Go to the previous, next section.
Like other editors, Emacs has commands for searching for occurrences of a string. The principal search command is unusual in that it is incremental; it begins to search before you have finished typing the search string. There are also nonincremental search commands more like those of other editors.
Besides the usual
replace-string command that finds all
occurrences of one string and replaces them with another, Emacs has a fancy
replacement command called
query-replace which asks interactively
which occurrences to replace.
An incremental search begins searching as soon as you type the first character of the search string. As you type in the search string, Emacs shows you where the string (as you have typed it so far) would be found. When you have typed enough characters to identify the place you want, you can stop. Depending on what you will do next, you may or may not need to terminate the search explicitly with an ESC first.
C-s starts an incremental search. C-s reads characters from the keyboard and positions the cursor at the first occurrence of the characters that you have typed. If you type C-s and then F, the cursor moves right after the first `F'. Type an O, and see the cursor move to after the first `FO'. After another O, the cursor is after the first `FOO' after the place where you started the search. Meanwhile, the search string `FOO' has been echoed in the echo area.
The echo area display ends with three dots when actual searching is going on. When search is waiting for more input, the three dots are removed. (On slow terminals, the three dots are not displayed.)
If you make a mistake in typing the search string, you can erase characters with DEL. Each DEL cancels the last character of search string. This does not happen until Emacs is ready to read another input character; first it must either find, or fail to find, the character you want to erase. If you do not want to wait for this to happen, use C-g as described below.
When you are satisfied with the place you have reached, you can type ESC, which stops searching, leaving the cursor where the search brought it. Also, any command not specially meaningful in searches stops the searching and is then executed. Thus, typing C-a would exit the search and then move to the beginning of the line. ESC is necessary only if the next command you want to type is a printing character, DEL, ESC, or another control character that is special within searches (C-q, C-w, C-r, C-s or C-y).
Sometimes you search for `FOO' and find it, but not the one you expected to find. There was a second `FOO' that you forgot about, before the one you were looking for. In this event, type another C-s to move to the next occurrence of the search string. This can be done any number of times. If you overshoot, you can cancel some C-s characters with DEL.
After you exit a search, you can search for the same string again by typing just C-s C-s: the first C-s is the key that invokes incremental search, and the second C-s means "search again".
If your string is not found at all, the echo area says `Failing I-Search'. The cursor is after the place where Emacs found as much of your string as it could. Thus, if you search for `FOOT', and there is no `FOOT', you might see the cursor after the `FOO' in `FOOL'. At this point there are several things you can do. If your string was mistyped, you can rub some of it out and correct it. If you like the place you have found, you can type ESC or some other Emacs command to "accept what the search offered". Or you can type C-g, which removes from the search string the characters that could not be found (the `T' in `FOOT'), leaving those that were found (the `FOO' in `FOOT'). A second C-g at that point cancels the search entirely, returning point to where it was when the search started.
If a search is failing and you ask to repeat it by typing another C-s, it starts again from the beginning of the buffer. Repeating a failing reverse search with C-r starts again from the end. This is called wrapping around. `Wrapped' appears in the search prompt once this has happened.
The C-g "quit" character does special things during searches; just what it does depends on the status of the search. If the search has found what you specified and is waiting for input, C-g cancels the entire search. The cursor moves back to where you started the search. If C-g is typed when there are characters in the search string that have not been found--because Emacs is still searching for them, or because it has failed to find them--then the search string characters which have not been found are discarded from the search string. With them gone, the search is now successful and waiting for more input, so a second C-g will cancel the entire search.
To search for a control character such as C-s or DEL or ESC, you must quote it by typing C-q first. This function of C-q is analogous to its meaning as an Emacs command: it causes the following character to be treated the way a graphic character would normally be treated in the same context. You can also specify a quoted character in octal while searching, just as you can for insertion. See section Basic Editing Commands.
You can change to searching backwards with C-r. If a search fails because the place you started was too late in the file, you should do this. Repeated C-r keeps looking for more occurrences backwards. A C-s starts going forwards again. C-r in a search can be cancelled with DEL.
If you know initially that you want to search backwards, you can
use C-r instead of C-s to start the search, because C-r
is also a key running a command (
isearch-backward) to search
The characters C-y and C-w can be used in incremental search to grab text from the buffer into the search string. This makes it convenient to search for another occurrence of text at point. C-w copies the word after point as part of the search string, advancing point over that word. Another C-s to repeat the search will then search for a string including that word. C-y is similar to C-w but copies all the rest of the current line into the search string.
