head 1.19; access; symbols perseant-exfatfs-base-20250801:1.19 perseant-exfatfs-base-20240630:1.19 perseant-exfatfs:1.19.0.36 perseant-exfatfs-base:1.19 cjep_sun2x:1.19.0.34 cjep_sun2x-base:1.19 cjep_staticlib_x-base1:1.19 cjep_staticlib_x:1.19.0.32 cjep_staticlib_x-base:1.19 phil-wifi-20200421:1.19 phil-wifi-20200411:1.19 phil-wifi-20200406:1.19 pgoyette-compat-merge-20190127:1.19 pgoyette-compat-20190127:1.19 pgoyette-compat-20190118:1.19 pgoyette-compat-1226:1.19 pgoyette-compat-1126:1.19 pgoyette-compat-1020:1.19 pgoyette-compat-0930:1.19 pgoyette-compat-0906:1.19 pgoyette-compat-0728:1.19 pgoyette-compat-0625:1.19 pgoyette-compat-0521:1.19 pgoyette-compat-0502:1.19 pgoyette-compat-0422:1.19 pgoyette-compat-0415:1.19 pgoyette-compat-0407:1.19 pgoyette-compat-0330:1.19 pgoyette-compat-0322:1.19 pgoyette-compat-0315:1.19 pgoyette-compat:1.19.0.30 pgoyette-compat-base:1.19 perseant-stdc-iso10646:1.19.0.28 perseant-stdc-iso10646-base:1.19 prg-localcount2-base3:1.19 prg-localcount2-base2:1.19 prg-localcount2-base1:1.19 prg-localcount2:1.19.0.26 prg-localcount2-base:1.19 pgoyette-localcount-20170426:1.19 bouyer-socketcan-base1:1.19 pgoyette-localcount-20170320:1.19 bouyer-socketcan:1.19.0.24 bouyer-socketcan-base:1.19 pgoyette-localcount-20170107:1.19 pgoyette-localcount-20161104:1.19 localcount-20160914:1.19 pgoyette-localcount-20160806:1.19 pgoyette-localcount-20160726:1.19 pgoyette-localcount:1.19.0.22 pgoyette-localcount-base:1.19 netbsd-5-2-3-RELEASE:1.19 netbsd-5-1-5-RELEASE:1.19 yamt-pagecache-base9:1.19 yamt-pagecache-tag8:1.19 tls-earlyentropy:1.19.0.18 tls-earlyentropy-base:1.19 riastradh-xf86-video-intel-2-7-1-pre-2-21-15:1.19 riastradh-drm2-base3:1.19 netbsd-5-2-2-RELEASE:1.19 netbsd-5-1-4-RELEASE:1.19 netbsd-5-2-1-RELEASE:1.19 netbsd-5-1-3-RELEASE:1.19 agc-symver:1.19.0.20 agc-symver-base:1.19 tls-maxphys-base:1.19 yamt-pagecache-base8:1.19 netbsd-5-2:1.19.0.16 yamt-pagecache-base7:1.19 netbsd-5-2-RELEASE:1.19 netbsd-5-2-RC1:1.19 yamt-pagecache-base6:1.19 yamt-pagecache-base5:1.19 yamt-pagecache-base4:1.19 netbsd-5-1-2-RELEASE:1.19 netbsd-5-1-1-RELEASE:1.19 yamt-pagecache-base3:1.19 yamt-pagecache-base2:1.19 yamt-pagecache:1.19.0.14 yamt-pagecache-base:1.19 bouyer-quota2-nbase:1.19 bouyer-quota2:1.19.0.12 bouyer-quota2-base:1.19 matt-nb5-pq3:1.19.0.10 matt-nb5-pq3-base:1.19 netbsd-5-1:1.19.0.8 netbsd-5-1-RELEASE:1.19 netbsd-5-1-RC4:1.19 netbsd-5-1-RC3:1.19 netbsd-5-1-RC2:1.19 netbsd-5-1-RC1:1.19 netbsd-5-0-2-RELEASE:1.19 netbsd-5-0-1-RELEASE:1.19 jym-xensuspend-nbase:1.19 netbsd-5-0:1.19.0.6 netbsd-5-0-RELEASE:1.19 netbsd-5-0-RC4:1.19 netbsd-5-0-RC3:1.19 netbsd-5-0-RC2:1.19 jym-xensuspend:1.19.0.4 jym-xensuspend-base:1.19 netbsd-5-0-RC1:1.19 netbsd-5:1.19.0.2 netbsd-5-base:1.19 matt-mips64-base2:1.18 matt-mips64:1.18.0.22 mjf-devfs2:1.18.0.20 mjf-devfs2-base:1.18 netbsd-4-0-1-RELEASE:1.18 wrstuden-revivesa-base-3:1.18 wrstuden-revivesa-base-2:1.18 wrstuden-fixsa-newbase:1.18 wrstuden-revivesa-base-1:1.18 yamt-pf42-base4:1.18 yamt-pf42-base3:1.18 hpcarm-cleanup-nbase:1.18 yamt-pf42-baseX:1.18 yamt-pf42-base2:1.18 wrstuden-revivesa:1.18.0.18 wrstuden-revivesa-base:1.18 yamt-pf42:1.18.0.16 yamt-pf42-base:1.18 keiichi-mipv6-nbase:1.18 keiichi-mipv6:1.18.0.14 keiichi-mipv6-base:1.18 matt-armv6-nbase:1.18 matt-armv6-prevmlocking:1.18 wrstuden-fixsa-base-1:1.18 netbsd-4-0:1.18.0.12 netbsd-4-0-RELEASE:1.18 cube-autoconf:1.18.0.10 cube-autoconf-base:1.18 netbsd-4-0-RC5:1.18 netbsd-4-0-RC4:1.18 netbsd-4-0-RC3:1.18 netbsd-4-0-RC2:1.18 netbsd-4-0-RC1:1.18 matt-armv6:1.18.0.8 matt-armv6-base:1.18 matt-mips64-base:1.18 hpcarm-cleanup:1.18.0.6 hpcarm-cleanup-base:1.18 netbsd-3-1-1-RELEASE:1.17 netbsd-3-0-3-RELEASE:1.17 wrstuden-fixsa:1.18.0.4 wrstuden-fixsa-base:1.18 abandoned-netbsd-4-base:1.17 abandoned-netbsd-4:1.17.0.8 netbsd-3-1:1.17.0.10 netbsd-3-1-RELEASE:1.17 netbsd-3-0-2-RELEASE:1.17 netbsd-3-1-RC4:1.17 netbsd-3-1-RC3:1.17 netbsd-3-1-RC2:1.17 netbsd-3-1-RC1:1.17 netbsd-4:1.18.0.2 netbsd-4-base:1.18 chap-midi-nbase:1.17 netbsd-3-0-1-RELEASE:1.17 chap-midi:1.17.0.6 chap-midi-base:1.17 netbsd-3-0:1.17.0.4 netbsd-3-0-RELEASE:1.17 netbsd-3-0-RC6:1.17 netbsd-3-0-RC5:1.17 netbsd-3-0-RC4:1.17 netbsd-3-0-RC3:1.17 netbsd-3-0-RC2:1.17 netbsd-3-0-RC1:1.17 netbsd-2-0-3-RELEASE:1.15.2.2 netbsd-2-1:1.15.4.1.0.2 netbsd-2-1-RELEASE:1.15.4.1 netbsd-2-1-RC6:1.15.4.1 netbsd-2-1-RC5:1.15.4.1 netbsd-2-1-RC4:1.15.4.1 netbsd-2-1-RC3:1.15.4.1 netbsd-2-1-RC2:1.15.4.1 netbsd-2-1-RC1:1.15.4.1 netbsd-2-0-2-RELEASE:1.15.2.2 netbsd-3:1.17.0.2 netbsd-3-base:1.17 netbsd-2-0-1-RELEASE:1.15 netbsd-2:1.15.0.4 netbsd-2-base:1.15 netbsd-2-0-RELEASE:1.15 netbsd-2-0-RC5:1.15 netbsd-2-0-RC4:1.15 netbsd-2-0-RC3:1.15 netbsd-2-0-RC2:1.15 netbsd-2-0-RC1:1.15 netbsd-2-0:1.15.0.2 netbsd-2-0-base:1.15 netbsd-1-6-PATCH002-RELEASE:1.14 netbsd-1-6-PATCH002:1.14 netbsd-1-6-PATCH002-RC4:1.14 netbsd-1-6-PATCH002-RC3:1.14 netbsd-1-6-PATCH002-RC2:1.14 netbsd-1-6-PATCH002-RC1:1.14 netbsd-1-6-PATCH001:1.14 netbsd-1-6-PATCH001-RELEASE:1.14 netbsd-1-6-PATCH001-RC3:1.14 netbsd-1-6-PATCH001-RC2:1.14 netbsd-1-6-PATCH001-RC1:1.14 fvdl_fs64_base:1.14 netbsd-1-6-RELEASE:1.14 netbsd-1-6-RC3:1.14 netbsd-1-6-RC2:1.14 netbsd-1-6-RC1:1.14 netbsd-1-6:1.14.0.2 netbsd-1-6-base:1.14 netbsd-1-5-PATCH003:1.12 netbsd-1-5-PATCH002:1.12 netbsd-1-5-PATCH001:1.12 nvi_1_79:1.1.1.6 netbsd-1-5-RELEASE:1.12 netbsd-1-5-BETA2:1.12 netbsd-1-5-BETA:1.12 netbsd-1-4-PATCH003:1.11 netbsd-1-5-ALPHA2:1.12 netbsd-1-5:1.12.0.4 netbsd-1-5-base:1.12 minoura-xpg4dl-base:1.12 minoura-xpg4dl:1.12.0.2 netbsd-1-4-PATCH002:1.11 wrstuden-devbsize-19991221:1.11 wrstuden-devbsize:1.11.0.6 wrstuden-devbsize-base:1.11 comdex-fall-1999:1.11.0.4 comdex-fall-1999-base:1.11 netbsd-1-4-PATCH001:1.11 netbsd-1-4-RELEASE:1.11 netbsd-1-4:1.11.0.2 netbsd-1-4-base:1.11 netbsd-1-3-PATCH003:1.10 netbsd-1-3-PATCH003-CANDIDATE2:1.10 netbsd-1-3-PATCH003-CANDIDATE1:1.10 netbsd-1-3-PATCH003-CANDIDATE0:1.10 netbsd-1-3-PATCH002:1.10 netbsd-1-3-PATCH001:1.10 netbsd-1-3-RELEASE:1.10 netbsd-1-3-BETA:1.10 netbsd-1-3:1.10.0.2 netbsd-1-3-base:1.10 netbsd-1-2-PATCH001:1.10 netbsd-1-2-RELEASE:1.10 netbsd-1-2-BETA:1.10 netbsd-1-2:1.10.0.4 netbsd-1-2-base:1.10 nvi_1_66:1.1.1.5 BOSTIC:1.1.1 netbsd-1-1-PATCH001:1.7 netbsd-1-1-RELEASE:1.7 netbsd-1-1:1.7.0.2 netbsd-1-1-base:1.7 netbsd-1-0-PATCH06:1.6 netbsd-1-0-PATCH05:1.6 netbsd-1-0-PATCH04:1.6 netbsd-1-0-PATCH03:1.6 netbsd-1-0-PATCH02:1.6 netbsd-1-0-PATCH1:1.6 netbsd-1-0-PATCH0:1.6 netbsd-1-0-RELEASE:1.6 netbsd-1-0:1.6.0.2 nvi-1-34b:1.1.1.4 nvi-1-33b:1.1.1.3 netbsd-1-0-base:1.4 nvi-1-11b:1.1.1.2 nvi-1-03:1.1.1.1 bostic-nvi:1.1.1; locks; strict; comment @ * @; 1.19 date 2008.10.29.16.49.37; author christos; state dead; branches; next 1.18; 1.18 date 2006.09.27.19.58.00; author christos; state Exp; branches; next 1.17; 1.17 date 2005.02.12.12.53.23; author aymeric; state Exp; branches; next 1.16; 1.16 date 2004.11.05.19.50.12; author dsl; state Exp; branches; next 1.15; 1.15 date 2004.02.13.17.56.18; author wiz; state Exp; branches 1.15.2.1 1.15.4.1; next 1.14; 1.14 date 2002.04.09.01.47.34; author thorpej; state Exp; branches; next 1.13; 1.13 date 2001.03.31.11.37.50; author aymeric; state Exp; branches; next 1.12; 1.12 date 2000.03.17.02.23.17; author christos; state Exp; branches; next 1.11; 1.11 date 98.01.09.08.08.05; author perry; state Exp; branches; next 1.10; 1.10 date 96.05.20.04.49.48; author cgd; state Exp; branches; next 1.9; 1.9 date 96.05.20.04.31.56; author mrg; state Exp; branches; next 1.8; 1.8 date 96.05.20.03.48.01; author mrg; state Exp; branches; next 1.7; 1.7 date 94.12.24.17.12.42; author cgd; state Exp; branches; next 1.6; 1.6 date 94.08.17.20.13.00; author cgd; state Exp; branches 1.6.2.1; next 1.5; 1.5 date 94.08.17.16.36.16; author cgd; state Exp; branches; next 1.4; 1.4 date 94.03.28.04.29.47; author cgd; state Exp; branches; next 1.3; 1.3 date 94.03.03.23.26.31; author mycroft; state Exp; branches; next 1.2; 1.2 date 94.01.24.06.40.41; author cgd; state Exp; branches; next 1.1; 1.1 date 94.01.24.05.53.07; author cgd; state Exp; branches 1.1.1.1; next ; 1.15.2.1 date 2005.02.12.12.24.13; author aymeric; state Exp; branches; next 1.15.2.2; 1.15.2.2 date 2005.02.12.12.46.26; author aymeric; state Exp; branches; next ; 1.15.4.1 date 2005.05.06.14.42.22; author riz; state Exp; branches; next ; 1.6.2.1 date 94.08.17.20.13.00; author cgd; state dead; branches; next 1.6.2.2; 1.6.2.2 date 94.08.17.20.13.01; author cgd; state Exp; branches; next ; 1.1.1.1 date 94.01.24.05.53.08; author cgd; state Exp; branches; next 1.1.1.2; 1.1.1.2 date 94.03.28.02.56.01; author cgd; state Exp; branches; next 1.1.1.3; 1.1.1.3 date 94.08.17.16.17.02; author cgd; state Exp; branches; next 1.1.1.4; 1.1.1.4 date 94.08.17.19.18.50; author cgd; state Exp; branches; next 1.1.1.5; 1.1.1.5 date 96.05.20.01.57.16; author mrg; state Exp; branches; next 1.1.1.6; 1.1.1.6 date 2001.03.31.11.29.48; author aymeric; state Exp; branches; next ; desc @@ 1.19 log @bye old vi! @ text @/* $NetBSD: ex_subst.c,v 1.18 2006/09/27 19:58:00 christos Exp $ */ /*- * Copyright (c) 1992, 1993, 1994 * The Regents of the University of California. All rights reserved. * Copyright (c) 1992, 1993, 1994, 1995, 1996 * Keith Bostic. All rights reserved. * * See the LICENSE file for redistribution information. */ #include "config.h" #include #ifndef lint #if 0 static const char sccsid[] = "@@(#)ex_subst.c 10.37 (Berkeley) 9/15/96"; #else __RCSID("$NetBSD: ex_subst.c,v 1.18 2006/09/27 19:58:00 christos Exp $"); #endif #endif /* not lint */ #include #include #include #include #include #include #include #include #include #include #include #include "../common/common.h" #include "../vi/vi.h" #define SUB_FIRST 0x01 /* The 'r' flag isn't reasonable. */ #define SUB_MUSTSETR 0x02 /* The 'r' flag is required. */ static int re_conv __P((SCR *, char **, size_t *, int *)); static int re_cscope_conv __P((SCR *, char **, size_t *, int *)); static int re_sub __P((SCR *, char *, char **, size_t *, size_t *, regmatch_t [10])); static int re_tag_conv __P((SCR *, char **, size_t *, int *)); static int s __P((SCR *, EXCMD *, char *, regex_t *, u_int)); /* * ex_s -- * [line [,line]] s[ubstitute] [[/;]pat[/;]/repl[/;] [cgr] [count] [#lp]] * * Substitute on lines matching a pattern. * * PUBLIC: int ex_s __P((SCR *, EXCMD *)); */ int ex_s(sp, cmdp) SCR *sp; EXCMD *cmdp; { regex_t *re; size_t blen, len; u_int flags; int delim; char *bp, *ptrn, *rep, *p, *t; /* * Skip leading white space. * * !!! * Historic vi allowed any non-alphanumeric to serve as the * substitution command delimiter. * * !!! * If the arguments are empty, it's the same as &, i.e. we * repeat the last substitution. */ if (cmdp->argc == 0) goto subagain; for (p = cmdp->argv[0]->bp, len = cmdp->argv[0]->len; len > 0; --len, ++p) { if (!isblank(*p)) break; } if (len == 0) subagain: return (ex_subagain(sp, cmdp)); delim = *p++; if (isalnum(delim) || delim == '\\') return (s(sp, cmdp, p, &sp->subre_c, SUB_MUSTSETR)); /* * !!! * The full-blown substitute command reset the remembered * state of the 'c' and 'g' suffixes. */ sp->c_suffix = sp->g_suffix = 0; /* * Get the pattern string, toss escaping