home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Usenet 1994 January
/
usenetsourcesnewsgroupsinfomagicjanuary1994.iso
/
sources
/
unix
/
volume21
/
strsed
/
part01
next >
Wrap
Text File
|
1990-03-22
|
31KB
|
1,071 lines
Subject: v21i028: String function to do sed(1)/tr(1) manipulations, Part01/03
Newsgroups: comp.sources.unix
Approved: rsalz@uunet.UU.NET
X-Checksum-Snefru: 2996bd1c 3a2bfb34 9ce711a4 292ee33f
Submitted-by: Terry Jones <terry@pcsbst.pcs.com>
Posting-number: Volume 21, Issue 28
Archive-name: strsed/part01
Strsed is another string function. It does regular expression search and
replace (including tr(1)-like transliteration) in the style of the
ed/edit/ex text editors.
Here are a few examples:
strsed(s, "s/fred/joe/", 0) Change "fred" to "joe"
strsed(s, "g/fred/joe/", 0) Change ALL "fred" to "joe"
strsed(s, "g/a//", 0) Delete all a's
strsed(s, "s/fred.*/{a-z}{A-Z}/", 0)
Change fred and what follows to upper case
int range[2];
strsed(s, "/[0-9]*/", range)
Search for a number in s; indices of the matched portion are
returned in range[0] & range[1].
It is general enough to be able to emulate most of the usual more
complicated string functions (including strchr, strrchr, strpbrk, strspn,
strcspn and strtok), and can perform complicated string manipulations that
are otherwise a major headache.
The GNU regex package is needed for linking. The necessary source files
are regex.c and regex.h which were included in the GNU 18.55 (and 18.54)
emacs distribution. Both are expected to be in the current directory when
you type 'make'.
#!/bin/sh
# shar: Shell Archiver (v1.22)
# Packed Thursday March, 8. 1990, 20:45:41 MEZ by terry
# from directory /user1/terry/s/regex
#
# This is part 1 of a multipart archive
# do not concatenate these parts, unpack them in order with /bin/sh
#
# Run the following text with /bin/sh to create:
# README
# strsed.3c
# strsed.c
# Makefile
# check1.c
# check2.c
# examples1
# examples2
# MANIFEST
#
if test -r s2_seq_.tmp
then echo "Must unpack archives in sequence!"
next=`cat s2_seq_.tmp`; echo "Please unpack part $next next"
exit 1; fi
echo "x - extracting README (Text)"
sed 's/^X//' << 'SHAR_EOF' > README &&
X
XREADME for strsed
X
XStrsed is another string function. It does regular expression search
Xand replace (including tr(1)-like transliteration) in the style of the
Xed/edit/ex text editors.
X
XHere are a few examples:
X
X (change "fred" to "joe") strsed(s, "s/fred/joe/", 0);
X
X (change all "fred"s to "joe") strsed(s, "g/fred/joe/", 0);
X
X (delete all a's) strsed(s, "g/a//", 0);
X
X (change fred and what follows
X to upper case) strsed(s, "s/fred.*/{a-z}{A-Z}/", 0);
X
X (strip repeated letters
X in the ranges a-z and 0-9) strsed(s, "g/\([a-z0-9]\)\1+/\1/", 0);
X
X (search for a number in s) int range[2];
X strsed(s, "/[0-9]*/", range);
X /* indices of the matched portion are
X returned in range[0] & range[1] */
X
X
Xand there's more... See the man page in 'strsed.3c' for details.
X
XIt is general enough to be able to emulate most of the usual more
Xcomplicated string functions (including strchr, strrchr, strpbrk,
Xstrspn, strcspn and strtok), and can perform complicated string
Xmanipulations that are otherwise a major headache.
X
XThe GNU regex package is needed for linking. The necessary source
Xfiles are regex.c and regex.h which were included in the GNU 18.55
X(and 18.54) emacs distribution. Both are expected to be in the
Xcurrent directory when you type 'make'.
X
XStrsed is distributed with no conditions or guarantees attached.
XBear in mind however that the GNU license applies to the regex
Xsources.
