7 Customizing Filling and Line Breaking

Since there’s a lot of normal text in comments and string literals, CC Mode provides features to edit these like in text mode. It does this by hooking in on the different line breaking functions and tuning relevant variables as necessary.

To make Emacs recognize comments and treat text in them as normal paragraphs, CC Mode makes several standard variables24 buffer-local and modifies them according to the language syntax and the comment line prefix.

User Option: c-comment-prefix-regexp

This style variable contains the regexp used to recognize the comment line prefix, which is the line decoration that starts every line in a comment. The variable is either the comment line prefix itself, or (more usually) an association list with different values for different languages. The symbol for the major mode is looked up in the alist to get the regexp for the language, and if it isn’t found then the special symbol ‘other’ is looked up instead.

When a comment line gets divided by M-j or the like, CC Mode inserts the comment line prefix from a neighboring line at the start of the new line. The default value of c-comment-prefix-regexp is ‘//+\\|\\**’, which matches C++ style line comments like

// blah blah

with two or more slashes in front of them, and the second and subsequent lines of C style block comments like

/*
 * blah blah
 */

with zero or more stars at the beginning of every line. If you change this variable, please make sure it still matches the comment starter (i.e. //) of line comments and the line prefix inside block comments.

Also note that since CC Mode uses the value of c-comment-prefix-regexp to set up several other variables at mode initialization, there won’t be any effect if you just change it inside a CC Mode buffer. You need to call the command c-setup-paragraph-variables too, to update those other variables. That’s also the case if you modify c-comment-prefix-regexp in a mode hook, since CC Mode will already have set up these variables before calling the hook.

In comments, CC Mode uses c-comment-prefix-regexp to adapt the line prefix from the other lines in the comment.

CC Mode uses adaptive fill mode (see Adaptive Fill in GNU Emacs Manual) to make Emacs correctly keep the line prefix when filling paragraphs. That also makes Emacs preserve the text indentation inside the comment line prefix. E.g. in the following comment, both paragraphs will be filled with the left margins of the texts kept intact:

/* Make a balanced b-tree of the nodes in the incoming
 * stream.  But, to quote the famous words of Donald E.
 * Knuth,
 *
 *     Beware of bugs in the above code; I have only
 *     proved it correct, not tried it.
 */

It’s also possible to use other adaptive filling packages, notably Kyle E. Jones’ Filladapt package25, which handles things like bulleted lists nicely. There’s a convenience function c-setup-filladapt that tunes the relevant variables in Filladapt for use in CC Mode. Call it from a mode hook, e.g. with something like this in your .emacs:

(defun my-c-mode-common-hook ()
  (c-setup-filladapt)
  (filladapt-mode 1))
(add-hook 'c-mode-common-hook 'my-c-mode-common-hook)
User Option: c-block-comment-prefix

Normally the comment line prefix inserted for a new line inside a comment is deduced from other lines in it. However there’s one situation when there’s no hint about what the prefix should look like, namely when a block comment is broken for the first time. This style variable26 is used then as the comment prefix. It defaults to ‘* 27, which makes a comment

/* Got O(n^2) here, which is a Bad Thing. */

break into

/* Got O(n^2) here, which
 * is a Bad Thing. */

Note that it won’t work to adjust the indentation by putting leading spaces in c-block-comment-prefix, since CC Mode still uses the normal indentation engine to indent the line. Thus, the right way to fix the indentation is by customizing the c syntactic symbol. It defaults to c-lineup-C-comments, which handles the indentation of most common comment styles, see Line-Up Functions.

User Option: c-ignore-auto-fill

When auto fill mode is enabled, CC Mode can selectively ignore it depending on the context the line break would occur in, e.g. to never break a line automatically inside a string literal. This variable takes a list of symbols for the different contexts where auto-filling never should occur:

string

Inside a string or character literal.

c

Inside a C style block comment.

c++

Inside a C++ style line comment.

cpp

Inside a preprocessor directive.

code

Anywhere else, i.e. in normal code.

By default, c-ignore-auto-fill is set to (string cpp code), which means that when auto-fill mode is activated, auto-filling only occurs in comments. In literals, it’s often desirable to have explicit control over newlines. In preprocessor directives, the necessary ‘\’ escape character before the newline is not automatically inserted, so an automatic line break would produce invalid code. In normal code, line breaks are normally dictated by some logical structure in the code rather than the last whitespace character, so automatic line breaks there will produce poor results in the current implementation.

If inside a comment and comment-multi-line (see Auto Fill in GNU Emacs Manual is non-nil, the indentation and line prefix are preserved. If inside a comment and comment-multi-line is nil, a new comment of the same type is started on the next line and indented as appropriate for comments.

Note that CC Mode sets comment-multi-line to t at startup. The reason is that M-j could otherwise produce sequences of single line block comments for texts that should logically be treated as one comment, and the rest of the paragraph handling code (e.g. M-q and M-a) can’t cope with that, which would lead to inconsistent behavior.


Footnotes

(24)

comment-start, comment-end, comment-start-skip, paragraph-start, paragraph-separate, paragraph-ignore-fill-prefix, adaptive-fill-mode, adaptive-fill-regexp, and adaptive-fill-first-line-regexp.

(25)

It’s available from http://www.wonderworks.com/. As of version 2.12, it does however lack a feature that makes it work suboptimally when c-comment-prefix-regexp matches the empty string (which it does by default). A patch for that is available from the CC Mode web site.

(26)

In versions before 5.26, this variable was called c-comment-continuation-stars. As a compatibility measure, CC Mode still uses the value on that variable if it’s set.

(27)

Actually, this default setting of c-block-comment-prefix typically gets overridden by the default style gnu, which sets it to blank. You can see the line splitting effect described here by setting a different style, e.g. k&r See Choosing a Style.