18 is assigned class-close syntax.
Line 2 introduces the inheritance list for the class so it is assigned
the inher-intro syntax, and line 3, which continues the
inheritance list is given inher-cont syntax.
Hitting C-c C-s on line 5 shows the following analysis:
((inclass . 1) (access-label . 67))
The primary syntactic symbol for this line is access-label as
this a label keyword that specifies access protection in C++. However,
because this line is also a top-level construct inside a class
definition, the analysis actually shows two syntactic symbols. The
other syntactic symbol assigned to this line is inclass.
Similarly, line 6 is given both inclass and topmost-intro
syntax:
((inclass . 58) (topmost-intro . 60))
Line 7 introduces a C++ member initialization list and as such is given
member-init-intro syntax. Note that in this case it is
not assigned inclass since this is not considered a
top-level construct. Lines 8 through 10 are all assigned
member-init-cont since they continue the member initialization
list started on line 7.
Line 11's analysis is a bit more complicated:
((inclass . 1) (inline-open))
This line is assigned a syntax of both inline-open and
inclass because it opens an in-class C++ inline method
definition. This is distinct from, but related to, the C++ notion of an
inline function in that its definition occurs inside an enclosing class
definition, which in C++ implies that the function should be inlined.
If though, the definition of the Bass constructor appeared
outside the class definition, the construct would be given the
defun-open syntax, even if the keyword inline appeared
before the method name, as in:
class Bass
: public Guitar,
public Amplifiable
{
public:
Bass();
}
inline
Bass::Bass()
: eString( new BassString( 0.105 )),
aString( new BassString( 0.085 )),
dString( new BassString( 0.065 )),
gString( new BassString( 0.045 ))
{
eString.tune( 'E' );
aString.tune( 'A' );
dString.tune( 'D' );
gString.tune( 'G' );
}
Returning to the previous example, line 16 is given inline-close
syntax, while line 12 is given defun-block-open syntax, and lines
13 through 15 are all given statement syntax. Line 17 is
interesting in that its syntactic analysis list contains three
elements:
((friend) (inclass . 58) (topmost-intro . 380))
The friend syntactic symbol is a modifier that typically does not
have a relative buffer position.
Template definitions introduce yet another syntactic symbol:
1: ThingManager <int, 2: Framework::Callback *, 3: Mutex> framework_callbacks;
Here, line 1 is analyzed as a topmost-intro, but lines 2 and 3
are both analyzed as template-args-cont lines.
Here is another (totally contrived) example which illustrates how syntax is assigned to various conditional constructs:
1: void spam( int index )
2: {
3: for( int i=0; i<index; i++ )
4: {
5: if( i == 10 )
6: {
7: do_something_special();
8: }
9: else
10: do_something( i );
11: }
12: do {
13: another_thing( i-- );
14: }
15: while( i > 0 );
16: }
Only the lines that illustrate new syntactic symbols will be discussed.
Line 4 has a brace which opens a conditional's substatement block. It
is thus assigned substatement-open syntax, and since line 5 is
the first line in the substatement block, it is assigned
substatement-block-intro syntax. Lines 6 and 7 are assigned
similar syntax. Line 8 contains the brace that closes the inner
substatement block. It is given the syntax block-close,
as are lines 11 and 14.
Line 9 is a little different -- since it contains the keyword
else matching the if statement introduced on line 5, it is
given the else-clause syntax. Note also that line 10 is slightly
different too. Because else is considered a conditional
introducing keyword (25), and because
the following substatement is not a brace block, line 10 is assigned the
substatement syntax.
One other difference is seen on line 15. The while construct
that closes a do conditional is given the special syntax
do-while-closure if it appears on a line by itself. Note that if
the while appeared on the same line as the preceding close brace,
that line would have been assigned block-close syntax instead.
Switch statements have their own set of syntactic symbols. Here's an example:
1: void spam( enum Ingredient i )
2: {
3: switch( i ) {
4: case Ham:
5: be_a_pig();
6: break;
7: case Salt:
8: drink_some_water();
9: break;
10: default:
11: {
12: what_is_it();
13: break;
14: }
15: }
14: }
Here, lines 4, 7, and 10 are all assigned case-label syntax,
while lines 5 and 8 are assigned statement-case-intro. Line 11
is treated slightly differently since it contains a brace that opens a
block -- it is given statement-case-open syntax.
There are a set of syntactic symbols that are used to recognize
constructs inside of brace lists. A brace list is defined as an
enum or aggregate initializer list, such as might statically
initialize an array of structs. For example:
1: static char* ingredients[] =
2: {
3: "Ham",
4: "Salt",
5: NULL
6: }
Following convention, line 2 in this example is assigned
brace-list-open syntax, and line 3 is assigned
brace-list-intro syntax. Likewise, line 6 is assigned
brace-list-close syntax. Lines 4 and 5 however, are assigned
brace-list-entry syntax, as would all subsequent lines in this
initializer list.