All the characters special in incremental search can be changed by setting the following variables:
Incremental search on a slow terminal uses a modified style of display that is designed to take less time. Instead of redisplaying the buffer at each place the search gets to, it creates a new single-line window and uses that to display the line that the search has found. The single-line window comes into play as soon as point gets outside of the text that is already on the screen.
When the search is terminated, the single-line window is removed. Only at this time is the window in which the search was done redisplayed to show its new value of point.
The three dots at the end of the search string, normally used to indicate that searching is going on, are not displayed in slow style display.
The slow terminal style of display is used when the terminal baud rate is
less than or equal to the value of the variable
The number of lines to use in slow terminal search display is controlled
by the variable
search-slow-window-lines. 1 is its normal value.
Emacs also has conventional nonincremental search commands, which require you to type the entire search string before searching begins.
To do a nonincremental search, first type C-s ESC. This enters the minibuffer to read the search string; terminate the string with RET, and then the search is done. If the string is not found the search command gets an error.
The way C-s ESC works is that the C-s invokes incremental search, which is specially programmed to invoke nonincremental search if the argument you give it is empty. (Such an empty argument would otherwise be useless.) C-r ESC also works this way.
Forward and backward nonincremental searches are implemented by the
search-backward. These commands
may be bound to keys in the usual manner. The reason that incremental
search is programmed to invoke them as well is that C-s ESC
is the traditional sequence of characters used in Emacs to invoke
However, nonincremental searches performed using C-s ESC do
search-forward right away. The first thing done is to see
if the next character is C-w, which requests a word search.
Word search searches for a sequence of words without regard to how the words are separated. More precisely, you type a string of many words, using single spaces to separate them, and the string can be found even if there are multiple spaces, newlines or other punctuation between the words.
Word search is useful in editing documents formatted by text formatters. If you edit while looking at the printed, formatted version, you can't tell where the line breaks are in the source file. With word search, you can search without having to know them.
Word search is a special case of nonincremental search and is invoked with C-s ESC C-w. This is followed by the search string, which must always be terminated with RET. Being nonincremental, this search does not start until the argument is terminated. It works by constructing a regular expression and searching for that. See section Regular Expression Search.
A backward word search can be done by C-r ESC C-w.
Forward and backward word searches are implemented by the commands
word-search-backward. These commands
may be bound to keys in the usual manner. The reason that incremental
search is programmed to invoke them as well is that C-s ESC C-w
is the traditional Emacs sequence of keys for word search.
A regular expression (regexp, for short) is a pattern that denotes a set of strings, possibly an infinite set. Searching for matches for a regexp is a very powerful operation that editors on Unix systems have traditionally offered. In GNU Emacs, you can search for the next match for a regexp either incrementally or not.
Incremental search for a regexp is done by typing C-M-s
isearch-forward-regexp). This command reads a search string
incrementally just like C-s, but it treats the search string as a
regexp rather than looking for an exact match against the text in the
buffer. Each time you add text to the search string, you make the regexp
longer, and the new regexp is searched for. A reverse regexp search command,
isearch-backward-regexp, also exists but no key runs it.
All of the control characters that do special things within an ordinary incremental search have the same function in incremental regexp search. Typing C-s or C-r immediately after starting the search retrieves the last incremental search regexp used; that is to say, incremental regexp and non-regexp searches have independent defaults.
Note that adding characters to the regexp in an incremental regexp search does not make the cursor move back and start again. Perhaps it ought to; I am not sure. As it stands, if you have searched for `foo' and you add `\|bar', the search will not check for a `bar' in the buffer before the `foo'.
Nonincremental search for a regexp is done by the functions
re-search-backward. You can invoke
these with M-x, or bind them to keys. Also, you can call
re-search-forward by way of incremental regexp search with
Regular expressions have a syntax in which a few characters are special constructs and the rest are ordinary. An ordinary character is a simple regular expression which matches that character and nothing else. The special characters are `$', `^', `.', `*', `+', `?', `[', `]' and `\'; no new special characters will be defined. Any other character appearing in a regular expression is ordinary, unless a `\' precedes it.