X
XThis was developed in a (basically) System V environment using gcc
X(1.37), but there should be no problem at all in compiling it on BSD
Xor elsewhere - everything is very standard.
X
XTo compile, copy or link the GNU regex.{c,h} files into the current
Xdirectory and type 'make'. If you do not have gcc then change the
XMakefile to use cc. Some compilers (NOT gcc) (notably MIPS' cc) have
Xbeen known to have problems compiling regex.c into something that will
Xrun without coredumping.
X
XThis will produce 'strsed.o'. As well as this, two checking programs
Xwill be compiled and example files will be run. If all is well you
Xwill get the two messages:
X
X Substitution and transliteration tests passed successfully.
X Searching tests passed successfully.
X
XIndicating that things are OK.
X
X
X
XHave fun...
X
XTerry Jones
X(terry@pcsbst.pcs.com or ...!{pyramid,unido}!pcsbst!distel!terry)
X
XPCS Computer Systeme GmbH
XPfalzer-Wald-Str 36
X8000 Munchen 90
XWest Germany 49-89-68004288
SHAR_EOF
chmod 0644 README || echo "restore of README fails"
set `wc -c README`;Sum=$1
if test "$Sum" != "2531"
then echo original size 2531, current size $Sum;fi
echo "x - extracting strsed.3c (Text)"
sed 's/^X//' << 'SHAR_EOF' > strsed.3c &&
X.\"
X.\" $Header: /user1/terry/s/regex/RCS/strsed.3c,v 1.3 90/03/07 15:48:59 terry Exp Locker: terry $
X.\"
X.\"
X.\" $Log: strsed.3c,v $
X.\\" Revision 1.3 90/03/07 15:48:59 terry
X.\\" Lots of little things.
X.\\"
X.\\" Revision 1.2 90/03/06 22:48:00 terry
X.\\" Changed the bit about the 3rd arg - which is now necessary.
X.\\"
X.\\" Revision 1.1 90/01/18 20:03:54 terry
X.\\" Initial revision
X.\\"
X.\"
X.TH STRSED 3C
X.SH NAME
Xstrsed \- ed(1)/tr(1)\-like substitute and replace function.
X.SH SYNOPSIS
X\fBchar *strsed(string, command, 0)
X.br
Xchar *string;
X.br
Xchar *command;\fR
X.br
X.sp
Xor
X.sp
X\fBchar *strsed(string, command, range)
X.br
Xchar *string;
X.br
Xchar *command;
X.br
Xint range[2];\fR
X.SH DESCRIPTION
X.B Strsed
Xis a regular expression pattern match and replace
Xfunction that also combines
X.B tr(1)
Xlike transliteration. The GNU regex package is used
Xfor the regular expression matching.
X.PP
X.B Strsed
Xcan be used to provide the
Xfunctionality of most of the other more "complicated"
Xstring functions (e.g.
X.BR strchr ,
X.BR strrchr ,
X.BR strpbrk ,
X.BR strspn ,
X.BR strcspn ,
Xand
X.BR strtok ),
Xalthough less efficiently
Xin each case, due to its generality.
X.B Strsed
Xis a very powerful and general function that can be used
Xto carry out complicated string manipulations such as
Xthose that are possible in text editors.
X.SH USAGE
X.I String
Xshould be a null\-terminated character string.
XA copy is made and will be operated on according to the
Xsearch and replace instructions contained in
X.IR command .
XUnless an error occurs (see ERRORS), the passed character strings
X.I string
Xand
X.I command
Xare
X.I never
Xcorrupted, and the string that is returned may always
Xbe passed to
X.B free()
Xsince its space is obtained from
X.BR malloc() .
X.PP
XBoth
X.I string
Xand
X.I command
Xmay contain the following C\-like escape sequences:
X.sp
X \eb Backspace.
X.br
X \ef Formfeed.
X.br
X \en Newline.
X.br
X \er Carriage Return.
X.br
X \es Space.
X.br
X \et Horizontal Tab.
X.br
X \ev Vertical Tab.