characters. * * !!! * Historic vi accepted any of the following forms: * * :s/abc/def/ change "abc" to "def" * :s/abc/def change "abc" to "def" * :s/abc/ delete "abc" * :s/abc delete "abc" * * QUOTING NOTE: * * Only toss an escaping character if it escapes a delimiter. * This means that "s/A/\\\\f" replaces "A" with "\\f". It * would be nice to be more regular, i.e. for each layer of * escaping a single escaping character is removed, but that's * not how the historic vi worked. */ for (ptrn = t = p;;) { if (p[0] == '\0' || p[0] == delim) { if (p[0] == delim) ++p; /* * !!! * Nul terminate the pattern string -- it's passed * to regcomp which doesn't understand anything else. */ *t = '\0'; break; } if (p[0] == '\\') if (p[1] == delim) ++p; else if (p[1] == '\\') *t++ = *p++; *t++ = *p++; } /* * If the pattern string is empty, use the last RE (not just the * last substitution RE). */ if (*ptrn == '\0') { if (sp->re == NULL) { ex_emsg(sp, NULL, EXM_NOPREVRE); return (1); } /* Re-compile the RE if necessary. */ if (!F_ISSET(sp, SC_RE_SEARCH) && re_compile(sp, sp->re, sp->re_len, NULL, NULL, &sp->re_c, RE_C_SEARCH)) return (1); flags = 0; } else { /* * !!! * Compile the RE. Historic practice is that substitutes set * the search direction as well as both substitute and search * RE's. We compile the RE twice, as we don't want to bother * ref counting the pattern string and (opaque) structure. */ if (re_compile(sp, ptrn, t - ptrn, &sp->re, &sp->re_len, &sp->re_c, RE_C_SEARCH)) return (1); if (re_compile(sp, ptrn, t - ptrn, &sp->subre, &sp->subre_len, &sp->subre_c, RE_C_SUBST)) return (1); flags = SUB_FIRST; sp->searchdir = FORWARD; } re = &sp->re_c; /* * Get the replacement string. * * The special character & (\& if O_MAGIC not set) matches the * entire RE. No handling of & is required here, it's done by * re_sub(). * * The special character ~ (\~ if O_MAGIC not set) inserts the * previous replacement string into this replacement string. * Count ~'s to figure out how much space we need. We could * special case nonexistent last patterns or whether or not * O_MAGIC is set, but it's probably not worth the effort. * * QUOTING NOTE: * * Only toss an escaping character if it escapes a delimiter or * if O_MAGIC is set and it escapes a tilde. * * !!! * If the entire replacement pattern is "%", then use the last * replacement pattern. This semantic was added to vi in System * V and then percolated elsewhere, presumably around the time * that it was added to their version of ed(1). */ if (p[0] == '\0' || p[0] == delim) { if (p[0] == delim) ++p; if (sp->repl != NULL) free(sp->repl); sp->repl = NULL; sp->repl_len = 0; } else if (p[0] == '%' && (p[1] == '\0' || p[1] == delim)) p += p[1] == delim ? 2 : 1; else { for (rep = p, len = 0; p[0] != '\0' && p[0] != delim; ++p, ++len) if (p[0] == '~') len += sp->repl_len; GET_SPACE_RET(sp, bp, blen, len); for (t = bp, len = 0, p = rep;;) { if (p[0] == '\0' || p[0] == delim) { if (p[0] == delim) ++p; break; } if (p[0] == '\\') { if (p[1] == delim) ++p; else if (p[1] == '\\') { *t++ = *p++; ++len; } else if (p[1] == '~') { ++p; if (!O_ISSET(sp, O_MAGIC)) goto tilde; } } else if (p[0] == '~' && O_ISSET(sp, O_MAGIC)) { tilde: ++p; memcpy(t, sp->repl, sp->repl_len); t += sp->repl_len; len += sp->repl_len; continue; } *t++ = *p++; ++len; } if ((sp->repl_len = len) != 0) { if (sp->repl != NULL) free(sp->repl); if ((sp->repl = malloc(len)) == NULL) { msgq(sp, M_SYSERR, NULL); FREE_SPACE(sp, bp, blen); return (1); } memcpy(sp->repl, bp, len); } FREE_SPACE(sp, bp, blen); } return (s(sp, cmdp, p, re, flags)); } /* * ex_subagain -- * [line [,line]] & [cgr] [count] [#lp]] * * Substitute using the last substitute RE and replacement pattern. * * PUBLIC: int ex_subagain __P((SCR *, EXCMD *)); */ int ex_subagain(sp, cmdp) SCR *sp; EXCMD *cmdp; { if (sp->subre == NULL) { ex_emsg(sp, NULL, EXM_NOPREVRE); return (1); } if (!F_ISSET(sp, SC_RE_SUBST) && re_compile(sp, sp->subre, sp->subre_len, NULL, NULL, &sp->subre_c, RE_C_SUBST)) return (1); return (s(sp, cmdp, cmdp->argc ? cmdp->argv[0]->bp : NULL, &sp->subre_c, 0)); } /* * ex_subtilde -- * [line [,line]] ~ [cgr] [count] [#lp]] * * Substitute using the last RE and last substitute replacement pattern. * * PUBLIC: int ex_subtilde __P((SCR *, EXCMD *)); */ int ex_subtilde(sp, cmdp) SCR *sp; EXCMD *cmdp; { if (sp->re == NULL) { ex_emsg(sp, NULL, EXM_NOPREVRE); return (1); } if (!F_ISSET(sp, SC_RE_SEARCH) && re_compile(sp, sp->re, sp->re_len, NULL, NULL, &sp->re_c, RE_C_SEARCH)) return (1); return (s(sp, cmdp, cmdp->argc ? cmdp->argv[0]->bp : NULL, &sp->re_c, 0)); } /* * s -- * Do the substitution. This stuff is *really* tricky. There are lots of * special cases, and general nastiness. Don't mess with it unless you're * pretty confident. * * The nasty part of the substitution is what happens when the replacement * string contains newlines. It's a bit tricky -- consider the information * that has to be retained for "s/f\(o\)o/^M\1^M\1/". The solution here is * to build a set of newline offsets which we use to break the line up later, * when the replacement is done. Don't change it unless you're *damned* * confident. */ #define NEEDNEWLINE(sp) { \ if (sp->newl_len == sp->newl_cnt) { \ sp->newl_len += 25; \ REALLOC(sp, sp->newl, size_t *, \ sp->newl_len * sizeof(size_t)); \ if (sp->newl == NULL) { \ sp->newl_len = 0; \ return (1); \ } \ } \ } #define BUILD(sp, l, len) { \ if (lbclen + (len) > lblen) { \ lblen += MAX(lbclen + (len), 256); \ REALLOC(sp, lb, char *, lblen); \ if (lb == NULL) { \ lbclen = 0; \ return (1); \ } \ } \ memcpy(lb + lbclen, l, len); \ lbclen += len; \ } #define NEEDSP(sp, len, pnt) { \ if (lbclen + (len) > lblen) { \ lblen += MAX(lbclen + (len), 256); \ REALLOC(sp, lb, char *, lblen); \ if (lb == NULL) { \ lbclen = 0; \ return (1); \ } \ pnt = lb + lbclen; \ } \ } static int s(sp, cmdp, s, re, flags) SCR *sp; EXCMD *cmdp; char *s; regex_t *re; u_int flags; { EVENT ev; MARK from, to; TEXTH tiq; recno_t elno, lno, slno; long llno; regmatch_t match[10]; size_t blen, cnt, last, lbclen, lblen, len, llen; size_t offset, saved_offset, scno; int cflag, lflag, nflag, pflag, rflag; int didsub, do_eol_match, eflags, empty_ok, eval; int linechanged, matched, quit, rval; char *bp, *lb; #ifndef REG_STARTEND char c; #endif NEEDFILE(sp, cmdp); slno = sp->lno; scno = sp->cno; /* * !!! * Historically, the 'g' and 'c' suffices were always toggled as flags, * so ":s/A/B/" was the same as ":s/A/B/ccgg". If O_EDCOMPATIBLE was * not set, they were initialized to 0 for all substitute commands. If * O_EDCOMPATIBLE was set, they were initialized to 0 only if the user * specified substitute/replacement patterns (see ex_s()). */ if (!O_ISSET(sp, O_EDCOMPATIBLE)) sp->c_suffix = sp->g_suffix = 0; /* * Historic vi permitted the '#', 'l' and 'p' options in vi mode, but * it only displayed the last change. I'd disallow them, but they are * useful in combination with the [v]global commands. In the current * model the problem is combining them with the 'c' flag -- the screen * would have to flip back and forth between the confirm screen and the * ex print screen, which would be pretty awful. We do display all * changes, though, for what that's worth. * * !!! * Historic vi was fairly strict about the order of "options", the * count, and "flags". I'm somewhat fuzzy on the difference between * options and flags, anyway, so this is a simpler approach, and we * just take it them in whatever order the user gives them. (The ex * usage statement doesn't reflect this.) */ cflag = lflag = nflag = pflag = rflag = 0; if (s == NULL) goto noargs; for (lno = OOBLNO; *s != '\0'; ++s) switch (*s) { case ' ': case '\t': continue; case '+': ++cmdp->flagoff; break; case '-': --cmdp->flagoff; break; case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': if (lno != OOBLNO) goto usage; errno = 0; llno = strtoul(s, &s, 10); if (*s == '\0') /* Loop increment correction. */ --s; lno = llno; if (llno != lno) { errno = ERANGE; llno = LONG_MAX; } if (errno == ERANGE) { if (llno == LONG_MAX) msgq(sp, M_ERR, "153|Count overflow"); else if (llno == LONG_MIN) msgq(sp, M_ERR, "154|Count underflow"); else msgq(sp, M_SYSERR, NULL); return (1); } /* * In historic vi, the count was inclusive from the * second address. */ cmdp->addr1.lno = cmdp->addr2.lno; cmdp->addr2.lno += lno - 1; if (!db_exist(sp, cmdp->addr2.lno) && db_last(sp, &cmdp->addr2.lno)) return (1); break; case '#': nflag = 1; break; case 'c': sp->c_suffix = !sp->c_suffix; /* Ex text structure initialization. */ if (F_ISSET(sp, SC_EX)) { memset(&tiq, 0, sizeof(TEXTH)); CIRCLEQ_INIT(&tiq); } break; case 'g': sp->g_suffix = !sp->g_suffix; break; case 'l': lflag = 1; break; case 'p': pflag = 1; break; case 'r': if (LF_ISSET(SUB_FIRST)) { msgq(sp, M_ERR, "155|Regular expression specified; r flag meaningless"); return (1); } if (!F_ISSET(sp, SC_RE_SEARCH)) { ex_emsg(sp, NULL, EXM_NOPREVRE); return (1); } rflag = 1; re = &sp->re_c; break; default: goto usage; } if (*s != '\0' || !rflag && LF_ISSET(SUB_MUSTSETR)) { usage: ex_emsg(sp, cmdp->cmd->usage, EXM_USAGE); return (1); } noargs: if (F_ISSET(sp, SC_VI) && sp->c_suffix && (lflag || nflag || pflag)) { msgq(sp, M_ERR, "156|The #, l and p flags may not be combined with the c flag in vi mode"); return (1); } /* * bp: if interactive, line cache * blen: if interactive, line cache length * lb: build buffer pointer. * lbclen: current length of built buffer. * lblen; length of build buffer. */ bp = lb = NULL; blen = lbclen = lblen = 0; /* For each line... */ for (matched = quit = 0, lno = cmdp->addr1.lno, elno = cmdp->addr2.lno; !quit && lno <= elno; ++lno) { /* Someone's unhappy, time to stop. */ if (INTERRUPTED(sp)) break; /* Get the line. */ if (db_get(sp, lno, DBG_FATAL, &s, &llen)) goto err; /* * Make a local copy if doing confirmation -- when calling * the confirm routine we're likely to lose the cached copy. */ if (sp->c_suffix) { if (bp == NULL) { GET_SPACE_RET(sp, bp, blen, llen); } else ADD_SPACE_RET(sp, bp, blen, llen); if (bp) memcpy(bp, s, llen); s = bp; } /* Start searching from the beginning. */ offset = 0; len = llen; /* Reset the build buffer offset. */ lbclen = 0; /* Reset empty match flag. */ empty_ok = 1; /* * We don't want to have to do a setline if the line didn't * change -- keep track of whether or not this line changed. * If doing confirmations, don't want to keep setting the * line if change is refused -- keep track of substitutions. */ didsub = linechanged = 0; /* New line, do an EOL match. */ do_eol_match = 1; /* It's not nul terminated, but we pretend it is. */ #ifdef REG_STARTEND eflags = REG_STARTEND; #else eflags = 0; #endif /* * The search area is from s + offset to the EOL. * * Generally, match[0].rm_so is the offset of the start * of the match from the start of the search, and offset * is the offset of the start of the last search. */ nextmatch: #ifdef REG_STARTEND match[0].rm_so = 0; match[0].rm_eo = len; #else c = s[len]; s[len] = '\0'; #endif /* Get the next match. */ eval = regexec(re, (char *)s + offset, 10, match, eflags); #ifndef REG_STARTEND s[len] = c; #endif /* * There wasn't a match or if there was an error, deal with * it. If there was a previous match in this line, resolve * the changes into the database. Otherwise, just move on. */ if (eval == REG_NOMATCH) goto endmatch; if (eval != 0) { re_error(sp, eval, re); goto err; } matched = 1; /* Only the first search can match an anchored expression. */ eflags |= REG_NOTBOL; /* * !!! * It's possible to match 0-length strings -- for example, the * command s;a*;X;, when matched against the string "aabb" will * result in "XbXbX", i.e. the matches are "aa", the space * between the b's and the space between the b's and the end of * the string. There is a similar space between the beginning * of the string and the a's. The rule that we use (because vi * historically used it) is that any 0-length match, occurring * immediately after a match, is ignored. Otherwise, the above * example would have resulted in "XXbXbX". Another example is * incorrectly using " *" to replace groups of spaces with one * space. * * The way we do this is that if we just had a successful match, * the starting offset does not skip characters, and the match * is empty, ignore the match and move forward. If there's no * more characters in the string, we were attempting to match * after the last character, so quit. */ if (!empty_ok && match[0].rm_so == 0 && match[0].rm_eo == 0) { empty_ok = 1; if (len == 0) goto endmatch; BUILD(sp, s + offset, 1) ++offset; --len; goto nextmatch; } /* Confirm change. */ if (sp->c_suffix) { /* * Set the cursor position for confirmation. Note, * if we matched on a '$', the cursor may be past * the end of line. */ from.lno = to.lno = lno; from.cno = match[0].rm_so + offset; to.cno = match[0].rm_eo + offset; /* * Both ex and vi have to correct for a change before * the first character in the line. */ if (llen == 0) from.cno = to.cno = 0; if (F_ISSET(sp, SC_VI)) { /* * Only vi has to correct for a change after * the last character in the line. * * XXX * It would be nice to change the vi code so * that we could display a cursor past EOL. */ if (to.cno >= llen) to.cno = llen - 1; if (from.cno >= llen) from.cno = llen - 1; sp->lno = from.lno; sp->cno = from.cno; if (vs_refresh(sp, 1)) goto err; vs_update(sp, msg_cat(sp, "169|Confirm change? [n]", NULL), NULL); if (v_event_get(sp, &ev, 0, 0)) goto err; switch (ev.e_event) { case E_CHARACTER: break; case E_EOF: case E_ERR: case E_INTERRUPT: goto lquit; default: v_event_err(sp, &ev); goto lquit; } } else { if (ex_print(sp, cmdp, &from, &to, 0) || ex_scprint(sp, &from, &to)) goto lquit; if (ex_txt(sp, &tiq, 0, TXT_CR)) goto err; ev.e_c = tiq.cqh_first->lb[0]; } switch (ev.e_c) { case CH_YES: break; default: case CH_NO: didsub = 0; BUILD(sp, s +offset, match[0].rm_eo); goto skip; case CH_QUIT: /* Set the quit/interrupted flags. */ lquit: quit = 1; F_SET(sp->gp, G_INTERRUPTED); /* * Resolve any changes, then return to (and * exit from) the main loop. */ goto endmatch; } } /* * Set the cursor to the last position changed, converting * from 1-based to 0-based. */ sp->lno = lno; sp->cno = match[0].rm_so; /* Copy the bytes before the match into the build buffer. */ BUILD(sp, s + offset, match[0].rm_so); /* Substitute the matching bytes. */ didsub = 1; if (re_sub(sp, s + offset, &lb, &lbclen, &lblen, match)) goto err; /* Set the change flag so we know this line was modified. */ linechanged = 1; /* Move past the matched bytes. */ skip: offset += match[0].rm_eo; len -= match[0].rm_eo; /* A match cannot be followed by an empty pattern. */ empty_ok = 0; /* * If doing a global change with confirmation, we have to * update the screen. The basic idea is to store the line * so the screen update routines can find it, and restart. */ if (didsub && sp->c_suffix && sp->g_suffix) { /* * The new search offset will be the end of the * modified line. */ saved_offset = lbclen; /* Copy the rest of the line. */ if (len) BUILD(sp, s + offset, len) /* Set the new offset. */ offset = saved_offset; /* Store inserted lines, adjusting the build buffer. */ last = 0; if (sp->newl_cnt) { for (cnt = 0; cnt < sp->newl_cnt; ++cnt, ++lno, ++elno) { if (db_insert(sp, lno, lb + last, sp->newl[cnt] - last)) goto err; last = sp->newl[cnt] + 1; ++sp->rptlines[L_ADDED]; } lbclen -= last; offset -= last; sp->newl_cnt = 0; } /* Store and retrieve the line. */ if (db_set(sp, lno, lb + last, lbclen)) goto err; if (db_get(sp, lno, DBG_FATAL, &s, &llen)) goto err; ADD_SPACE_RET(sp, bp, blen, llen) memcpy(bp, s, llen); s = bp; len = llen - offset; /* Restart the build. */ lbclen = 0; BUILD(sp, s, offset); /* * If we haven't already done the after-the-string * match, do one. Set REG_NOTEOL so the '$' pattern * only matches once. */ if (!do_eol_match) goto endmatch; if (offset == len) { do_eol_match = 0; eflags |= REG_NOTEOL; } goto nextmatch; } /* * If it's a global: * * If at the end of the string, do a test for the after * the string match. Set REG_NOTEOL so the '$' pattern * only matches once. */ if (sp->g_suffix && do_eol_match) { if (len == 0) { do_eol_match = 0; eflags |= REG_NOTEOL; } goto nextmatch; } endmatch: if (!linechanged) continue; /* Copy any remaining bytes into the build buffer. */ if (len) BUILD(sp, s + offset, len) /* Store inserted lines, adjusting the build buffer. */ last = 0; if (sp->newl_cnt) { for (cnt = 0; cnt < sp->newl_cnt; ++cnt, ++lno, ++elno) { if (db_insert(sp, lno, lb + last, sp->newl[cnt] - last)) goto err; last = sp->newl[cnt] + 1; ++sp->rptlines[L_ADDED]; } lbclen -= last; sp->newl_cnt = 0; } /* Store the changed line. */ if (db_set(sp, lno, lb + last, lbclen)) goto err; /* Update changed line counter. */ if (sp->rptlchange != lno) { sp->rptlchange = lno; ++sp->rptlines[L_CHANGED]; } /* * !!! * Display as necessary. Historic practice is to only * display the last line of a line split into multiple * lines. */ if (lflag || nflag || pflag) { from.lno = to.lno = lno; from.cno = to.cno = 0; if (lflag) (void)ex_print(sp, cmdp, &from, &to, E_C_LIST); if (nflag) (void)ex_print(sp, cmdp, &from, &to, E_C_HASH); if (pflag) (void)ex_print(sp, cmdp, &from, &to, E_C_PRINT); } } /* * !!! * Historically, vi attempted to leave the cursor at the same place if * the substitution was done at the current cursor position. Otherwise * it moved it to the first non-blank of the last line changed. There * were some problems: for example, :s/$/foo/ with the cursor on the * last character of the line left the cursor on the last character, or * the & command with multiple occurrences of the matching string in the * line usually left the cursor in a fairly random position. * * We try to do the same thing, with the exception that if the user is * doing substitution with confirmation, we move to the last line about * which the user was consulted, as opposed to the last line that they * actually changed. This prevents a screen flash if the user doesn't * change many of the possible lines. */ if (!sp->c_suffix && (sp->lno != slno || sp->cno != scno)) { sp->cno = 0; (void)nonblank(sp, sp->lno, &sp->cno); } /* * If not in a global command, and nothing matched, say so. * Else, if none of the lines displayed, put something up. */ rval = 0; if (!matched) { if (!F_ISSET(sp, SC_EX_GLOBAL)) { msgq(sp, M_ERR, "157|No match found"); goto err; } } else if (!lflag && !nflag && !pflag) F_SET(cmdp, E_AUTOPRINT); if (0) { err: rval = 1; } if (bp != NULL) FREE_SPACE(sp, bp, blen); if (lb != NULL) free(lb); return (rval); } /* * re_compile -- * Compile the RE. * * PUBLIC: int re_compile __P((SCR *, * PUBLIC: char *, size_t, char **, size_t *, regex_t *, u_int)); */ int re_compile(sp, ptrn, plen, ptrnp, lenp, rep, flags) SCR *sp; char *ptrn, **ptrnp; size_t plen, *lenp; regex_t *rep; u_int flags; { size_t len; int reflags, replaced, rval; char *p; /* Set RE flags. */ reflags = 0; if (!LF_ISSET(RE_C_CSCOPE | RE_C_TAG)) { if (O_ISSET(sp, O_EXTENDED)) reflags |= REG_EXTENDED; if (O_ISSET(sp, O_IGNORECASE)) reflags |= REG_ICASE; if (O_ISSET(sp, O_ICLOWER)) { for (p = ptrn, len = plen; len > 0; ++p, --len) if (isupper((unsigned char)*p)) break; if (len == 0) reflags |= REG_ICASE; } } /* If we're replacing a saved value, clear the old one. */ if (LF_ISSET(RE_C_SEARCH) && F_ISSET(sp, SC_RE_SEARCH)) { regfree(&sp->re_c); F_CLR(sp, SC_RE_SEARCH); } if (LF_ISSET(RE_C_SUBST) && F_ISSET(sp, SC_RE_SUBST)) { regfree(&sp->subre_c); F_CLR(sp, SC_RE_SUBST); } /* * If we're saving the string, it's a pattern we haven't seen before, * so convert the vi-style RE's to POSIX 1003.2 RE's. Save a copy for * later recompilation. Free any previously saved value. */ if (ptrnp != NULL) { if (LF_ISSET(RE_C_CSCOPE)) { if (re_cscope_conv(sp, &ptrn, &plen, &replaced)) return (1); /* * XXX * Currently, the match-any- expression used in * re_cscope_conv() requires extended RE's. This may * not be right or safe. */ reflags |= REG_EXTENDED; } else if (LF_ISSET(RE_C_TAG)) { if (re_tag_conv(sp, &ptrn, &plen, &replaced)) return (1); } else if (re_conv(sp, &ptrn, &plen, &replaced)) return (1); /* Discard previous pattern. */ if (*ptrnp != NULL) { free(*ptrnp); *ptrnp = NULL; } if (lenp != NULL) *lenp = plen; /* * Copy the string into allocated memory. * * XXX * Regcomp isn't 8-bit clean, so the pattern is nul-terminated * for now. There's just no other solution. */ MALLOC(sp, *ptrnp, char *, plen + 1); if (*ptrnp != NULL) { memcpy(*ptrnp, ptrn, plen); (*ptrnp)[plen] = '\0'; } /* Free up conversion-routine-allocated memory. */ if (replaced) FREE_SPACE(sp, ptrn, 0); if (*ptrnp == NULL) return (1); ptrn = *ptrnp; } /* * XXX * Regcomp isn't 8-bit clean, so we just lost if the pattern * contained a nul. Bummer! */ if ((rval = regcomp(rep, ptrn, /* plen, */ reflags)) != 0) { if (!LF_ISSET(RE_C_SILENT)) re_error(sp, rval, rep); return (1); } if (LF_ISSET(RE_C_SEARCH)) F_SET(sp, SC_RE_SEARCH); if (LF_ISSET(RE_C_SUBST)) F_SET(sp, SC_RE_SUBST); return (0); } /* * re_conv -- * Convert vi's regular expressions into something that the * the POSIX 1003.2 RE functions can handle. * * There are three conversions we make to make vi's RE's (specifically * the global, search, and substitute patterns) work with POSIX RE's. * * 1: If O_MAGIC is not set, strip backslashes from the magic character * set (.[*~) that have them, and add them to the ones that don't. * 2: If O_MAGIC is not set, the string "\~" is replaced with the text * from the last substitute command's replacement string. If O_MAGIC * is set, it's the string "~". * 3: The pattern \ does "word" searches, convert it to use the * new RE escapes. * * !!!