External language definition blocks also have their own syntactic symbols. In this example:
1: extern "C"
2: {
3: int thing_one( int );
4: int thing_two( double );
5: }
line 2 is given the extern-lang-open syntax, while line 5 is given
the extern-lang-close syntax. The analysis for line 3 yields:
((inextern-lang) (topmost-intro . 14)), where
inextern-lang is a modifier similar in purpose to inclass.
Similarly, C++ namespace constructs have their own associated syntactic symbols. In this example:
1: namespace foo
2: {
3: void xxx() {}
4: }
line 2 is given the namespace-open syntax, while line 4 is given
the namespace-close syntax. The analysis for line 3 yields:
((innamespace) (topmost-intro . 17)), where innamespace is
a modifier similar in purpose to inextern-lang and inclass.
A number of syntactic symbols are associated with parenthesis lists, a.k.a argument lists, as found in function declarations and function calls. This example illustrates these:
1: void a_function( int line1,
2: int line2 );
3:
4: void a_longer_function(
5: int line1,
6: int line2
7: );
8:
9: void call_them( int line1, int line2 )
10: {
11: a_function(
12: line1,
13: line2
14: );
15:
16: a_longer_function( line1,
17: line2 );
18: }
Lines 5 and 12 are assigned arglist-intro syntax since they are
the first line following the open parenthesis, and lines 7 and 14 are
assigned arglist-close syntax since they contain the parenthesis
that closes the argument list.
Lines that continue argument lists can be assigned one of two syntactic
symbols. For example, Lines 2 and 17
are assigned arglist-cont-nonempty syntax. What this means
is that they continue an argument list, but that the line containing the
parenthesis that opens the list is not empty following the open
parenthesis. Contrast this against lines 6 and 13 which are assigned
arglist-cont syntax. This is because the parenthesis that opens
their argument lists is the last character on that line.
Note that there is no arglist-open syntax. This is because any
parenthesis that opens an argument list, appearing on a separate line,
is assigned the statement-cont syntax instead.
A few miscellaneous syntactic symbols that haven't been previously covered are illustrated by this C++ example:
1: void Bass::play( int volume )
2: const
3: {
4: /* this line starts a multi-line
5: * comment. This line should get `c' syntax */
6:
7: char* a_multiline_string = "This line starts a multi-line \
8: string. This line should get `string' syntax.";
9:
10: note:
11: {
12: #ifdef LOCK
13: Lock acquire();
14: #endif // LOCK
15: slap_pop();
16: cout << "I played "
17: << "a note\n";
18: }
19: }
The lines to note in this example include:
func-decl-cont syntax;
defun-block-intro and
comment-intro syntax;
c syntax;
defun-block-intro. Note that the appearance of the
comment on lines 4 and 5 do not cause line 6 to be assigned
statement syntax because comments are considered to be
syntactic whitespace, which are ignored when analyzing
code;
string syntax;
label syntax;
block-open syntax;
cpp-macro syntax.
stream-op syntax.
Multi-line C preprocessor macros are now (somewhat) supported. At least CC Mode now recognizes the fact that it is inside a multi-line macro, and it properly skips such macros as syntactic whitespace. In this example:
1: #define LIST_LOOP(cons, listp) \
2: for (cons = listp; !NILP (cons); cons = XCDR (cons)) \
3: if (!CONSP (cons)) \
4: signal_error ("Invalid list format", listp); \
5: else
line 1 is given the syntactic symbol cpp-macro. This first line
of a macro is always given this symbol. The second and subsequent lines
(e.g. lines 2 through 5) are given the cpp-macro-cont syntactic
symbol, with a relative buffer position pointing to the # which
starts the macro definition.
In Objective-C buffers, there are three additional syntactic symbols assigned to various message calling constructs. Here's an example illustrating these:
1: - (void)setDelegate:anObject
2: withStuff:stuff
3: {
4: [delegate masterWillRebind:self
5: toDelegate:anObject
6: withExtraStuff:stuff];
7: }
Here, line 1 is assigned objc-method-intro syntax, and line 2 is
assigned objc-method-args-cont syntax. Lines 5 and 6 are both
assigned objc-method-call-cont syntax.
Two other syntactic symbols can appear in old style, non-prototyped C code (26):
1: int add_three_integers(a, b, c)
2: int a;
3: int b;
4: int c;
5: {
6: return a + b + c;
7: }
Here, line 2 is the first line in an argument declaration list and so is
given the knr-argdecl-intro syntactic symbol. Subsequent lines
(i.e. lines 3 and 4 in this example), are given knr-argdecl
syntax.