For example, `f' is not a special character, so it is ordinary, and therefore `f' is a regular expression that matches the string `f' and no other string. (It does not match the string `ff'.) Likewise, `o' is a regular expression that matches only `o'.
Any two regular expressions a and b can be concatenated. The result is a regular expression which matches a string if a matches some amount of the beginning of that string and b matches the rest of the string.
As a simple example, we can concatenate the regular expressions `f' and `o' to get the regular expression `fo', which matches only the string `fo'. Still trivial. To do something nontrivial, you need to use one of the special characters. Here is a list of them.
`*' always applies to the smallest possible preceding expression. Thus, `fo*' has a repeating `o', not a repeating `fo'.
The matcher processes a `*' construct by matching, immediately, as many repetitions as can be found. Then it continues with the rest of the pattern. If that fails, backtracking occurs, discarding some of the matches of the `*'-modified construct in case that makes it possible to match the rest of the pattern. For example, matching `ca*ar' against the string `caaar', the `a*' first tries to match all three `a's; but the rest of the pattern is `ar' and there is only `r' left to match, so this try fails. The next alternative is for `a*' to match only two `a's. With this choice, the rest of the regexp matches successfully.
Character ranges can also be included in a character set, by writing two characters with a `-' between them. Thus, `[a-z]' matches any lower-case letter. Ranges may be intermixed freely with individual characters, as in `[a-z$%.]', which matches any lower case letter or `$', `%' or period.
Note that the usual special characters are not special any more inside a character set. A completely different set of special characters exists inside character sets: `]', `-' and `^'.
To include a `]' in a character set, you must make it the first character. For example, `a]' matches `]' or `a'. To include a `-', write `---', which is a range containing only `-'. To include `^', make it other than the first character in the set.
`^' is not special in a character set unless it is the first character. The character following the `^' is treated as if it were first (`-' and `]' are not special there).
Note that a complement character set can match a newline, unless newline is mentioned as one of the characters not to match.
Because `\' quotes special characters, `\$' is a regular expression which matches only `$', and `\[' is a regular expression which matches only `[', and so on.
Note: for historical compatibility, special characters are treated as ordinary ones if they are in contexts where their special meanings make no sense. For example, `*foo' treats `*' as ordinary since there is no preceding expression on which the `*' can act. It is poor practice to depend on this behavior; better to quote the special character anyway, regardless of where is appears.
For the most part, `\' followed by any character matches only that character. However, there are several exceptions: characters which, when preceded by `\', are special constructs. Such characters are always ordinary when encountered on their own. Here is a table of `\' constructs.
Thus, `foo\|bar' matches either `foo' or `bar' but no other string.
`\|' applies to the largest possible surrounding expressions. Only a surrounding `\( ... \)' grouping can limit the grouping power of `\|'.
Full backtracking capability exists to handle multiple uses of `\|'.
This last application is not a consequence of the idea of a parenthetical grouping; it is a separate feature which happens to be assigned as a second meaning to the same `\( ... \)' construct because there is no conflict in practice between the two meanings. Here is an explanation of this feature:
The strings matching the first nine `\( ... \)' constructs appearing in a regular expression are assigned numbers 1 through 9 in order that the open-parentheses appear in the regular expression. `\1' through `\9' may be used to refer to the text matched by the corresponding `\( ... \)' construct.
For example, `\(.*\)\1' matches any newline-free string that is composed of two identical halves. The `\(.*\)' matches the first half, which may be anything, but the `\1' that follows must match the same exact text.
Here is a complicated regexp, used by Emacs to recognize the end of a sentence together with any whitespace that follows. It is given in Lisp syntax to enable you to distinguish the spaces from the tab characters. In Lisp syntax, the string constant begins and ends with a double-quote. `\"' stands for a double-quote as part of the regexp, `\\' for a backslash as part of the regexp, `\t' for a tab and `\n' for a newline.
"[.?!]\"')]*\\($\\|\t\\| \\)[ \t\n]*"
This contains four parts in succession: a character set matching period, `?' or `!'; a character set matching close-brackets, quotes or parentheses, repeated any number of times; an alternative in backslash-parentheses that matches end-of-line, a tab or two spaces; and a character set matching whitespace characters, repeated any number of times.