/XXX * This doesn't exactly match the historic behavior of vi because we do * the ~ substitution before calling the RE engine, so magic characters * in the replacement string will be expanded by the RE engine, and they * weren't historically. It's a bug. */ static int re_conv(sp, ptrnp, plenp, replacedp) SCR *sp; char **ptrnp; size_t *plenp; int *replacedp; { size_t blen, len, needlen; int magic; char *bp, *p, *t; /* * First pass through, we figure out how much space we'll need. * We do it in two passes, on the grounds that most of the time * the user is doing a search and won't have magic characters. * That way we can skip most of the memory allocation and copies. */ magic = 0; for (p = *ptrnp, len = *plenp, needlen = 0; len > 0; ++p, --len) switch (*p) { case '\\': if (len > 1) { --len; switch (*++p) { case '<': magic = 1; needlen += sizeof(RE_WSTART); break; case '>': magic = 1; needlen += sizeof(RE_WSTOP); break; case '~': if (!O_ISSET(sp, O_MAGIC)) { magic = 1; needlen += sp->repl_len; } break; case '.': case '[': case '*': if (!O_ISSET(sp, O_MAGIC)) { magic = 1; needlen += 1; } break; default: needlen += 2; } } else needlen += 1; break; case '~': if (O_ISSET(sp, O_MAGIC)) { magic = 1; needlen += sp->repl_len; } break; case '.': case '[': case '*': if (!O_ISSET(sp, O_MAGIC)) { magic = 1; needlen += 2; } break; default: needlen += 1; break; } if (!magic) { *replacedp = 0; return (0); } /* Get enough memory to hold the final pattern. */ *replacedp = 1; GET_SPACE_RET(sp, bp, blen, needlen); for (p = *ptrnp, len = *plenp, t = bp; len > 0; ++p, --len) switch (*p) { case '\\': if (len > 1) { --len; switch (*++p) { case '<': memcpy(t, RE_WSTART, sizeof(RE_WSTART) - 1); t += sizeof(RE_WSTART) - 1; break; case '>': memcpy(t, RE_WSTOP, sizeof(RE_WSTOP) - 1); t += sizeof(RE_WSTOP) - 1; break; case '~': if (O_ISSET(sp, O_MAGIC)) *t++ = '~'; else { memcpy(t, sp->repl, sp->repl_len); t += sp->repl_len; } break; case '.': case '[': case '*': if (O_ISSET(sp, O_MAGIC)) *t++ = '\\'; *t++ = *p; break; default: *t++ = '\\'; *t++ = *p; } } else *t++ = '\\'; break; case '~': if (O_ISSET(sp, O_MAGIC)) { memcpy(t, sp->repl, sp->repl_len); t += sp->repl_len; } else *t++ = '~'; break; case '.': case '[': case '*': if (!O_ISSET(sp, O_MAGIC)) *t++ = '\\'; *t++ = *p; break; default: *t++ = *p; break; } *ptrnp = bp; *plenp = t - bp; return (0); } /* * re_tag_conv -- * Convert a tags search path into something that the POSIX * 1003.2 RE functions can handle. */ static int re_tag_conv(sp, ptrnp, plenp, replacedp) SCR *sp; char **ptrnp; size_t *plenp; int *replacedp; { size_t blen, len; int lastdollar; char *bp, *p, *t; len = *plenp; /* Max memory usage is 2 times the length of the string. */ *replacedp = 1; GET_SPACE_RET(sp, bp, blen, len * 2); p = *ptrnp; t = bp; /* If the last character is a '/' or '?', we just strip it. */ if (len > 0 && (p[len - 1] == '/' || p[len - 1] == '?')) --len; /* If the next-to-last or last character is a '$', it's magic. */ if (len > 0 && p[len - 1] == '$') { --len; lastdollar = 1; } else lastdollar = 0; /* If the first character is a '/' or '?', we just strip it. */ if (len > 0 && (p[0] == '/' || p[0] == '?')) { ++p; --len; } /* If the first or second character is a '^', it's magic. */ if (p[0] == '^') { *t++ = *p++; --len; } /* * Escape every other magic character we can find, meanwhile stripping * the backslashes ctags inserts when escaping the search delimiter * characters. */ for (; len > 0; --len) { if (p[0] == '\\' && (p[1] == '/' || p[1] == '?')) { ++p; --len; } else if (strchr("^.[]$*", p[0])) *t++ = '\\'; *t++ = *p++; } if (lastdollar) *t++ = '$'; *ptrnp = bp; *plenp = t - bp; return (0); } /* * re_cscope_conv -- * Convert a cscope search path into something that the POSIX * 1003.2 RE functions can handle. */ static int re_cscope_conv(sp, ptrnp, plenp, replacedp) SCR *sp; char **ptrnp; size_t *plenp; int *replacedp; { size_t blen, len, nspaces; char *bp, *p, *t; /* * Each space in the source line printed by cscope represents an * arbitrary sequence of spaces, tabs, and comments. */ #define CSCOPE_RE_SPACE "([ \t]|/\\*([^*]|\\*/)*\\*/)*" for (nspaces = 0, p = *ptrnp, len = *plenp; len > 0; ++p, --len) if (*p == ' ') ++nspaces; /* * Allocate plenty of space: * the string, plus potential escaping characters; * nspaces + 2 copies of CSCOPE_RE_SPACE; * ^, $, nul terminator characters. */ *replacedp = 1; len = (p - *ptrnp) * 2 + (nspaces + 2) * sizeof(CSCOPE_RE_SPACE) + 3; GET_SPACE_RET(sp, bp, blen, len); p = *ptrnp; t = bp; *t++ = '^'; memcpy(t, CSCOPE_RE_SPACE, sizeof(CSCOPE_RE_SPACE) - 1); t += sizeof(CSCOPE_RE_SPACE) - 1; for (len = *plenp; len > 0; ++p, --len) if (*p == ' ') { memcpy(t, CSCOPE_RE_SPACE, sizeof(CSCOPE_RE_SPACE) - 1); t += sizeof(CSCOPE_RE_SPACE) - 1; } else { if (strchr("\\^.[]$*+?()|{}", *p)) *t++ = '\\'; *t++ = *p; } memcpy(t, CSCOPE_RE_SPACE, sizeof(CSCOPE_RE_SPACE) - 1); t += sizeof(CSCOPE_RE_SPACE) - 1; *t++ = '$'; *ptrnp = bp; *plenp = t - bp; return (0); } /* * re_error -- * Report a regular expression error. * * PUBLIC: void re_error __P((SCR *, int, regex_t *)); */ void re_error(sp, errcode, preg) SCR *sp; int errcode; regex_t *preg; { size_t s; char *oe; s = regerror(errcode, preg, "", 0); if ((oe = malloc(s)) == NULL) msgq(sp, M_SYSERR, NULL); else { (void)regerror(errcode, preg, oe, s); msgq(sp, M_ERR, "RE error: %s", oe); free(oe); } } /* * re_sub -- * Do the substitution for a regular expression. */ static int re_sub(sp, ip, lbp, lbclenp, lblenp, match) SCR *sp; char *ip; /* Input line. */ char **lbp; size_t *lbclenp, *lblenp; regmatch_t match[10]; { enum { C_NOTSET, C_LOWER, C_ONELOWER, C_ONEUPPER, C_UPPER } conv; size_t lbclen, lblen; /* Local copies. */ size_t mlen; /* Match length. */ size_t rpl; /* Remaining replacement length. */ char *rp; /* Replacement pointer. */ int ch; int no; /* Match replacement offset. */ char *p, *t; /* Buffer pointers. */ char *lb; /* Local copies. */ lb = *lbp; /* Get local copies. */ lbclen = *lbclenp; lblen = *lblenp; /* * QUOTING NOTE: * * There are some special sequences that vi provides in the * replacement patterns. * & string the RE matched (\& if nomagic set) * \# n-th regular subexpression * \E end \U, \L conversion * \e end \U, \L conversion * \l convert the next character to lower-case * \L convert to lower-case, until \E, \e, or end of replacement * \u convert the next character to upper-case * \U convert to upper-case, until \E, \e, or end of replacement * * Otherwise, since this is the lowest level of replacement, discard * all escaping characters. This (hopefully) matches historic practice. */ #define OUTCH(ch, nltrans) { \ CHAR_T __ch = (ch); \ u_int __value = KEY_VAL(sp, __ch); \ if (nltrans && (__value == K_CR || __value == K_NL)) { \ NEEDNEWLINE(sp); \ sp->newl[sp->newl_cnt++] = lbclen; \ } else if (conv != C_NOTSET) { \ switch (conv) { \ case C_ONELOWER: \ conv = C_NOTSET; \ /* FALLTHROUGH */ \ case C_LOWER: \ if (isupper(__ch)) \ __ch = tolower(__ch); \ break; \ case C_ONEUPPER: \ conv = C_NOTSET; \ /* FALLTHROUGH */ \ case C_UPPER: \ if (islower(__ch)) \ __ch = toupper(__ch); \ break; \ default: \ abort(); \ } \ } \ NEEDSP(sp, 1, p); \ *p++ = __ch; \ ++lbclen; \ } conv = C_NOTSET; for (rp = sp->repl, rpl = sp->repl_len, p = lb + lbclen; rpl--;) { switch (ch = *rp++) { case '&': if (O_ISSET(sp, O_MAGIC)) { no = 0; goto subzero; } break; case '\\': if (rpl == 0) break; --rpl; switch (ch = *rp) { case '&': ++rp; if (!O_ISSET(sp, O_MAGIC)) { no = 0; goto subzero; } break; case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': no = *rp++ - '0'; subzero: if (match[no].rm_so == -1 || match[no].rm_eo == -1) break; mlen = match[no].rm_eo - match[no].rm_so; for (t = ip + match[no].rm_so; mlen--; ++t) OUTCH(*t, 0); continue; case 'e': case 'E': ++rp; conv = C_NOTSET; continue; case 'l': ++rp; conv = C_ONELOWER; continue; case 'L': ++rp; conv = C_LOWER; continue; case 'u': ++rp; conv = C_ONEUPPER; continue; case 'U': ++rp; conv = C_UPPER; continue; default: ++rp; break; } } OUTCH(ch, 1); } *lbp = lb; /* Update caller's information. */ *lbclenp = lbclen; *lblenp = lblen; return (0); } @ 1.18 log @Coverity CID 4171: bp can be null if len == 0. So check for it anyway and don't call memcpy unconditionally. @ text @d1 1 a1 1 /* $NetBSD: ex_subst.c,v 1.17 2005/02/12 12:53:23 aymeric Exp $ */ d19 1 a19 1 __RCSID("$NetBSD: ex_subst.c,v 1.17 2005/02/12 12:53:23 aymeric Exp $"); @ 1.17 log @Fix the RCSID's to be $NetBSD$ instead of $NetBSD @ text @d1 1 a1 1 /* $NetBSD: ex_subst.c,v 1.16 2004/11/05 19:50:12 dsl Exp $ */ d19 1 a19 1 __RCSID("$NetBSD$"); d535 2 a536 1 memcpy(bp, s, llen); @ 1.16 log @Add (unsigned char) cast to ctype functions @ text @d1 1 a1 1 /* $NetBSD: ex_subst.c,v 1.15 2004/02/13 17:56:18 wiz Exp $ */ d19 1 a19 1 __RCSID("$NetBSD"); @ 1.15 log @Spell the plural of suffix "suffixes", not "suffices". Inspired by PR 24400 by Todd Vierling. @ text @d1 1 a1 1 /* $NetBSD: ex_subst.c,v 1.14 2002/04/09 01:47:34 thorpej Exp $ */ d942 1 a942 1 if (isupper(*p)) @ 1.15.4.1 log @Pull up revision 1.17 (requested by aymeric in ticket #1195): Fix the RCSID's to be $NetBSD$ instead of $NetBSD @ text @d1 1 a1 1 /* $NetBSD: ex_subst.c,v 1.15 2004/02/13 17:56:18 wiz Exp $ */ d19 1 a19 1 __RCSID("$NetBSD$"); @ 1.15.2.1 log @Fix the RCSID's to be $NetBSD$ instead of $NetBSD @ text @d1 1 a1 1 /* $NetBSD: ex_subst.c,v 1.15 2004/02/13 17:56:18 wiz Exp $ */ d19 1 a19 1 __RCSID("$NetBSD$"); @ 1.15.2.2 log @Backout previous. Sorry. @ text @d19 1 a19 1 __RCSID("$NetBSD"); @ 1.14 log @Use __RCSID() and __COPYRIGHT(). @ text @d1 1 a1 1 /* $NetBSD: ex_subst.c,v 1.13 2001/03/31 11:37:50 aymeric Exp $ */ d96 1 a96 1 * state of the 'c' and 'g' suffices. @ 1.13 log @merge changes after import of nvi 1.79 @ text @d1 1 a1 1 /* $NetBSD: ex_subst.c,v 1.12 2000/03/17 02:23:17 christos Exp $ */ d14 1 d16 1 d18 3 @ 1.12 log @Don't depend on BSD regexec extensions (REG_STARTEND) Emulate it if it is not defined. @ text @d1 1 a1 1 /* $NetBSD: ex_subst.c,v 1.11 1998/01/09 08:08:05 perry Exp $ */ d15 1 a15 1 static const char sccsid[] = "@@(#)ex_subst.c 10.30 (Berkeley) 5/16/96"; d37 2 a38 2 static int re_conv __P((SCR *, char **, int *)); static int re_cscope_conv __P((SCR *, char **, int *)); d41 1 a41 1 static int re_tag_conv __P((SCR *, char **, int *)); d96 1 a96 1 * Get the pattern string, toss escaped characters. d108 1 a108 1 * Only toss an escape character if it escapes a delimiter. d111 1 a111 1 * escaping a single escape character is removed, but that's d144 3 a146 3 /* Compile the RE if necessary. */ if (!F_ISSET(sp, SC_RE_SEARCH) && re_compile(sp, sp->re, NULL, NULL, &sp->re_c, RE_C_SEARCH)) d157 2 a158 2 if (re_compile(sp, ptrn, &sp->re, &sp->re_len, &sp->re_c, RE_C_SEARCH)) d160 2 a161 2 if (re_compile(sp, ptrn, &sp->subre, &sp->subre_len, &sp->subre_c, RE_C_SUBST)) d184 1 a184 1 * Only toss an escape character if it escapes a delimiter or d227 1 a227 1 memmove(t, sp->repl, sp->repl_len); d243 1 a243 1 memmove(sp->repl, bp, len); d267 2 a268 2 if (!F_ISSET(sp, SC_RE_SUBST) && re_compile(sp, sp->subre, NULL, NULL, &sp->subre_c, RE_C_SUBST)) d291 2 a292 2 if (!F_ISSET(sp, SC_RE_SEARCH) && re_compile(sp, sp->re, NULL, NULL, &sp->re_c, RE_C_SEARCH)) d332 1 a332 1 memmove(lb + lbclen, l, len); \ d530 1 a530 1 memmove(bp, s, llen); d569 1 a569 1 nextmatch: d574 2 a575 2 c = ((char *)s)[len]; ((char *)s)[len] = '\0'; d577 1 d581 1 a581 1 ((char *)s)[len] = c; a583 1 d777 1 a777 1 memmove(bp, s, llen); d914 1 a914 1 * PUBLIC: char *, char **, size_t *, regex_t *, u_int)); d917 1 a917 1 re_compile(sp, ptrn, ptrnp, lenp, rep, flags) d920 1 a920 1 size_t *lenp; d936 1 a936 1 for (p = ptrn; *p != '\0'; ++p) d939 1 a939 1 if (*p == '\0') a958 1 replaced = 0; a959 4 if (*ptrnp != NULL) { free(*ptrnp); *ptrnp = NULL; } d961 1 a961 1 if (re_cscope_conv(sp, &ptrn, &replaced)) d971 1 a971 1 if (re_tag_conv(sp, &ptrn, &replaced)) d974 1 a974 1 if (re_conv(sp, &ptrn, &replaced)) d976 6 a981 1 len = strlen(ptrn); d983 20 a1002 2 *lenp = len; if ((*ptrnp = v_strdup(sp, ptrn, len)) == NULL) d1004 2 d1008 6 a1013 7 rval = regcomp(rep, ptrn, reflags); /* Free up any allocated memory. */ if (replaced) FREE_SPACE(sp, ptrn, 0); if (rval) { d1050 1 a1050 1 re_conv(sp, ptrnp, replacedp) d1053 1 d1056 1 a1056 1 size_t blen, needlen; d1064 1 a1064 1 * That way we can skip the malloc and memmove's. d1066 2 a1067 1 for (p = *ptrnp, magic = 0, needlen = 0; *p != '\0'; ++p) d1070 4 a1073 11 switch (*++p) { case '<': magic = 1; needlen += sizeof(RE_WSTART); break; case '>': magic = 1; needlen += sizeof(RE_WSTOP); break; case '~': if (!O_ISSET(sp, O_MAGIC)) { d1075 3 a1077 7 needlen += sp->repl_len; } break; case '.': case '[': case '*': if (!O_ISSET(sp, O_MAGIC)) { d1079 18 a1096 1 needlen += 1; d1098 2 a1099 4 break; default: needlen += 2; } d1125 3 a1127 7 /* * Get enough memory to hold the final pattern. * * XXX * It's nul-terminated, for now. */ GET_SPACE_RET(sp, bp, blen, needlen + 1); d1129 1 a1129 1 for (p = *ptrnp, t = bp; *p != '\0'; ++p) d1132 32 a1163 15 switch (*++p) { case '<': memmove(t, RE_WSTART, sizeof(RE_WSTART) - 1); t += sizeof(RE_WSTART) - 1; break; case '>': memmove(t, RE_WSTOP, sizeof(RE_WSTOP) - 1); t += sizeof(RE_WSTOP) - 1; break; case '~': if (O_ISSET(sp, O_MAGIC)) *t++ = '~'; else { memmove(t, sp->repl, sp->repl_len); t += sp->repl_len; d1165 1 a1165 9 break; case '.': case '[': case '*': if (O_ISSET(sp, O_MAGIC)) *t++ = '\\'; *t++ = *p; break; default: a1166 2 *t++ = *p; } d1170 1 a1170 1 memmove(t, sp->repl, sp->repl_len); a1185 1 *t = '\0'; d1188 1 a1188 1 *replacedp = 1; d1198 1 a1198 1 re_tag_conv(sp, ptrnp, replacedp) d1201 1 d1208 1 a1208 3 *replacedp = 0; len = strlen(p = *ptrnp); d1211 1 d1214 1 d1217 7 a1223 6 /* The last character is a '/' or '?', we just strip it. */ if (p[len - 1] == '/' || p[len - 1] == '?') p[len - 1] = '\0'; /* The next-to-last character is a '$', and it's magic. */ if (p[len - 2] == '$') { a1224 1 p[len - 2] = '\0'; d1228 2 a1229 2 /* The first character is a '/' or '?', we just strip it. */ if (p[0] == '/' || p[0] == '?') d1231 2 d1234 2 a1235 2 /* The second character is a '^', and it's magic. */ if (p[0] == '^') d1237 2 d1241 2 a1242 2 * Escape every other magic character we can find, stripping the * backslashes ctags inserts to escape the search delimiter d1245 2 a1246 3 while (p[0]) { /* Ctags escapes the search delimiter characters. */ if (p[0] == '\\' && (p[1] == '/' || p[1] == '?')) d1248 2 a1249 1 else if (strchr("^.[]$*", p[0])) a1254 1 *t++ = '\0'; d1257 1 a1257 1 *replacedp = 1; d1267 1 a1267 1 re_cscope_conv(sp, ptrnp, replacedp) d1270 1 d1274 1 a1274 1 char *bp, *p, *re; d1281 1 a1281 1 for (nspaces = 0, p = *ptrnp; *p != '\0'; ++p) d1287 3 a1289 3 * the string, plus potential escaping characters * nspaces + 2 copies of CSCOPE_RE_SPACE * ^, $, nul terminator characters d1291 1 d1295 11 a1305 9 p = bp; *p++ = '^'; memcpy(p, CSCOPE_RE_SPACE, sizeof(CSCOPE_RE_SPACE) - 1); p += sizeof(CSCOPE_RE_SPACE) - 1; for (re = *ptrnp; *re != '\0'; ++re) if (*re == ' ') { memcpy(p, CSCOPE_RE_SPACE, sizeof(CSCOPE_RE_SPACE) - 1); p += sizeof(CSCOPE_RE_SPACE) - 1; d1307 3 a1309 3 if (strchr("\\^.[]$*", *re)) *p++ = '\\'; *p++ = *re; d1312 3 a1314 4 memcpy(p, CSCOPE_RE_SPACE, sizeof(CSCOPE_RE_SPACE) - 1); p += sizeof(CSCOPE_RE_SPACE) - 1; *p++ = '$'; *p = '\0'; d1317 1 a1317 1 *replacedp = 1; d1387 1 a1387 1 * all escape characters. This (hopefully) follows historic practice. d1389 1 a1389 1 #define OUTCH(ch) { \ d1392 1 a1392 1 if (__value == K_CR || __value == K_NL) { \ d1448 1 a1448 1 OUTCH(*t); d1476 1 a1476 1 OUTCH(ch); @ 1.11 log @RCS Id Police. @ text @d1 1 a1 1 /* $NetBSD$ */ d368 3 d556 1 d558 3 d569 3 a571 1 nextmatch: match[0].rm_so = 0; d573 4 a576 1 d579 4 @ 1.10 log @a better (more careful) fix for problem fixed by rev 1.9 (the rolled-forward patch to fix a 'comparison can never be true due to limited range...' compiler warning). @ text @d1 2 @ 1.9 log @merge in fix lost in upgrade. @ text @d357 2 a358 2 recno_t elno, slno; long lno; d418 1 a418 1 lno = strtoul(s, &s, 10); d421 5 d427 1 a427 1 if (lno == LONG_MAX) d429 1 a429 1 else if (lno == LONG_MIN) @ 1.8 log @merge in nvi 1.66 @ text @d357 2 a358 1 recno_t elno, lno, slno; @ 1.7 log @use the correct-sized type for lno. @ text @d4 2 d7 1 a7 27 * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. d10 2 d13 1 a13 1 static char sccsid[] = "@@(#)ex_subst.c 8.59 (Berkeley) 8/17/94"; a23 1 #include a26 1 #include d29 2 a30 6 #include "compat.h" #include #include #include "vi.h" #include "excmd.h" d35 6 a40 4 static __inline int regsub __P((SCR *, char *, char **, size_t *, size_t *, regmatch_t [10])); static int substitute __P((SCR *, EXF *, EXCMDARG *, char *, regex_t *, u_int)); d43 1 a43 1 * ex_substitute -- d47 2 d51 1 a51 1 ex_substitute(sp, ep, cmdp) d53 1 a53 2 EXF *ep; EXCMDARG *cmdp; d55 1 a55 1 regex_t *re, lre; d58 1 a58 1 int delim, eval, reflags, replaced; d72 2 d80 2 a81 1 return (ex_subagain(sp, ep, cmdp)); d83 2 a84 3 if (isalnum(delim)) return (substitute(sp, ep, cmdp, p, &sp->subre, SUB_MUSTSETR)); d92 1 a92 1 d137 2 a138 2 if (!F_ISSET(sp, S_SRE_SET)) { msgq(sp, M_ERR, "No previous regular expression"); d141 5 a145 1 re = &sp->sre; a147 23 /* Set RE flags. */ reflags = 0; if (O_ISSET(sp, O_EXTENDED)) reflags |= REG_EXTENDED; if (O_ISSET(sp, O_IGNORECASE)) reflags |= REG_ICASE; /* Convert vi-style RE's to POSIX 1003.2 RE's. */ if (re_conv(sp, &ptrn, &replaced)) return (1); /* Compile the RE. */ eval = regcomp(&lre, (char *)ptrn, reflags); /* Free up any allocated memory. */ if (replaced) FREE_SPACE(sp, ptrn, 0); if (eval) { re_error(sp, eval, &lre); return (1); } a148 2 * Set saved RE. * d150 4 a153 2 * Historic practice is that substitutes set the search * direction as well as both substitute and search RE's. d155 8 a163 7 sp->sre = lre; F_SET(sp, S_SRE_SET); sp->subre = lre; F_SET(sp, S_SUBRE_SET); re = &lre; flags = SUB_FIRST; d165 1 d172 1 a172 1 * regsub(). d195 1 a195 1 FREE(sp->repl, sp->repl_len); d245 1 a245 1 return (substitute(sp, ep, cmdp, p, re, flags)); d253 2 d257 1 a257 1 ex_subagain(sp, ep, cmdp) d259 1 a259 2 EXF *ep; EXCMDARG *cmdp; d261 2 a262 2 if (!F_ISSET(sp, S_SUBRE_SET)) { msgq(sp, M_ERR, "No previous regular expression"); d265 5 a269 1 return (substitute(sp, ep, cmdp, cmdp->argv[0]->bp, &sp->subre, 0)); d277 2 d281 1 a281 1 ex_subtilde(sp, ep, cmdp) d283 1 a283 2 EXF *ep; EXCMDARG *cmdp; d285 2 a286 2 if (!F_ISSET(sp, S_SRE_SET)) { msgq(sp, M_ERR, "No previous regular expression"); d289 5 a293 1 return (substitute(sp, ep, cmdp, cmdp->argv[0]->bp, &sp->sre, 0)); d297 5 d306 1 a306 1 * when the replacement is done. Don't change it unless you're pretty damned a345 6 /* * substitute -- * Do the substitution. This stuff is *really* tricky. There are * lots of special cases, and general nastiness. Don't mess with it * unless you're pretty confident. */ d347 1 a347 1 substitute(sp, ep, cmdp, s, re, flags) d349 1 a349 2 EXF *ep; EXCMDARG *cmdp; d354 1 d356 2 a357 2 recno_t elno; long lno; d359 2 a360 1 size_t blen, cnt, last, lbclen, lblen, len, llen, offset, saved_offset; d366 5 d377 1 a377 1 * specified substitute/replacement patterns (see ex_substitute()). d399 2 d422 1 a422 1 msgq(sp, M_ERR, "Count overflow"); d424 1 a424 1 msgq(sp, M_ERR, "Count underflow"); d435 3 d444 6 d463 1 a463 1 "Regular expression specified; r flag meaningless"); d466 2 a467 3 if (!F_ISSET(sp, S_SRE_SET)) { msgq(sp, M_ERR, "No previous regular expression"); d471 1 a471 1 re = &sp->sre; d478 1 a478 1 usage: msgq(sp, M_ERR, "Usage: %s", cmdp->cmd->usage); d482 1 a482 1 if (IN_VI_MODE(sp) && sp->c_suffix && (lflag || nflag || pflag)) { d484 1 a484 1 "The #, l and p flags may not be combined with the c flag in vi mode"); d503 1 a503 3 if (INTERRUPTED(sp)) { if (!F_ISSET(sp, S_GLOBAL)) msgq(sp, M_INFO, "Interrupted"); a504 1 } d507 2 a508 4 if ((s = file_gline(sp, ep, lno, &llen)) == NULL) { GETLINE_ERR(sp, lno); goto ret1; } d569 1 a569 1 goto ret1; a611 5 * * XXX * We may want to "fix" this in the confirm routine, * if the confirm routine should be able to display * a cursor past EOL. d615 5 a619 1 to.cno = match[0].rm_eo; d622 9 a630 1 else { d635 29 d665 3 a667 2 switch (sp->s_confirm(sp, ep, &from, &to)) { case CONF_YES: d669 2 a670 1 case CONF_NO: d674 4 a677 7 case CONF_QUIT: /* Set the quit flag. */ quit = 1; /* If interruptible, pass the info back. */ if (F_ISSET(sp, S_INTERRUPTIBLE)) F_SET(sp, S_INTERRUPTED); d680 2 a681 2 * If any changes, resolve them, otherwise * return to the main loop. d687 7 d699 2 a700 2 if (regsub(sp, s + offset, &lb, &lbclen, &lblen, match)) goto ret1; d736 1 a736 1 if (file_iline(sp, ep, lno, d738 1 a738 1 goto ret1; d748 4 a751 6 if (file_sline(sp, ep, lno, lb + last, lbclen)) goto ret1; if ((s = file_gline(sp, ep, lno, &llen)) == NULL) { GETLINE_ERR(sp, lno); goto ret1; } d802 1 a802 1 if (file_iline(sp, ep, d804 1 a804 1 goto ret1; d813 2 a814 2 if (file_sline(sp, ep, lno, lb + last, lbclen)) goto ret1; d832 1 a832 1 ex_print(sp, ep, &from, &to, E_F_LIST); d834 1 a834 1 ex_print(sp, ep, &from, &to, E_F_HASH); d836 1 a836 1 ex_print(sp, ep, &from, &to, E_F_PRINT); a837 10 if (!sp->c_suffix) sp->lno = lno; /* * !!! * Move the cursor to the last line changed. */ if (!sp->c_suffix) sp->lno = lno; d842 7 a848 1 * Move the cursor to the first non-blank of the last line change. d850 5 a854 3 * XXX * This is NOT backward compatible with historic vi, which always * moved to the last line actually changed. d856 1 a856 1 if (!sp->c_suffix) { d858 1 a858 1 (void)nonblank(sp, ep, sp->lno, &sp->cno); d865 1 d867 4 a870 2 if (!F_ISSET(sp, S_GLOBAL)) msgq(sp, M_INFO, "No match found"); d872 1 a872 1 F_SET(EXP(sp), EX_AUTOPRINT); a873 1 rval = 0; d875 1 a875 1 ret1: rval = 1; d886 404 a1289 1 * regsub -- d1292 2 a1293 2 static __inline int regsub(sp, ip, lbp, lbclenp, lblenp, match) d1331 1 a1331 1 #define ADDCH(ch) { \ d1390 1 a1390 1 ADDCH(*t); d1418 1 a1418 1 ADDCH(ch); @ 1.6 log @clean up import. @ text @d387 2 a388 1 recno_t elno, lno; @ 1.6.2.1 log @file ex_subst.c was added on branch netbsd-1-0 on 1994-08-17 20:13:01 +0000 @ text @d1 1001 @ 1.6.2.2 log @clean up import. @ text @a0 1001 /*- * Copyright (c) 1992, 1993, 1994 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef lint static char sccsid[] = "@@(#)ex_subst.c 8.59 (Berkeley) 8/17/94"; #endif /* not lint */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include "compat.h" #include #include #include "vi.h" #include "excmd.h" #define SUB_FIRST 0x01 /* The 'r' flag isn't reasonable. */ #define SUB_MUSTSETR 0x02 /* The 'r' flag is required. */ static __inline int regsub __P((SCR *, char *, char **, size_t *, size_t *, regmatch_t [10])); static int substitute __P((SCR *, EXF *, EXCMDARG *, char *, regex_t *, u_int)); /* * ex_substitute -- * [line [,line]] s[ubstitute] [[/;]pat[/;]/repl[/;] [cgr] [count] [#lp]] * * Substitute on lines matching a pattern. */ int ex_substitute(sp, ep, cmdp) SCR *sp; EXF *ep; EXCMDARG *cmdp; { regex_t *re, lre; size_t blen, len; u_int flags; int delim, eval, reflags, replaced; char *bp, *ptrn, *rep, *p, *t; /* * Skip leading white space. * * !!! * Historic vi allowed any non-alphanumeric to serve as the * substitution command delimiter. * * !!! * If the arguments are empty, it's the same as &, i.e. we * repeat the last substitution. */ for (p = cmdp->argv[0]->bp, len = cmdp->argv[0]->len; len > 0; --len, ++p) { if (!isblank(*p)) break; } if (len == 0) return (ex_subagain(sp, ep, cmdp)); delim = *p++; if (isalnum(delim)) return (substitute(sp, ep, cmdp, p, &sp->subre, SUB_MUSTSETR)); /* * !!! * The full-blown substitute command reset the remembered * state of the 'c' and 'g' suffices. */ sp->c_suffix = sp->g_suffix = 0; /* * Get the pattern string, toss escaped characters. * * !!! * Historic vi accepted any of the following forms: * * :s/abc/def/ change "abc" to "def" * :s/abc/def change "abc" to "def" * :s/abc/ delete "abc" * :s/abc delete "abc" * * QUOTING NOTE: * * Only toss an escape character if it escapes a delimiter. * This means that "s/A/\\\\f" replaces "A" with "\\f". It * would be nice to be more regular, i.e. for each layer of * escaping a single escape character is removed, but that's * not how the historic vi worked. */ for (ptrn = t = p;;) { if (p[0] == '\0' || p[0] == delim) { if (p[0] == delim) ++p; /* * !!! * Nul terminate the pattern string -- it's passed * to regcomp which doesn't understand anything else. */ *t = '\0'; break; } if (p[0] == '\\') if (p[1] == delim) ++p; else if (p[1] == '\\') *t++ = *p++; *t++ = *p++; } /* * If the pattern string is empty, use the last RE (not just the * last substitution RE). */ if (*ptrn == '\0') { if (!F_ISSET(sp, S_SRE_SET)) { msgq(sp, M_ERR, "No previous regular expression"); return (1); } re = &sp->sre; flags = 0; } else { /* Set RE flags. */ reflags = 0; if (O_ISSET(sp, O_EXTENDED)) reflags |= REG_EXTENDED; if (O_ISSET(sp, O_IGNORECASE)) reflags |= REG_ICASE; /* Convert vi-style RE's to POSIX 1003.2 RE's. */ if (re_conv(sp, &ptrn, &replaced)) return (1); /* Compile the RE. */ eval = regcomp(&lre, (char *)ptrn, reflags); /* Free up any allocated memory. */ if (replaced) FREE_SPACE(sp, ptrn, 0); if (eval) { re_error(sp, eval, &lre); return (1); } /* * Set saved RE. * * !!! * Historic practice is that substitutes set the search * direction as well as both substitute and search RE's. */ sp->searchdir = FORWARD; sp->sre = lre; F_SET(sp, S_SRE_SET); sp->subre = lre; F_SET(sp, S_SUBRE_SET); re = &lre; flags = SUB_FIRST; } /* * Get the replacement string. * * The special character & (\& if O_MAGIC not set) matches the * entire RE. No handling of & is required here, it's done by * regsub(). * * The special character ~ (\~ if O_MAGIC not set) inserts the * previous replacement string into this replacement string. * Count ~'s to figure out how much space we need. We could * special case nonexistent last patterns or whether or not * O_MAGIC is set, but it's probably not worth the effort. * * QUOTING NOTE: * * Only toss an escape character if it escapes a delimiter or * if O_MAGIC is set and it escapes a tilde. * * !!! * If the entire replacement pattern is "%", then use the last * replacement pattern. This semantic was added to vi in System * V and then percolated elsewhere, presumably around the time * that it was added to their version of ed(1). */ if (p[0] == '\0' || p[0] == delim) { if (p[0] == delim) ++p; if (sp->repl != NULL) FREE(sp->repl, sp->repl_len); sp->repl = NULL; sp->repl_len = 0; } else if (p[0] == '%' && (p[1] == '\0' || p[1] == delim)) p += p[1] == delim ? 2 : 1; else { for (rep = p, len = 0; p[0] != '\0' && p[0] != delim; ++p, ++len) if (p[0] == '~') len += sp->repl_len; GET_SPACE_RET(sp, bp, blen, len); for (t = bp, len = 0, p = rep;;) { if (p[0] == '\0' || p[0] == delim) { if (p[0] == delim) ++p; break; } if (p[0] == '\\') { if (p[1] == delim) ++p; else if (p[1] == '\\') { *t++ = *p++; ++len; } else if (p[1] == '~') { ++p; if (!O_ISSET(sp, O_MAGIC)) goto tilde; } } else if (p[0] == '~' && O_ISSET(sp, O_MAGIC)) { tilde: ++p; memmove(t, sp->repl, sp->repl_len); t += sp->repl_len; len += sp->repl_len; continue; } *t++ = *p++; ++len; } if ((sp->repl_len = len) != 0) { if (sp->repl != NULL) free(sp->repl); if ((sp->repl = malloc(len)) == NULL) { msgq(sp, M_SYSERR, NULL); FREE_SPACE(sp, bp, blen); return (1); } memmove(sp->repl, bp, len); } FREE_SPACE(sp, bp, blen); } return (substitute(sp, ep, cmdp, p, re, flags)); } /* * ex_subagain -- * [line [,line]] & [cgr] [count] [#lp]] * * Substitute using the last substitute RE and replacement pattern. */ int ex_subagain(sp, ep, cmdp) SCR *sp; EXF *ep; EXCMDARG *cmdp; { if (!F_ISSET(sp, S_SUBRE_SET)) { msgq(sp, M_ERR, "No previous regular expression"); return (1); } return (substitute(sp, ep, cmdp, cmdp->argv[0]->bp, &sp->subre, 0)); } /* * ex_subtilde -- * [line [,line]] ~ [cgr] [count] [#lp]] * * Substitute using the last RE and last substitute replacement pattern. */ int ex_subtilde(sp, ep, cmdp) SCR *sp; EXF *ep; EXCMDARG *cmdp; { if (!F_ISSET(sp, S_SRE_SET)) { msgq(sp, M_ERR, "No previous regular expression"); return (1); } return (substitute(sp, ep, cmdp, cmdp->argv[0]->bp, &sp->sre, 0)); } /* * The nasty part of the substitution is what happens when the replacement * string contains newlines. It's a bit tricky -- consider the information * that has to be retained for "s/f\(o\)o/^M\1^M\1/". The solution here is * to build a set of newline offsets which we use to break the line up later, * when the replacement is done. Don't change it unless you're pretty damned * confident. */ #define NEEDNEWLINE(sp) { \ if (sp->newl_len == sp->newl_cnt) { \ sp->newl_len += 25; \ REALLOC(sp, sp->newl, size_t *, \ sp->newl_len * sizeof(size_t)); \ if (sp->newl == NULL) { \ sp->newl_len = 0; \ return (1); \ } \ } \ } #define BUILD(sp, l, len) { \ if (lbclen + (len) > lblen) { \ lblen += MAX(lbclen + (len), 256); \ REALLOC(sp, lb, char *, lblen); \ if (lb == NULL) { \ lbclen = 0; \ return (1); \ } \ } \ memmove(lb + lbclen, l, len); \ lbclen += len; \ } #define NEEDSP(sp, len, pnt) { \ if (lbclen + (len) > lblen) { \ lblen += MAX(lbclen + (len), 256); \ REALLOC(sp, lb, char *, lblen); \ if (lb == NULL) { \ lbclen = 0; \ return (1); \ } \ pnt = lb + lbclen; \ } \ } /* * substitute -- * Do the substitution. This stuff is *really* tricky. There are * lots of special cases, and general nastiness. Don't mess with it * unless you're pretty confident. */ static int substitute(sp, ep, cmdp, s, re, flags) SCR *sp; EXF *ep; EXCMDARG *cmdp; char *s; regex_t *re; u_int flags; { MARK from, to; recno_t elno, lno; regmatch_t match[10]; size_t blen, cnt, last, lbclen, lblen, len, llen, offset, saved_offset; int cflag, lflag, nflag, pflag, rflag; int didsub, do_eol_match, eflags, empty_ok, eval; int linechanged, matched, quit, rval; char *bp, *lb; /* * !!! * Historically, the 'g' and 'c' suffices were always toggled as flags, * so ":s/A/B/" was the same as ":s/A/B/ccgg". If O_EDCOMPATIBLE was * not set, they were initialized to 0 for all substitute commands. If * O_EDCOMPATIBLE was set, they were initialized to 0 only if the user * specified substitute/replacement patterns (see ex_substitute()). */ if (!O_ISSET(sp, O_EDCOMPATIBLE)) sp->c_suffix = sp->g_suffix = 0; /* * Historic vi permitted the '#', 'l' and 'p' options in vi mode, but * it only displayed the last change. I'd disallow them, but they are * useful in combination with the [v]global commands. In the current * model the problem is combining them with the 'c' flag -- the screen * would have to flip back and forth between the confirm screen and the * ex print screen, which would be pretty awful. We do display all * changes, though, for what that's worth. * * !!! * Historic vi was fairly strict about the order of "options", the * count, and "flags". I'm somewhat fuzzy on the difference between * options and flags, anyway, so this is a simpler approach, and we * just take it them in whatever order the user gives them. (The ex * usage statement doesn't reflect this.) */ cflag = lflag = nflag = pflag = rflag = 0; for (lno = OOBLNO; *s != '\0'; ++s) switch (*s) { case ' ': case '\t': continue; case '+': ++cmdp->flagoff; break; case '-': --cmdp->flagoff; break; case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': if (lno != OOBLNO) goto usage; errno = 0; lno = strtoul(s, &s, 10); if (*s == '\0') /* Loop increment correction. */ --s; if (errno == ERANGE) { if (lno == LONG_MAX) msgq(sp, M_ERR, "Count overflow"); else if (lno == LONG_MIN) msgq(sp, M_ERR, "Count underflow"); else msgq(sp, M_SYSERR, NULL); return (1); } /* * In historic vi, the count was inclusive from the * second address. */ cmdp->addr1.lno = cmdp->addr2.lno; cmdp->addr2.lno += lno - 1; break; case '#': nflag = 1; break; case 'c': sp->c_suffix = !sp->c_suffix; break; case 'g': sp->g_suffix = !sp->g_suffix; break; case 'l': lflag = 1; break; case 'p': pflag = 1; break; case 'r': if (LF_ISSET(SUB_FIRST)) { msgq(sp, M_ERR, "Regular expression specified; r flag meaningless"); return (1); } if (!F_ISSET(sp, S_SRE_SET)) { msgq(sp, M_ERR, "No previous regular expression"); return (1); } rflag = 1; re = &sp->sre; break; default: goto usage; } if (*s != '\0' || !rflag && LF_ISSET(SUB_MUSTSETR)) { usage: msgq(sp, M_ERR, "Usage: %s", cmdp->cmd->usage); return (1); } if (IN_VI_MODE(sp) && sp->c_suffix && (lflag || nflag || pflag)) { msgq(sp, M_ERR, "The #, l and p flags may not be combined with the c flag in vi mode"); return (1); } /* * bp: if interactive, line cache * blen: if interactive, line cache length * lb: build buffer pointer. * lbclen: current length of built buffer. * lblen; length of build buffer. */ bp = lb = NULL; blen = lbclen = lblen = 0; /* For each line... */ for (matched = quit = 0, lno = cmdp->addr1.lno, elno = cmdp->addr2.lno; !quit && lno <= elno; ++lno) { /* Someone's unhappy, time to stop. */ if (INTERRUPTED(sp)) { if (!F_ISSET(sp, S_GLOBAL)) msgq(sp, M_INFO, "Interrupted"); break; } /* Get the line. */ if ((s = file_gline(sp, ep, lno, &llen)) == NULL) { GETLINE_ERR(sp, lno); goto ret1; } /* * Make a local copy if doing confirmation -- when calling * the confirm routine we're likely to lose the cached copy. */ if (sp->c_suffix) { if (bp == NULL) { GET_SPACE_RET(sp, bp, blen, llen); } else ADD_SPACE_RET(sp, bp, blen, llen); memmove(bp, s, llen); s = bp; } /* Start searching from the beginning. */ offset = 0; len = llen; /* Reset the build buffer offset. */ lbclen = 0; /* Reset empty match flag. */ empty_ok = 1; /* * We don't want to have to do a setline if the line didn't * change -- keep track of whether or not this line changed. * If doing confirmations, don't want to keep setting the * line if change is refused -- keep track of substitutions. */ didsub = linechanged = 0; /* New line, do an EOL match. */ do_eol_match = 1; /* It's not nul terminated, but we pretend it is. */ eflags = REG_STARTEND; /* * The search area is from