X.br
X \ez Used to remove ambiguity if necessary.
X.br
X \e0\-9 A reference to a register.
X.br
X (except for \e0 in a regular expression.)
X.br
X \e0x3d The character whose value is 3d hexadecimal.
X.br
X \e0X3d The character whose value is 3d hexadecimal.
X.br
X \e040 The character whose value is 40 octal.
X.br
X \e32 The character whose value is 32 decimal.
X.sp
XThe NUL (0) character cannot be specified.
XA '\e' followed by one to three digits can be interpreted in
Xseveral ways. If one or two hex digits are preceeded by
Xan 'x' or an 'X',
Xthey will be taken as specifying a character in hexadecimal.
XIf there are exactly three
Xoctal digits and the first is in the range '0' to '3'
Xthen they are taken as specifying a character in octal.
XOtherwise a single digit
Xis taken to be a register reference and two or three digits
Xare interpreted as specifying a character in decimal.
X\ez can be used to
Xavoid problems with ambiguity. For instance,
X.B \e007
Xwill be interpreted by
X.B strsed
Xas octal 007. To specify the contents of register zero (\e0)
Xfollowed by the two characters "07", use
X.BR \e0\ez07 .
XThe \ez makes it clear what is meant (acting like a punctuation mark)
Xand is otherwise ignored.
X.PP
XStrsed allows
X.B ed(1)
Xlike regular expressions and substitutions
Xon
X.IR string .
XThe search and replace command is specified by
X.IR command .
XThe format of
X.I command
Xis either
X.sp
X\fB/search_pattern/replacement/\fR
X.br
Xor
X.br
X\fBg/search_pattern/replacement/\fR
X.PP
XIn the first form, the search and replace is performed once on
Xthe string, and in the second, the replacement is done globally
X(i.e. for every occurrence of the search pattern in
X.IR string .).
XA leading
X"s" in the above is silently ignored. This allows for a syntax
Xmore like that of
X.BR ed(1) .
Xe.g.
X.I s/e/x/
Xis the same as
X.IR /e/x/ .
X.PP
XIf
X.I replacement
Xis empty, then the matched text will
Xbe replaced by nothing \- i.e. deleted.
X.PP
X.I Search_pattern
Xis a full regular expression
X(see
X.BR ed(1) ),
Xincluding register
Xspecifications (i.e. \fB\e( ... \e)\fR) and register references,
X(e.g. \fB\e2\fR)
Xbut not the \fB{m,n}\fR
Xrepetition feature of
X.BR ed(1) .
X.PP
X.I Replacement
Xconsists of ordinary characters and/or register references
X(e.g. \fB\e1\fR or \fB\e2\fR.) \fB\e0\fR means the entire matched
Xtext. In addition, a
Xregister reference may be immediately followed by a
Xtransliteration request, of the form
X.br
X\fB{char\-list\-1}{char\-list\-2}\fR.
X.br
XThe characters from
X.I char\-list\-1
Xwill be transliterated into the corresponding ones from
X.I char\-list\-2
Xin the same manner as
X.BR tr(1) .
XIf the register reference before a transliteration
Xrequest is omitted, it defaults to \fB\e0\fR.
XWithin a transliteration request, the characters
X"}" and "\-" are metacharacters and must be escaped with
Xa leading \e if you want them to be interpreted literally.
XCharacter ranges such as a\-z are expanded in the same
Xfashion as
X.BR tr(1) .
XIf
X.I char\-list\-2
Xis shorter than
X.I char\-list\-1
Xthen
X.I char\-list\-2
Xis padded to be the same length as
X.I char\-list\-1
Xby repeating its last character as many times as are needed.
XFor example, the transliteration request
X.br
X.B {a\-z}{X}
X.br
Xwill transliterate all lower case letters into an 'X'.
XCharacter ranges may be increasing or decreasing.
X.PP
XUnusual character ranges (such as
X.BR a\-f\-0\-\e0x2d\-c )
Xare interpreted as running from their first character
Xto their last (so the above would be treated as
X.BR a\-c ).