C and its derivative languages are highly complex creatures. Often, ambiguous code situations arise that require @ccmode{} to scan large portions of the buffer to determine syntactic context. Such pathological code(27) can cause @ccmode{} to perform fairly badly. This section identifies some of the coding styles to watch out for, and suggests some workarounds that you can use to improve performance.
Because @ccmode{} has to scan the buffer backwards from the current insertion point, and because C's syntax is fairly difficult to parse in the backwards direction, @ccmode{} often tries to find the nearest position higher up in the buffer from which to begin a forward scan. The farther this position is from the current insertion point, the slower the mode gets. Some coding styles can even force @ccmode{} to scan from the beginning of the buffer for every line of code!
One of the simplest things you can do to reduce scan time, is make sure
any brace that opens a top-level construct(28) always appears in the
leftmost column. This is actually an Emacs constraint, as embodied in
the beginning-of-defun function which @ccmode{} uses
heavily. If you insist on hanging top-level open braces on the right
side of the line, then you might want to set the variable
defun-prompt-regexp to something reasonable (29), however that "something
reasonable" is difficult to define, so @ccmode{} doesn't do it
for you.
A special note about defun-prompt-regexp in Java mode: while much
of the early sample Java code seems to encourage a style where the brace
that opens a class is hung on the right side of the line, this is not a
good style to pursue in Emacs. @ccmode{} comes with a variable
c-Java-defun-prompt-regexp which tries to define a regular
expression usable for this style, but there are problems with it. In
some cases it can cause beginning-of-defun to hang(30). For this reason,
it is not used by default, but if you feel adventurous, you can set
defun-prompt-regexp to it in your mode hook. In any event,
setting and rely on defun-prompt-regexp will definitely slow
things down!
You will probably notice pathological behavior from @ccmode{} when working in files containing large amounts of C preprocessor macros. This is because Emacs cannot skip backwards over these lines as quickly as it can comment.
Previous versions of @ccmode{} had potential performance problems
when recognizing K&R style function argument declarations. This was
because there are ambiguities in the C syntax when K&R style argument
lists are used(31). @ccmode{} has adopted BOCM's convention for
limiting the search: it assumes that argdecls are indented at least one
space, and that the function headers are not indented at all. With
current versions of @ccmode{}, user customization of
c-recognize-knr-p is deprecated. Just don't put argdecls in
column zero!
You might want to investigate the speed-ups contained in the
file `cc-lobotomy.el', which comes as part of the @ccmode{}
distribution, but is completely unsupported.
As mentioned previous, @ccmode{} always trades speed for accuracy,
however it is recognized that sometimes you need speed and can sacrifice
some accuracy in indentation. The file `cc-lobotomy.el' contains
hacks that will "dumb down" @ccmode{} in some specific ways, making
that trade-off of accurancy for speed. I won't go into details of its
use here; you should read the comments at the top of the file, and look
at the variable cc-lobotomy-pith-list for details.
Q. How do I re-indent the whole file?
A. Visit the file and hit C-x h to mark the whole buffer. Then hit ESC C-\.
Q. How do I re-indent the entire function? ESC C-x doesn't work.
A. ESC C-x is reserved for future Emacs use. To re-indent the entire function hit C-c C-q.
Q. How do I re-indent the current block?
A. First move to the brace which opens the block with ESC C-u, then re-indent that expression with ESC C-q.
Q. Why doesn't the RET key indent the line to where the new text should go after inserting the newline?
A. Emacs' convention is that RET just adds a newline, and that C-j adds a newline and indents it. You can make RET do this too by adding this to your
c-mode-common-hook(see the sample `.emacs' file section Sample .emacs file):(define-key c-mode-base-map "\C-m" 'newline-and-indent)This is a very common question. If you want this to be the default behavior, don't lobby me, lobby RMS!
:-)Q. I put
(c-set-offset 'substatement-open 0)in my `.emacs' file but I get an error saying thatc-set-offset's function definition is void.A. This means that @ccmode{} wasn't loaded into your Emacs session by the time the
c-set-offsetcall was reached, mostly likely because @ccmode{} is being autoloaded. Instead of putting thec-set-offsetline in your top-level `.emacs' file, put it in yourc-mode-common-hook, or simply add the following to the top of your `.emacs' file:(require 'cc-mode)See the sample `.emacs' file section Sample .emacs file for details.
Q. How do I make strings, comments, keywords, and other constructs appear in different colors, or in bold face, etc.?
A. "Syntax Colorization" is a standard Emacs feature, controlled by
font-lock-mode. It is not part of @ccmode{}.Q. M-a and M-e used to move over entire balanced brace lists, but now they move into blocks. How do I get the old behavior back?
A. Use C-M-f and C-M-b to move over balanced brace blocks. Use M-a and M-e to move by statements, which will move into blocks.