s + offset to the EOL. * * Generally, match[0].rm_so is the offset of the start * of the match from the start of the search, and offset * is the offset of the start of the last search. */ nextmatch: match[0].rm_so = 0; match[0].rm_eo = len; /* Get the next match. */ eval = regexec(re, (char *)s + offset, 10, match, eflags); /* * There wasn't a match or if there was an error, deal with * it. If there was a previous match in this line, resolve * the changes into the database. Otherwise, just move on. */ if (eval == REG_NOMATCH) goto endmatch; if (eval != 0) { re_error(sp, eval, re); goto ret1; } matched = 1; /* Only the first search can match an anchored expression. */ eflags |= REG_NOTBOL; /* * !!! * It's possible to match 0-length strings -- for example, the * command s;a*;X;, when matched against the string "aabb" will * result in "XbXbX", i.e. the matches are "aa", the space * between the b's and the space between the b's and the end of * the string. There is a similar space between the beginning * of the string and the a's. The rule that we use (because vi * historically used it) is that any 0-length match, occurring * immediately after a match, is ignored. Otherwise, the above * example would have resulted in "XXbXbX". Another example is * incorrectly using " *" to replace groups of spaces with one * space. * * The way we do this is that if we just had a successful match, * the starting offset does not skip characters, and the match * is empty, ignore the match and move forward. If there's no * more characters in the string, we were attempting to match * after the last character, so quit. */ if (!empty_ok && match[0].rm_so == 0 && match[0].rm_eo == 0) { empty_ok = 1; if (len == 0) goto endmatch; BUILD(sp, s + offset, 1) ++offset; --len; goto nextmatch; } /* Confirm change. */ if (sp->c_suffix) { /* * Set the cursor position for confirmation. Note, * if we matched on a '$', the cursor may be past * the end of line. * * XXX * We may want to "fix" this in the confirm routine, * if the confirm routine should be able to display * a cursor past EOL. */ from.lno = to.lno = lno; from.cno = match[0].rm_so + offset; to.cno = match[0].rm_eo; if (llen == 0) from.cno = to.cno = 0; else { if (to.cno >= llen) to.cno = llen - 1; if (from.cno >= llen) from.cno = llen - 1; } switch (sp->s_confirm(sp, ep, &from, &to)) { case CONF_YES: break; case CONF_NO: didsub = 0; BUILD(sp, s +offset, match[0].rm_eo); goto skip; case CONF_QUIT: /* Set the quit flag. */ quit = 1; /* If interruptible, pass the info back. */ if (F_ISSET(sp, S_INTERRUPTIBLE)) F_SET(sp, S_INTERRUPTED); /* * If any changes, resolve them, otherwise * return to the main loop. */ goto endmatch; } } /* Copy the bytes before the match into the build buffer. */ BUILD(sp, s + offset, match[0].rm_so); /* Substitute the matching bytes. */ didsub = 1; if (regsub(sp, s + offset, &lb, &lbclen, &lblen, match)) goto ret1; /* Set the change flag so we know this line was modified. */ linechanged = 1; /* Move past the matched bytes. */ skip: offset += match[0].rm_eo; len -= match[0].rm_eo; /* A match cannot be followed by an empty pattern. */ empty_ok = 0; /* * If doing a global change with confirmation, we have to * update the screen. The basic idea is to store the line * so the screen update routines can find it, and restart. */ if (didsub && sp->c_suffix && sp->g_suffix) { /* * The new search offset will be the end of the * modified line. */ saved_offset = lbclen; /* Copy the rest of the line. */ if (len) BUILD(sp, s + offset, len) /* Set the new offset. */ offset = saved_offset; /* Store inserted lines, adjusting the build buffer. */ last = 0; if (sp->newl_cnt) { for (cnt = 0; cnt < sp->newl_cnt; ++cnt, ++lno, ++elno) { if (file_iline(sp, ep, lno, lb + last, sp->newl[cnt] - last)) goto ret1; last = sp->newl[cnt] + 1; ++sp->rptlines[L_ADDED]; } lbclen -= last; offset -= last; sp->newl_cnt = 0; } /* Store and retrieve the line. */ if (file_sline(sp, ep, lno, lb + last, lbclen)) goto ret1; if ((s = file_gline(sp, ep, lno, &llen)) == NULL) { GETLINE_ERR(sp, lno); goto ret1; } ADD_SPACE_RET(sp, bp, blen, llen) memmove(bp, s, llen); s = bp; len = llen - offset; /* Restart the build. */ lbclen = 0; BUILD(sp, s, offset); /* * If we haven't already done the after-the-string * match, do one. Set REG_NOTEOL so the '$' pattern * only matches once. */ if (!do_eol_match) goto endmatch; if (offset == len) { do_eol_match = 0; eflags |= REG_NOTEOL; } goto nextmatch; } /* * If it's a global: * * If at the end of the string, do a test for the after * the string match. Set REG_NOTEOL so the '$' pattern * only matches once. */ if (sp->g_suffix && do_eol_match) { if (len == 0) { do_eol_match = 0; eflags |= REG_NOTEOL; } goto nextmatch; } endmatch: if (!linechanged) continue; /* Copy any remaining bytes into the build buffer. */ if (len) BUILD(sp, s + offset, len) /* Store inserted lines, adjusting the build buffer. */ last = 0; if (sp->newl_cnt) { for (cnt = 0; cnt < sp->newl_cnt; ++cnt, ++lno, ++elno) { if (file_iline(sp, ep, lno, lb + last, sp->newl[cnt] - last)) goto ret1; last = sp->newl[cnt] + 1; ++sp->rptlines[L_ADDED]; } lbclen -= last; sp->newl_cnt = 0; } /* Store the changed line. */ if (file_sline(sp, ep, lno, lb + last, lbclen)) goto ret1; /* Update changed line counter. */ if (sp->rptlchange != lno) { sp->rptlchange = lno; ++sp->rptlines[L_CHANGED]; } /* * !!! * Display as necessary. Historic practice is to only * display the last line of a line split into multiple * lines. */ if (lflag || nflag || pflag) { from.lno = to.lno = lno; from.cno = to.cno = 0; if (lflag) ex_print(sp, ep, &from, &to, E_F_LIST); if (nflag) ex_print(sp, ep, &from, &to, E_F_HASH); if (pflag) ex_print(sp, ep, &from, &to, E_F_PRINT); } if (!sp->c_suffix) sp->lno = lno; /* * !!! * Move the cursor to the last line changed. */ if (!sp->c_suffix) sp->lno = lno; } /* * !!! * Move the cursor to the first non-blank of the last line change. * * XXX * This is NOT backward compatible with historic vi, which always * moved to the last line actually changed. */ if (!sp->c_suffix) { sp->cno = 0; (void)nonblank(sp, ep, sp->lno, &sp->cno); } /* * If not in a global command, and nothing matched, say so. * Else, if none of the lines displayed, put something up. */ if (!matched) { if (!F_ISSET(sp, S_GLOBAL)) msgq(sp, M_INFO, "No match found"); } else if (!lflag && !nflag && !pflag) F_SET(EXP(sp), EX_AUTOPRINT); rval = 0; if (0) { ret1: rval = 1; } if (bp != NULL) FREE_SPACE(sp, bp, blen); if (lb != NULL) free(lb); return (rval); } /* * regsub -- * Do the substitution for a regular expression. */ static __inline int regsub(sp, ip, lbp, lbclenp, lblenp, match) SCR *sp; char *ip; /* Input line. */ char **lbp; size_t *lbclenp, *lblenp; regmatch_t match[10]; { enum { C_NOTSET, C_LOWER, C_ONELOWER, C_ONEUPPER, C_UPPER } conv; size_t lbclen, lblen; /* Local copies. */ size_t mlen; /* Match length. */ size_t rpl; /* Remaining replacement length. */ char *rp; /* Replacement pointer. */ int ch; int no; /* Match replacement offset. */ char *p, *t; /* Buffer pointers. */ char *lb; /* Local copies. */ lb = *lbp; /* Get local copies. */ lbclen = *lbclenp; lblen = *lblenp; /* * QUOTING NOTE: * * There are some special sequences that vi provides in the * replacement patterns. * & string the RE matched (\& if nomagic set) * \# n-th regular subexpression * \E end \U, \L conversion * \e end \U, \L conversion * \l convert the next character to lower-case * \L convert to lower-case, until \E, \e, or end of replacement * \u convert the next character to upper-case * \U convert to upper-case, until \E, \e, or end of replacement * * Otherwise, since this is the lowest level of replacement, discard * all escape characters. This (hopefully) follows historic practice. */ #define ADDCH(ch) { \ CHAR_T __ch = (ch); \ u_int __value = KEY_VAL(sp, __ch); \ if (__value == K_CR || __value == K_NL) { \ NEEDNEWLINE(sp); \ sp->newl[sp->newl_cnt++] = lbclen; \ } else if (conv != C_NOTSET) { \ switch (conv) { \ case C_ONELOWER: \ conv = C_NOTSET; \ /* FALLTHROUGH */ \ case C_LOWER: \ if (isupper(__ch)) \ __ch = tolower(__ch); \ break; \ case C_ONEUPPER: \ conv = C_NOTSET; \ /* FALLTHROUGH */ \ case C_UPPER: \ if (islower(__ch)) \ __ch = toupper(__ch); \ break; \ default: \ abort(); \ } \ } \ NEEDSP(sp, 1, p); \ *p++ = __ch; \ ++lbclen; \ } conv = C_NOTSET; for (rp = sp->repl, rpl = sp->repl_len, p = lb + lbclen; rpl--;) { switch (ch = *rp++) { case '&': if (O_ISSET(sp, O_MAGIC)) { no = 0; goto subzero; } break; case '\\': if (rpl == 0) break; --rpl; switch (ch = *rp) { case '&': ++rp; if (!O_ISSET(sp, O_MAGIC)) { no = 0; goto subzero; } break; case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': no = *rp++ - '0'; subzero: if (match[no].rm_so == -1 || match[no].rm_eo == -1) break; mlen = match[no].rm_eo - match[no].rm_so; for (t = ip + match[no].rm_so; mlen--; ++t) ADDCH(*t); continue; case 'e': case 'E': ++rp; conv = C_NOTSET; continue; case 'l': ++rp; conv = C_ONELOWER; continue; case 'L': ++rp; conv = C_LOWER; continue; case 'u': ++rp; conv = C_ONEUPPER; continue; case 'U': ++rp; conv = C_UPPER; continue; default: ++rp; break; } } ADDCH(ch); } *lbp = lb; /* Update caller's information. */ *lbclenp = lbclen; *lblenp = lblen; return (0); } @ 1.5 log @clean up import. still have to hack some things. @ text @d35 1 a35 1 static const char sccsid[] = "@@(#)ex_subst.c 8.58 (Berkeley) 8/17/94"; @ 1.4 log @nvi 1.11(beta) from bostic. reconcile conflicts/kill rcsids. @ text @d35 1 a35 1 static char sccsid[] = "@@(#)ex_subst.c 8.39 (Berkeley) 3/22/94"; d63 2 a64 3 static int checkmatchsize __P((SCR *, regex_t *)); static inline int regsub __P((SCR *, char *, char **, size_t *, size_t *)); d110 7 d155 7 a161 5 /* If the pattern string is empty, use the last one. */ if (*ptrn == NULL) { if (!F_ISSET(sp, S_SUBRE_SET)) { msgq(sp, M_ERR, "No previous regular expression."); d164 1 a164 1 re = &sp->subre; d224 6 d231 3 a233 1 if (*p == '\0') { d238 3 a240 1 } else { d273 9 a281 6 if (sp->repl != NULL) FREE(sp->repl, sp->repl_len); if ((sp->repl = malloc(len)) == NULL) { msgq(sp, M_SYSERR, NULL); FREE_SPACE(sp, bp, blen); return (1); a282 2 memmove(sp->repl, bp, len); sp->repl_len = len; a284 3 if (checkmatchsize(sp, &sp->subre)) return (1); d301 1 a301 1 msgq(sp, M_ERR, "No previous regular expression."); d320 1 a320 1 msgq(sp, M_ERR, "No previous regular expression."); d388 1 d390 1 a390 1 int cflag, gflag, lflag, nflag, pflag, rflag; d392 1 a392 1 int linechanged, matched, quit, rval, teardown; d396 11 d422 1 a422 1 cflag = gflag = lflag = nflag = pflag = rflag = 0; d427 6 d444 1 a444 1 msgq(sp, M_ERR, "Count overflow."); d446 1 a446 1 msgq(sp, M_ERR, "Count underflow."); d462 1 a462 1 cflag = 1; d465 1 a465 1 gflag = 1; d476 1 a476 1 "Regular expression specified; r flag meaningless."); d479 1 a479 1 if (!F_ISSET(sp, S_SUBRE_SET)) { d481 1 a481 1 "No previous regular expression."); d485 1 d496 1 a496 1 if (IN_VI_MODE(sp) && cflag && (lflag || nflag || pflag)) { d498 1 a498 1 "The #, l and p flags may not be combined with the c flag in vi mode."); a501 3 /* Set up