XNote that it is
X.B not
Xpossible (in this release) to specify the complement of a
Xcharacter range in a transliteration request. However, this
Xcan be done in the
X.I search_pattern
Xby commencing a character class with a "^" in the normal
Xregular expression fashion.
X.PP
XThe highest register that can be referenced is \fB\e9\fR.
X.SH EXAMPLES
XHere are some example
X.I command
Xstrings that might be given to
X.BR strsed .
X.sp
X\fB/a/A/\fR # Change the first 'a' into an 'A'
X.br
X\fBg/a/A/\fR # Change every 'a' into an 'A'
X.br
X\fBg/://\fR # Delete every ':'
X.br
X\fBg/jack/jill/\fR # Change every 'jack' to a 'jill'
X.br
X\fB/[^\es\et]/X/\fR # Change the first non\-whitespace character into an 'X'.
X.sp
XSome more advanced examples...
X.sp
X\fB/\e([\es\et]*\e)\e([^\es\et]*\e)/\e1\e2{a\-z}{A\-Z}/\fR
X.sp
XThis converts the first non\-whitespace word to upper case,
Xpreserving any initial whitespace.
XIt catches the first run of spaces and TABs into register
Xone \e([\es\et]*\e), and then the following run of non\-white
Xcharacters into register two \e([^\es\et]*\e). The replacement,
X\e1\e2{a\-z}{A\-Z} specifies register 1 (the whitespace) followed
Xby the contents of register 2 transliterated into uppercase.
XThis would produce
X.br
X" SPOTTED pinto bean"
X.br
Xif called on the string
X.br
X" spotted pinto bean".
X.sp
X\fBg/\e([a\-z]\e)\e1+/\e1/\fR
X.sp
XThis is a very useful example and performs the same function
Xas tr \-s. That is, it squeezes runs of identical characters
X(in the range a to z) down to a single instance of that
Xcharacter. So "beeee good" becomes "be god". The "+" is the
Xregular expression notation meaning "one or more".
X.sp
X\fBg/\e([\et\es]*\e)\e(.\e)\e([^\et\es]*\e)/\e1\e2{a\-z}{A\-Z}\e3/\fR
X.sp
XThis example capitalises the first letter of each word in the string,
Xand preserves all whitespace. It catches three things,
X.br
X1) the initial whitespace \e([\et\es]*\e) in register 1
X.br
X2) the next letter \e(.\e) in register 2
X.br
X3) the following nonwhite letters \e([^\et\es]*\e) in register 3
X.br
Xand then prints them out as they were found, with the only
Xdifference being the uppercase conversion of the contents of
Xregister 2. Given the string
X.br
X" this is a line "
X.br
Xthis command would
Xreturn
X.br
X" This Is A Line ".
X.br
XIf the initial 'g' was not present in the command, then the capitalisation
Xwould only be done to the first word in the string.
XIt is important to understand this difference well.
X.SH SEARCHING ONLY
X.B Strsed
Xmay be used to search for a regular expression in a
Xstring, but perform no action. The portion of the string
Xthat matched will be returned in the third argument
X.IR range .
XIn this case
X.I command
Xshould be of the form
X.IR /pattern/ .
XOn return,
X.I range[0]
Xwill contain an index into the original
Xstring to indicate where the match began, and
X.I range[1]
Xwill index the first character after the end of the match. For
Xexample, after the call
X.sp
Xstrsed("two big macs please", "/b.*c/", range);
X.sp
X.I range[0]
Xwill contain 4 and
X.I range[1]
Xwill contain 11.
XIf not match is found, both elements of
X.I range
Xwill contain \-1.
X.SH ERRORS
XIf
X.B strsed
Xdetects any error it returns NULL. This can happen if the
Xsyntax of
X.I command
Xis incorrect, if the regular expression in
X.I command
Xis incorrect, if space cannot be obtained from
X.BR malloc ,
Xor for other similar reasons. Note that it is
X.B not
Xan error if the empty string is returned.
X.SH "COMPILING AND LINKING STRSED"
X.B Strsed
Xshould be compiled with the \-O and \-c options of your C compiler.
