|
|
#define INIT declarations #define GETC () getc code #define PEEK () peekc code #define UNGETC () ungetcc code #define RETURN (ptr) return code #define ERROR (val) error code#include <regexp.h>
char *compile(char *instring, char *expbuf, const char *endbuf, int eof);
int step(const char *string, const char *expbuf);
int advance(const char *string, const char *expbuf);
extern char *loc1, *loc2, *locs;
This page describes general-purpose,
regular expression matching routines in the
form of
ed(C),
defined in
<regexp.h>.
Programs such as
ed(C),
sed(C),
grep(C),
expr(C),
and so on, that perform regular expression matching
use this source file.
In this way,
only this file need be changed to maintain regular expression compatibility.
The advance and step functions do pattern matching give an character string and a compiled regular expression as input.
The interface to this file is complex.
Programs that include this file must have
the following five macros declared before the
#include <regexp.h>
statement.
These macros are used by the compile routine.
Error | Meaning | |
---|---|---|
11 | Range endpoint too large. | |
16 | Bad number. | |
25 | ````\''digit'' out of range. | |
36 | Illegal or missing delimiter. | |
41 | No remembered search string. | |
42 | ``\(~\)'' imbalance. | |
43 | Too many ``\(''. | |
44 | More than 2 numbers given in ``\{~\}''. | |
45 | ``}'' expected after ``\''. | |
46 | First number exceeds second in ``\{~\}''. | |
49 | ``[]'' imbalance. | |
50 | Regular expression overflow. |
The syntax of the compile routine is as follows:
compile(instring, expbuf, endbuf, eof)The first parameter instring is never used explicitly by the compile routine but is useful for programs that pass down different pointers to input characters. It is sometimes used in the INIT declaration (see below). Programs that call functions to input characters or have characters in an external array can pass down a value of
((char ) 0)
for this parameter.
The next parameter expbuf is a character pointer. It points to the place where the compiled regular expression is placed.
The parameter endbuf is one more than the highest address where the compiled regular expression may be placed. If the compiled expression cannot fit in (endbuf-expbuf) bytes, a call to ERROR(50) is made.
The parameter eof is the character which marks the end of the regular expression. For example, in ed(C), this character is usually a ``/''.
Each program that includes this file must have a #define statement for INIT. This definition is placed right after the declaration for the function compile and the opening curly brace (``{''). It is used for dependent declarations and initializations. Most often it is used to set a register variable to point to the beginning of the regular expression so that this register variable can be used in the declarations for GETC(), PEEKC() and UNGETC(). Otherwise it can be used to declare external variables that might be used by GETC(), PEEKC() and UNGETC(). See the example below of the declarations taken from grep(C).
There are other functions in this file which perform actual regular expression matching, one of which is the function step. The call to step is as follows:
step(string, expbuf)The first parameter to step is a pointer to a string of characters to be checked for a match. The second parameter, expbuf, is the compiled regular expression which was obtained by a call of the function compile.
The function step returns non-zero if the given string matches the regular expression, and zero if the expressions do not match. If there is a match, two external character pointers are set as a side effect to the call to step. The variable set in step is loc1. This is a pointer to the first character that matched the regular expression. The variable loc2, which is set by the function advance, points to the character after the last character that matches the regular expression. Thus if the regular expression matches the entire line, loc1 points to the first character of string and loc2 points to the null at the end of string.
step uses the external variable circf
which is set by compile if the regular expression begins
with ``^''.
If this is set, then
step
tries to match
the regular expression to the beginning of the string only.
If more than one regular expression is to be
compiled before the first is executed, the value
of
circf
should be saved for each compiled expression,
and
circf
should be set to that saved value before each call
to
step.
The function
advance
is called from
step
with the same arguments as
step.
The purpose of
step
is to step through the
string
argument and call
advance
until
advance
returns non-zero indicating a match or until the end of
string
is reached.
If one wants to constrain
string
to the beginning of the
line in all cases,
step
need not be called; simply call
advance.
When advance encounters a ``'' or ``\{~\}'' sequence in the regular expression, it advances its pointer to the string to be matched as far as possible and recursively calls itself trying to match the rest of the string to the rest of the regular expression. As long as there is no match, advance backs up along the string until it finds a match or reaches the point in the string that initially matched the ``'' or ``\{~\}''. It is sometimes desirable to stop this backing up before the initial point in the string is reached. If the external character pointer locs is equal to the point in the string at sometime during the backing up process, advance breaks out of the loop that backs up and returns zero. This is used by ed(C) and sed(C) for substitutions done globally (not just the first occurrence, but the whole line) so, for example, expressions like s/y//g do not loop forever.
The additional external variables
sed and nbra
are used for special purposes.
#define INIT register char *sp = instring; #define GETC() (*sp++) #define PEEKC() (*sp) #define UNGETC(c) (--sp) #define RETURN(c) return; #define ERROR(c) regerr()#include <regexp.h> ... (void) compile(*argv, expbuf, &expbuf[ESIZE], '\0'); ... if (step(linebuf, expbuf)) succeed();
X/Open Portability Guide, Issue 3, 1989
.
circf,
sed
and
nbra
are not part of any currently supported standard;
they are an extension of AT&T System V provided by the Santa Cruz Operation.