@ccmode{} is now standard with the latest versions of Emacs 19 and XEmacs 19. It is also the standard for Emacs 20 and XEmacs 20. You would typically just use the version that comes with your X/Emacs. These may be slightly out of date due to release schedule skew, so you should always check the canonical site for the latest version.
World Wide Web:
http://www.python.org/ftp/emacs/
Anonymous FTP:
ftp://ftp.python.org/pub/emacs/
There are many files under these directories; you can pick up the entire
distribution (named cc-mode.tar.gz; a gzip'd tar file), or any of
the individual files, including PostScript documentation.
If you do not have World Wide Web, or anonymous ftp access, you can get the distribution through an anonymous ftp-to-mail gateway, such as the one run by DEC at:
ftpmail@decwrl.dec.com
To get @ccmode{} via email, send the following message in the body of your mail to that address:
reply <a valid net address back to you> connect ftp.python.org binary uuencode chdir pub/emacs get cc-mode.tar.gz
or just send the message "help" for more information on ftpmail. Response times will vary with the number of requests in the queue. I am in no way connected to this service, so I make no claims or guarantees about its availability!
;; Here's a sample .emacs file that might help you along the way. Just
;; copy this region and paste it into your .emacs file. You may want to
;; change some of the actual values.
(defconst my-c-style
'((c-tab-always-indent . t)
(c-comment-only-line-offset . 4)
(c-hanging-braces-alist . ((substatement-open after)
(brace-list-open)))
(c-hanging-colons-alist . ((member-init-intro before)
(inher-intro)
(case-label after)
(label after)
(access-label after)))
(c-cleanup-list . (scope-operator
empty-defun-braces
defun-close-semi))
(c-offsets-alist . ((arglist-close . c-lineup-arglist)
(substatement-open . 0)
(case-label . 4)
(block-open . 0)
(knr-argdecl-intro . -)))
(c-echo-syntactic-information-p . t)
)
"My C Programming Style")
;; Customizations for all of c-mode, c++-mode, and objc-mode
(defun my-c-mode-common-hook ()
;; add my personal style and set it for the current buffer
(c-add-style "PERSONAL" my-c-style t)
;; offset customizations not in my-c-style
(c-set-offset 'member-init-intro '++)
;; other customizations
(setq tab-width 8
;; this will make sure spaces are used instead of tabs
indent-tabs-mode nil)
;; we like auto-newline and hungry-delete
(c-toggle-auto-hungry-state 1)
;; keybindings for all supported languages. We can put these in
;; c-mode-base-map because c-mode-map, c++-mode-map, objc-mode-map,
;; java-mode-map, and idl-mode-map inherit from it.
(define-key c-mode-base-map "\C-m" 'newline-and-indent)
)
(add-hook 'c-mode-common-hook 'my-c-mode-common-hook)
filladapt to
fill comments.
c-indent-exp has not been fully optimized. It essentially
equivalent to hitting TAB (c-indent-command) on every
line. Some information is cached from line to line, but such caching
invariable causes inaccuracies in analysis in some bizarre situations.
To report bugs, use the C-c C-b (c-submit-bug-report)
command. This provides vital information I need to reproduce your
problem. Make sure you include a concise, but complete code example.
Please try to boil your example down to just the essential code needed
to reproduce the problem, and include an exact recipe of steps needed to
expose the bug. Be especially sure to include any code that appears
before your bug example, if you think it might affect my ability
to reproduce it.
Bug reports are now sent to the following email addresses:
cc-mode-help@python.org and
bug-gnu-emacs@gnu.org; the latter is mirrored on the
Usenet newsgroup gnu.emacs.bug. You can send other questions and
suggestions (kudos? ;-) to cc-mode-help@python.org, or
help-gnu-emacs@gnu.org which is mirrored on newsgroup
gnu.emacs.help.
If you want to get announcements of new CC Mode releases, send the
word subscribe in the body of a message to
cc-mode-announce-request@python.org. Announcements will also be
posted to the Usenet newsgroup gnu.emacs.sources. Note that the
cc-mode-victims@python.org mailing list was recently
decommissioned.
Jump to: - - . - a - b - c - d - e - f - g - h - i - j - k - l - m - n - o - p - r - s - t - u - w
Since all @ccmode{} commands are prepended with the string
`c-', each appears under its c-<thing> name and its
<thing> (c-) name.
Jump to:
a
-
b
-
c
-
d
-
e
-
f
-
h
-
i
-
j
-
l
-
m
-
n
-
o
-
s
-
t
-
u
-
v
Jump to: # - < - > - b - c - d - e - m - r - t
Since all @ccmode{} variables are prepended with the string
`c-', each appears under its c-<thing> name and its
<thing> (c-) name.
Jump to:
b
-
c
-
d
-
e
-
f
-
h
-
i
-
j
-
l
-
m
-
o
-
p
-
r
-
s
-
t
This document was generated on 6 November 2000 using texi2html 1.56k.