XIt has no main() function. When you come to link, you use strsed.o
Xand regex.o from the GNU 18.55 (or 18.54) emacs distribution.
X.SH "OBSCURE NOTE ON REGULAR EXPRESSIONS"
XIt is possible (but not too likely) that the regular expression
Xlanguage that is recognised may differ
Xslightly from installation to installation. This is because the
XGNU regular expression package may compiled with different settings
Xfor recognition of meta-characters. So on one machine, the character
X"|" might be taken as being the OR operator, whilst somewhere else
Xyou need to give "\e|" \- or vice-versa. This could be a pain in the
Xneck, but there's not alot that can be done about it. If you
X.I really
Xneed
Xto know the difference in a portable way, look in regex.h to see
Xwhat things are defined and then act accordingly when constructing
Xcommands for
X.BR strsed .
X.SH AUTHOR
XTerry Jones
X.br
XPCS Computer Systeme GmbH
X.br
XPfaelzer\-Wald\-Str 36
X.br
X8000 Muenchen 90
X.br
XWest Germany 49\-89\-68004288
X.sp
Xterry@distel.pcs.com or ...!{pyramid,unido}!pcsbst!distel!terry
X.sp
XJanuary 8th, 1990.
X.SH ACKNOWLEDGEMENTS
XMany thanks to Jordan K. (mother) Hubbard for discussions, bugfinding,
Xhandholding, forcing me to use emacs and torrents of (usually) uncalled\-for abuse.
X.SH SEE ALSO
X.I ed(1), tr(1)
SHAR_EOF
chmod 0644 strsed.3c || echo "restore of strsed.3c fails"
set `wc -c strsed.3c`;Sum=$1
if test "$Sum" != "10496"
then echo original size 10496, current size $Sum;fi
echo "x - extracting strsed.c (Text)"
sed 's/^X//' << 'SHAR_EOF' > strsed.c &&
X#ifndef lint
Xstatic char *rcsid = "$Header: /user1/terry/s/regex/RCS/strsed.c,v 1.17 90/03/08 20:44:32 terry Exp Locker: terry $";
X#endif lint
X
X/*
X * Strsed.c
X *
X * ed(1)/tr(1)-like search, replace, transliterate. See the
X * manpage for details.
X *
X * Usage:
X *
X * strsed(string, pattern, 0);
X * char *string;
X * char *pattern;
X * or
X * strsed(string, pattern, range);
X * char *string;
X * char *pattern;
X * int range[2];
X *
X *
X * Terry Jones
X * terry@distel.pcs.com
X * ...!{pyramid,unido}!pcsbst!distel!terry
X *
X * PCS Computer Systeme GmbH
X * Pfaelzer-Wald-Str 36
X * 8000 Muenchen 90
X * West Germany 49-89-68004288
X *
X * January 8th, 1990.
X *
X */
X
X/*
X * $Log: strsed.c,v $
X * Revision 1.17 90/03/08 20:44:32 terry
X * Final cleanup.
X *
X * Revision 1.16 90/03/07 15:46:35 terry
X * Changed backslash_eliminate to only malloc on
X * REPLACEMENT type. Added ".*" optimisation so that
X * the regex functions are never called.
X *
X * Revision 1.15 90/03/06 22:27:49 terry
X * Removed varargs stuff since the 3rd argument is now
X * compulsory. Cleaned up. A few comments even.
X *
X * Revision 1.14 90/03/06 21:50:28 terry
X * Touched up memory stuff. Added mem_find(). Changed
X * buf_sz and buf_inc to be a reasonable refelection
X * of the length of the input.
X *
X * Revision 1.13 90/03/06 20:22:48 terry
X * Major rearrangements. Added mem(), mem_init(), mem_save(),
X * mem_free() to handle memory in a vastly improved fashion.
X * Calls to malloc are minimised as far as possible.
X *
X * Revision 1.12 90/03/06 13:23:33 terry
X * Made map static.
X *
X * Revision 1.11 90/01/10 15:51:12 terry
X * checked in with -k by terry at 90.01.18.20.03.08.