Note that the above example shows how to write this regexp when
entering it as part of an Emacs Lisp program. To enter the same regexp
in an interactive command such as
re-search-forward you must
spell it differently:
[.?!]"')]*\($\|^Q^I\| \)[ ^Q^I^Q^J]*
All sorts of searches in Emacs normally ignore the case of the text they are searching through; if you specify searching for `FOO', then `Foo' and `foo' are also considered a match. Regexps, and in particular character sets, are included: `[aB]' would match `a' or `A' or `b' or `B'.
If you do not want this feature, set the variable
nil. Then all letters must match exactly, including case. This
is a per-buffer variable; altering the variable affects only the current
buffer, but there is a default value which you can change as well.
See section Local Variables.
Global search-and-replace operations are not needed as often in Emacs as
they are in other editors, but they are available. In addition to the
replace-string command which is like that found in most
editors, there is a
query-replace command which asks you, for each
occurrence of the pattern, whether to replace it.
The replace commands all replace one string (or regexp) with one
replacement string. It is possible to perform several replacements in
parallel using the command
expand-region-abbrevs. See section Controlling Abbrev Expansion.
To replace every instance of `foo' after point with `bar', use the command M-x replace-string with the two arguments `foo' and `bar'. Replacement occurs only after point, so if you want to cover the whole buffer you must go to the beginning first. All occurrences up to the end of the buffer are replaced; to limit replacement to part of the buffer, narrow to that part of the buffer before doing the replacement (see section Narrowing).
replace-string exits, point is left at the last occurrence
replaced. The value of point when the
replace-string command was
issued is remembered on the mark ring; C-u C-SPC moves back
A numeric argument restricts replacement to matches that are surrounded by word boundaries.
replace-string replaces exact matches for a single string. The
replace-regexp replaces any match for a specified
replace-regexp, the newstring need not be constant. It
can refer to all or part of what is matched by the regexp. `\&'
in newstring stands for the entire text being replaced.
`\d' in newstring, where d is a digit, stands for
whatever matched the d'th parenthesized grouping in regexp.
M-x replace-regexp RET c[ad]+r RET \&-safe RET
would replace (for example) `cadr' with `cadr-safe' and `cddr' with `cddr-safe'.
M-x replace-regexp RET \(c[ad]+r\)-safe RET \1 RET
would perform exactly the opposite replacements. To include a `\' in the text to replace with, you must give `\\'.
If the arguments to a replace command are in lower case, it preserves case when it makes a replacement. Thus, the command
M-x replace-string RET foo RET bar RET
replaces a lower case `foo' with a lower case `bar', `FOO'
with `BAR', and `Foo' with `Bar'. If upper case letters are
used in the second argument, they remain upper case every time that
argument is inserted. If upper case letters are used in the first
argument, the second argument is always substituted exactly as given, with
no case conversion. Likewise, if the variable
case-replace is set
nil, replacement is done without case conversion. If
case-fold-search is set to
nil, case is significant in
matching occurrences of `foo' to replace; also, case conversion of the
replacement string is not done.
If you want to change only some of the occurrences of `foo' to
`bar', not all of them, then you cannot use an ordinary
replace-string. Instead, use M-% (
This command finds occurrences of `foo' one by one, displays each
occurrence and asks you whether to replace it. A numeric argument to
query-replace tells it to consider only occurrences that are bounded
by word-delimiter characters.
Aside from querying,
query-replace works just like
The things you can type when you are shown an occurrence of string or a match for regexp are:
nil, as it normally is.
query-replace, so if you want to do further replacement you must use C-x ESC to restart (see section Repeating Minibuffer Commands).
If you type any other character, the
query-replace is exited, and
the character executed as a command. To restart the
use C-x ESC, which repeats the
query-replace because it
used the minibuffer to read its arguments. See section Repeating Minibuffer Commands.
To replace every occurrence, you can start
query-replace at the
beginning of the buffer and type !, or you can use the
replace-string command at the beginning of the buffer. To
replace every occurrence in a part of the buffer, narrow to that part
and then run
query-replace at the
beginning of it. See section Narrowing.
Here are some other commands that find matches for a regular expression. They all operate from point to the end of the buffer.
The buffer `*Occur*' containing the output serves as a menu for finding the occurrences in their original context. Find an occurrence as listed in `*Occur*', position point there and type C-c C-c; this switches to the buffer that was searched and moves point to the original of the same occurrence.
Go to the previous, next section.