X *
X * Revision 1.11 90/01/10 15:51:12 terry
X * *** empty log message ***
X *
X * Revision 1.10 90/01/10 12:48:40 terry
X * Fixed handling of perverted character ranges in nextch().
X * a-f-c now means a-c.
X *
X * Revision 1.9 90/01/10 12:03:48 terry
X * Pounded on space allocation, added more_space,
X * remove free() in build_map, tested tiny buffer sizes etc.
X *
X * Revision 1.8 90/01/09 18:15:12 terry
X * added backslash elimination to str.
X * altered backslash_elimantion to take one of three types
X * REGEX, NORMAL or REPLACEMENT depending on the
X * elimination desired. Changed interpretation of \
X * followed by a single digit to be that character if the
X * type of elimination is NORMAL. i.e. \4 = ^D.
X *
X * Revision 1.7 90/01/09 17:05:05 terry
X * Frozen version for release to comp.sources.unix
X *
X * Revision 1.6 90/01/09 16:47:54 terry
X * Altered pure searching return values to be -1
X *
X * Revision 1.5 90/01/09 14:54:34 terry
X * *** empty log message ***
X *
X * Revision 1.4 90/01/09 14:51:04 terry
X * removed #include <stdio> silliness.
X *
X * Revision 1.2 90/01/09 10:48:22 terry
X * Fixed handling of } and - metacharacters inside
X * transliteration request strings in backslash_eliminate().
X *
X * Revision 1.1 90/01/08 17:41:35 terry
X * Initial revision
X *
X *
X */
X
X#include <ctype.h>
X#include <string.h>
X#include <malloc.h>
X#include "regex.h"
X
X#define BYTEWIDTH 8
X#define REGEX 0
X#define REPLACEMENT 1
X#define NORMAL 2
X
X/*
X * And this is supposed to make freeing easier. It's a little hard to
X * keep track of what can and cannot be freed in what follows, so I
X * ignore it and every time a malloc is done for one of the things
X * below (and these are the only ones possible) we free if need be and
X * then alloc some more if it can't be avoided. No-one (who is going
X * to free) needs to call malloc then. And no-one need call free.
X * Wonderful in theory...
X */
X
X#define MEM_STR 0
X#define MEM_PAT 1
X#define MEM_FROM 2
X#define MEM_TO 3
X#define MEM_NEWSTR 4
X#define MEM_MAP 5
X#define MEM_MAP_SAVE 6
X
X#define MEM_SLOTS 7
X
X/*
X * This calls mem_free(), which free()s all the allocated storage EXCEPT
X * for the piece whose address is 'n'. If something goes wrong below
X * we call RETURN(0) and if we want to return some address we call RETURN
X * with the address to be returned.
X */
X
X#define RETURN(n) \
X mem_free(n); \
X return (char *)n;
X
Xstatic struct {
X char *s;
X int size;
X int used;
X} mem_slots[MEM_SLOTS];
X
X
X#define more_space(need) \
X if (need && space != -1){ \
X if (space - (need) < 0){ \
X buf_sz += buf_inc + (need) - space; \
X if (!(new_str = (char *)realloc(new_str, (unsigned)buf_sz))){ \
X RETURN(0); \
X } \
X mem_slots[MEM_NEWSTR].s = new_str; \
X mem_slots[MEM_NEWSTR].size = buf_sz; \
X space = buf_inc; \
X } \
X else{ \
X space -= need; \
X } \
X }
X
X
Xchar *
Xstrsed(string, pattern, range)
Xregister char *string;
Xregister char *pattern;
Xint *range;
X{
X extern char *re_compile_pattern();
X extern int re_search();
X
X static char *backslash_eliminate();
X static char *mem();
X static void mem_init();
X static void mem_free();
X
X char *from;
X char *new_str;
X char *pat;
X char *str;
X char *tmp;
X char *to;
X static char map[1 << BYTEWIDTH];
X int buf_sz;
X int buf_inc;
X int global = 0;
X int match;
X int new_pos = 0;
X int search_only = 0;
X int seenbs = 0;
X int space;
X int match_all = 0;
X register int str_len;
X static int first_time = 1;
X static struct re_pattern_buffer re_comp_buf;
X struct re_registers regs;
X
X if (!string || !pattern){
X RETURN(0);
X }
X
X /*
X * If this is the first time we've been called, clear the memory slots.
X */
X if (first_time){
X mem_init();
X }
X
X /*
X * Take our own copies of the string and pattern since we promised
X * in the man page not to hurt the originals.
X */
X str = mem(MEM_STR, strlen(string) + 1);
X str[0] = '\0';
X strcat(str, string);
X pat = mem(MEM_PAT, strlen(pattern) + 1);
X pat[0] = '\0';
X strcat(pat, pattern);
X
X /*
X * If escape sequences are not already removed elsewhere, remove
X * them from the string. If you don't know what you're doing here
X * or are in any doubt, don't define ESCAPED_STRING.
X */
X#ifndef ESCAPED_STRING
X if (!(str = backslash_eliminate(str, NORMAL, MEM_STR))){
X RETURN(0);
X }
X#endif
X
X str_len = strlen(str);
X
X /*
X * Set up the size of our buffer (in which we build the
X * newstring, and the size by which we increment it when
X * (and if) the need arises. There shouldn't be too much
X * growth in the average case. Of course some people will
X * go and do things like
X *
X * strsed(string, "s/.*$/\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0")
X *
X * and they will be somewhat penalised. Oh well.
X *
X */
X
X buf_sz = str_len < 8 ? 16 : str_len << 1;
X buf_inc = buf_sz;
X
X /*
X * Get the action.
X * s = substitue and g = global.
X * anything else is invalid.
X *
X */
X while (*pat && *pat != '/'){
X switch (*pat){
X case 'g':{
X global = 1;
X break;
X }
X case 's':{
X break;
X }
X default:{
X RETURN(0);
X }
X }
X pat++;
X }
X
X if (!*pat){
X RETURN(0);
X }
X
X pat++;
X
X /*
X * Now split 'pat' into its two components. These are delimited (or
X * should be) by (unquoted) '/'. The first we point to with 'from'
X * and the second with 'to'.
X *
X * Someone should write a function to make this sort of thing trivial...
X *
X */
X
X from = to = pat;
X
X while (*to){
X if (seenbs){
X seenbs = 0;
X }
X else{
X if (*to == '\\'){
X seenbs = 1;
X }
X else if (*to == '/'){
X break;
X }
X }
X to++;
X }
X
X if (!*to){
X RETURN(0);
X }
X
X *to++ = '\0';
X
X if (*to){
X tmp = to + strlen(to) - 1;
X
X /*
X * Back up to the last non-whitespace char in 'to'
X *
X */
X
X while (*tmp == ' ' || *tmp == '\t'){
X tmp--;
X }
X
X /*
X * Make sure that, if there was a character,
X * that it was a / and wasn't preceded by \.
X *
X */
X
X if (*tmp && !(*tmp = '/' && *(tmp - 1) != '\\')){
X RETURN(0);
X }
X
X *tmp = '\0';
X }
X else{
X /*
X * Search only.
X * It doesn't make sense to say
X *
X * strsed(string, "g/abc/", range)
X *
X * because we are only searching and returning the
X * matched indexes. So turn off global (in case it's on)
X * so that we will return just the first instance.
X *
X * If no range has been given either, then there's no
X * point in going on.
X *
X */
X
X if (!range){
X RETURN(0);
X }
X
X global = 0;
X search_only = 1;
X }
X
X /*
X * Eliminate backslashes and character ranges etc.
X *
X */
X
X if (!(from = backslash_eliminate(from, REGEX, MEM_FROM)) ||
X !(to = backslash_eliminate(to, REPLACEMENT, MEM_TO))){
X RETURN(0);
X }
X
X /*
X * If the first char of 'to' is '\0' then we are deleting or
X * searching only. We don't have to worry about space since
X * the transformed string will be less than or equal in length
X * to the original. We just overwrite.
X * We set space = -1 so that later on we can avoid worrying
X * about overflow etc.
X *
X * Otherwise, we are doing a substitution. Here we have to
X * worry about space because the replacement may be larger
X * than the original. malloc some room and if we overflow it
X * later we will realloc. slows things down if the new string
X * turns out to be too much bigger. oh well.
X *
X */
X
X if (*to){
X if (!(new_str = mem(MEM_NEWSTR, buf_sz + 1))){
X RETURN(0);
X }
X space = buf_sz;
X }
X else{
X new_str = str;
X space = -1;
X }
X
X /*
X * Do things to get ready for the regex functions.
X * Don't do anything though if the regex in 'from' is ".*"
X * We handle that below. (Just a special case optimisation).
X *
X */
X
X if (from[0] == '.' && from[1] == '*' && from[2] == '\0'){
X register int i;
X match_all = 1;
X /*
X * For safety's sake, clear out the register values.
X * There might be a register reference in the replacement.
X * There will be nothing in the register (since the search
X * pattern was ".*"). Since we aren't calling the regex
X * stuff we can't rely on it to set these to -1.
X */
X for (i = 0; i < RE_NREGS; i++){
X regs.start[i] = -1;
X }
X }
X else{
X if (first_time){
X if (!(re_comp_buf.buffer = (char *)malloc((unsigned)200))){
X RETURN(0);
X }
X
X re_comp_buf.allocated = 200;
X
X if (!(re_comp_buf.fastmap = (char *)malloc((unsigned)1 << BYTEWIDTH))){
X RETURN(0);
X }
X
X first_time = 0;
X }
X
X re_comp_buf.translate = 0;
X re_comp_buf.used = 0;
X
X /*
X * Compile the r.e.
X *
X */
X if (re_compile_pattern(from, strlen(from), &re_comp_buf)){
X RETURN(0);
X }
X }
X
X
X /*
X * Now get on with the matching/replacing etc.
X *
X */
X
X do {
X if (match_all){
X /* Fake a match instead of calling re_search(). */
X match = 1;
X regs.start[0] = 0;
X regs.end[0] = str_len;
X }
X else{
X match = re_search(&re_comp_buf, str, str_len, 0, str_len, ®s);
X }
X
X if (search_only){
X /*
X * Show what happened and return.
X *
X */
X
X range[0] = match == -1 ? -1 : regs.start[0];
X range[1] = match == -1 ? -1 : regs.end[0];
X RETURN(str);
X }
X
X if (match != -1){
X
X /*
X * Copy that portion that was not matched. It will
X * be unchanged in the output string.
X *
X */
X more_space(regs.start[0]);
X strncpy(new_str + new_pos, str, regs.start[0]);
X new_pos += regs.start[0];
X
X /*
X * Put in the replacement text (if any).
X * We substitute the contents of 'to', watching for register
X * references.
X */
X
X tmp = to;
X while (*tmp){
X if (*tmp == '\\' && isdigit(*(tmp + 1))){
X
X /* A register reference. */
X
X register int reg = *(tmp + 1) - '0';
X int translit = 0;
X int need = regs.end[reg] - regs.start[reg];
X
X /*
X * Check for a transliteration request.
X *
X */
X if (*(tmp + 2) == '{'){
X /* A transliteration table. Build the map. */
X static char *build_map();
X if (!(tmp = build_map(tmp + 2, map))){
X RETURN(0);
X }
X translit = 1;
X }
X else{
X tmp += 2;
X translit = 0;
X }
X
X more_space(need);
X
X /*
X * Copy in the register contents (if it matched), transliterating if need be.
X *
X */
X if (regs.start[reg] != -1){
X register int i;
X for (i = regs.start[reg]; i < regs.end[reg]; i++){
X new_str[new_pos++] = translit ? map[str[i]] : str[i];
X }
X }
X }
X else{
X /* A plain character, put it in. */
SHAR_EOF
echo "End of part 1"
echo "File strsed.c is continued in part 2"
echo "2" > s2_seq_.tmp
exit 0