interrupts. */ teardown = !intr_init(sp); d517 1 a517 1 if (F_ISSET(sp, S_INTERRUPTED)) { d519 1 a519 1 msgq(sp, M_INFO, "Interrupted."); d526 1 a526 1 return (1); d533 1 a533 1 if (cflag) { d569 3 a571 3 * Generally, sp->match[0].rm_so is the offset of the start * of the match from the start of the search, and offset is * the offset of the start of the last search. d573 2 a574 2 nextmatch: sp->match[0].rm_so = 0; sp->match[0].rm_eo = len; d577 1 a577 2 eval = regexec(re, (char *)s + offset, re->re_nsub + 1, sp->match, eflags); d615 1 a615 2 if (!empty_ok && sp->match[0].rm_so == 0 && sp->match[0].rm_eo == 0) { d626 1 a626 1 if (cflag) { d638 2 a639 2 from.cno = sp->match[0].rm_so + offset; to.cno = sp->match[0].rm_eo; d653 1 a653 1 BUILD(sp, s +offset, sp->match[0].rm_eo); d672 1 a672 15 BUILD(sp, s + offset, sp->match[0].rm_so); /* * Cursor moves to last line changed, unless doing confirm, * in which case don't move it. * * !!! * Historic vi just put the cursor on the first non-blank * of the last line changed. We move to the beginning of * the next substitution. */ if (!cflag) { sp->lno = lno; sp->cno = lbclen; } d676 1 a676 1 if (regsub(sp, s + offset, &lb, &lbclen, &lblen)) d683 2 a684 2 skip: offset += sp->match[0].rm_eo; len -= sp->match[0].rm_eo; d694 1 a694 1 if (didsub && cflag && gflag) { d761 1 a761 1 if (gflag && do_eol_match) { d796 4 a799 1 ++sp->rptlines[L_CHANGED]; d801 6 a806 1 /* Display as necessary. */ d817 23 d848 1 a848 1 msgq(sp, M_INFO, "No match found."); a856 3 if (teardown) intr_end(sp); d859 2 d868 2 a869 2 static inline int regsub(sp, ip, lbp, lbclenp, lblenp) d874 1 d909 1 a909 1 u_int __value = term_key_val(sp, __ch); \ d952 1 a953 1 ++rp; d961 5 a965 6 subzero: if (sp->match[no].rm_so == -1 || sp->match[no].rm_eo == -1) continue; mlen = sp->match[no].rm_eo - sp->match[no].rm_so; for (t = ip + sp->match[no].rm_so; mlen--; ++t) a999 18 return (0); } static int checkmatchsize(sp, re) SCR *sp; regex_t *re; { /* Build nsub array as necessary. */ if (sp->matchsize < re->re_nsub + 1) { sp->matchsize = re->re_nsub + 1; REALLOC(sp, sp->match, regmatch_t *, sp->matchsize * sizeof(regmatch_t)); if (sp->match == NULL) { sp->matchsize = 0; return (1); } } @ 1.3 log @Do the right thing if the pattern has an escaped escape character followed immediately by a delimiter. @ text @d2 1 a2 1 * Copyright (c) 1992, 1993 d35 1 a35 2 /* from: static char sccsid[] = "@@(#)ex_substitute.c 8.33 (Berkeley) 1/9/94"; */ static char *rcsid = "$Id: ex_subst.c,v 1.2 1994/01/24 06:40:41 cgd Exp $"; d39 2 d42 1 d45 3 d50 1 d53 4 a58 1 #include "interrupt.h" a65 1 static void subst_intr __P((int)); d141 1 a141 1 if (p[0] == '\\') { d144 1 a144 7 else if (p[1] == '\\') { /* * Skip over an escaped escape character; * otherwise the check for an escaped * delimiter will be confused on the next * iteration. */ a145 2 } } a201 3 * The special character ~ (\~ if O_MAGIC not set) inserts the * previous replacement string into this replacement string. * d206 6 d215 1 a215 2 * an escape character, or if O_MAGIC is set and it escapes a * tilde. a222 5 /* * Count ~'s to figure out how much space we need. We could * special case nonexistent last patterns or whether or not * O_MAGIC is set, but it's probably not worth the effort. */ d235 1 a235 1 if (p[1] == '\\' || p[1] == delim) d237 4 a240 1 else if (p[1] == '~') { d310 1 a310 1 /* a369 1 DECLARE_INTERRUPTS; d373 1 d375 1 a375 2 int linechanged, matched, quit, rval; int cflag, gflag, lflag, nflag, pflag, rflag; d467 2 a468 2 if (!F_ISSET(sp, S_GLOBAL)) SET_UP_INTERRUPTS(subst_intr); d632 1 a632 1 d650 2 a651 1 * of the last line changed. This might be better. d655 1 a655 1 sp->cno = sp->match[0].rm_so + offset; d810 2 a811 3 interrupt_err: if (!F_ISSET(sp, S_GLOBAL)) TEAR_DOWN_INTERRUPTS; d849 1 a849 1 * \# n-th regular subexpression a972 22 } /* * subst_intr -- * Set the interrupt bit in any screen that is interruptible. * * XXX * In the future this may be a problem. The user should be able to move to * another screen and keep typing while this runs. If so, and the user has * more than one substitute running, it will be hard to decide which one to * stop. */ static void subst_intr(signo) int signo; { SCR *sp; for (sp = __global_list->dq.cqh_first; sp != (void *)&__global_list->dq; sp = sp->q.cqe_next) if (F_ISSET(sp, S_INTERRUPTIBLE)) F_SET(sp, S_INTERRUPTED); @ 1.2 log @more Ids than you'll ever want. @ text @d36 1 a36 1 static char *rcsid = "$Id$"; d133 13 a145 2 if (p[0] == '\\' && p[1] == delim) ++p; @ 1.1 log @Initial revision @ text @d35 2 a36 1 static char sccsid[] = "@@(#)ex_substitute.c 8.33 (Berkeley) 1/9/94"; @ 1.1.1.1 log @nvi 1.03, from ftp.cs.berkeley.edu, per keith bostic's permission. @ text @@ 1.1.1.2 log @nvi/nex 1.11beta from bostic. @ text @d2 1 a2 1 * Copyright (c) 1992, 1993, 1994 d35 1 a35 1 static char sccsid[] = "@@(#)ex_subst.c 8.39 (Berkeley) 3/22/94"; a38 2 #include #include a39 1 #include a41 3 #include #include #include a43 1 #include a45 4 #include "compat.h" #include #include d48 1 d56 1 d132 2 a133 5 if (p[0] == '\\') if (p[1] == delim) ++p; else if (p[1] == '\\') *t++ = *p++; d190 3 a196 6 * The special character ~ (\~ if O_MAGIC not set) inserts the * previous replacement string into this replacement string. * Count ~'s to figure out how much space we need. We could * special case nonexistent last patterns or whether or not * O_MAGIC is set, but it's probably not worth the effort. * d200 2 a201 1 * if O_MAGIC is set and it escapes a tilde. d209 5 d226 1 a226 1 if (p[1] == delim) d228 1 a228 4 else if (p[1] == '\\') { *t++ = *p++; ++len; } else if (p[1] == '~') { d298 1 a298 1 /* d358 1 d362 2 a364 2 int didsub, do_eol_match, eflags, empty_ok, eval; int linechanged, matched, quit, rval, teardown; d456 2 a457 2 /* Set up interrupts. */ teardown = !intr_init(sp); d621 1 a621 1 d639 1 a639 2 * of the last line changed. We move to the beginning of * the next substitution. d643 1 a643 1 sp->cno = lbclen; d798 3 a800 2 if (teardown) intr_end(sp); d838 1 a838 1 * \# n-th regular subexpression d962 22 @ 1.1.1.3 log @new public version of nvi @ text @d35 1 a35 1 static const char sccsid[] = "@@(#)ex_subst.c 8.58 (Berkeley) 8/17/94"; d63 3 a65 2 static __inline int regsub __P((SCR *, char *, char **, size_t *, size_t *, regmatch_t [10])); a110 7 * !!! * The full-blown substitute command reset the remembered * state of the 'c' and 'g' suffices. */ sp->c_suffix = sp->g_suffix = 0; /* d149 5 a153 7 /* * If the pattern string is empty, use the last RE (not just the * last substitution RE). */ if (*ptrn == '\0') { if (!F_ISSET(sp, S_SRE_SET)) { msgq(sp, M_ERR, "No previous regular expression"); d156 1 a156 1 re = &sp->sre; a215 6 * * !!! * If the entire replacement pattern is "%", then use the last * replacement pattern. This semantic was added to vi in System * V and then percolated elsewhere, presumably around the time * that it was added to their version of ed(1). d217 1 a217 3 if (p[0] == '\0' || p[0] == delim) { if (p[0] == delim) ++p; d222 1 a222 3 } else if (p[0] == '%' && (p[1] == '\0' || p[1] == delim)) p += p[1] == delim ? 2 : 1; else { d255 6 a260 9 if ((sp->repl_len = len) != 0) { if (sp->repl != NULL) free(sp->repl); if ((sp->repl = malloc(len)) == NULL) { msgq(sp, M_SYSERR, NULL); FREE_SPACE(sp, bp, blen); return (1); } memmove(sp->repl, bp, len); d262 2 d266 3 d285 1 a285 1 msgq(sp, M_ERR, "No previous regular expression"); d304 1 a304 1 msgq(sp, M_ERR, "No previous regular expression"); a371 1 regmatch_t match[10]; d373 1 a373 1 int cflag, lflag, nflag, pflag, rflag; d375 1 a375 1 int linechanged, matched, quit, rval; a378 11 * !!! * Historically, the 'g' and 'c' suffices were always toggled as flags, * so ":s/A/B/" was the same as ":s/A/B/ccgg". If O_EDCOMPATIBLE was * not set, they were initialized to 0 for all substitute commands. If * O_EDCOMPATIBLE was set, they were initialized to 0 only if the user * specified substitute/replacement patterns (see ex_substitute()). */ if (!O_ISSET(sp, O_EDCOMPATIBLE)) sp->c_suffix = sp->g_suffix = 0; /* d394 1 a394 1 cflag = lflag = nflag = pflag = rflag = 0; a398 6 continue; case '+': ++cmdp->flagoff; break; case '-': --cmdp->flagoff; d410 1 a410 1 msgq(sp, M_ERR, "Count overflow"); d412 1 a412 1 msgq(sp, M_ERR, "Count underflow"); d428 1 a428 1 sp->c_suffix = !sp->c_suffix; d431 1 a431 1 sp->g_suffix = !sp->g_suffix; d442 1 a442 1 "Regular expression specified; r flag meaningless"); d445 1 a445 1 if (!F_ISSET(sp, S_SRE_SET)) { d447 1 a447 1 "No previous regular expression"); a450 1 re = &sp->sre; d461 1 a461 1 if (IN_VI_MODE(sp) && sp->c_suffix && (lflag || nflag || pflag)) { d463 1 a463 1 "The #, l and p flags may not be combined with the c flag in vi mode"); d467 3 d485 1 a485 1 if (INTERRUPTED(sp)) { d487 1 a487 1 msgq(sp, M_INFO, "Interrupted"); d494 1 a494 1 goto ret1; d501 1 a501 1 if (sp->c_suffix) { d537 3 a539 3 * Generally, match[0].rm_so is the offset of the start * of the match from the start of the search, and offset * is the offset of the start of the last search. d541 2 a542 2 nextmatch: match[0].rm_so = 0; match[0].rm_eo = len; d545 2 a546 1 eval = regexec(re, (char *)s + offset, 10, match, eflags); d584 2 a585 1 if (!empty_ok && match[0].rm_so == 0 && match[0].rm_eo == 0) { d596 1 a596 1 if (sp->c_suffix) { d608 2 a609 2 from.cno = match[0].rm_so + offset; to.cno = match[0].rm_eo; d623 1 a623 1 BUILD(sp, s +offset, match[0].rm_eo); d642 15 a656 1 BUILD(sp, s + offset, match[0].rm_so); d660 1 a660 1 if (regsub(sp, s + offset, &lb, &lbclen, &lblen, match)) d667 2 a668 2 skip: offset += match[0].rm_eo; len -= match[0].rm_eo; d678 1 a678 1 if (didsub && sp->c_suffix && sp->g_suffix) { d745 1 a745 1 if (sp->g_suffix && do_eol_match) { d780 1 a780 4 if (sp->rptlchange != lno) { sp->rptlchange = lno; ++sp->rptlines[L_CHANGED]; } d782 1 a782 6 /* * !!! * Display as necessary. Historic practice is to only * display the last line of a line split into multiple * lines. */ a792 23 if (!sp->c_suffix) sp->lno = lno; /* * !!! * Move the cursor to the last line changed. */ if (!sp->c_suffix) sp->lno = lno; } /* * !!! * Move the cursor to the first non-blank of the last line change. * * XXX * This is NOT backward compatible with historic vi, which always * moved to the last line actually changed. */ if (!sp->c_suffix) { sp->cno = 0; (void)nonblank(sp, ep, sp->lno, &sp->cno); d801 1 a801 1 msgq(sp, M_INFO, "No match found"); d810 3 a814 2 if (lb != NULL) free(lb); d822 2 a823 2 static __inline int regsub(sp, ip, lbp, lbclenp, lblenp, match) a827 1 regmatch_t match[10]; d862 1 a862 1 u_int __value = KEY_VAL(sp, __ch); \ a904 1 ++rp; d906 1 d914 6 a919 5 subzero: if (match[no].rm_so == -1 || match[no].rm_eo == -1) break; mlen = match[no].rm_eo - match[no].rm_so; for (t = ip + match[no].rm_so; mlen--; ++t) d954 18 @ 1.1.1.4 log @new public version of nvi @ text @d35 1 a35 1 static char sccsid[] = "@@(#)ex_subst.c 8.59 (Berkeley) 8/17/94"; @ 1.1.1.5 log @import of nvi 1.66 @ text @a3 2 * Copyright (c) 1992, 1993, 1994, 1995, 1996 * Keith Bostic. All rights reserved. d5 27 a31 1 * See the LICENSE file for redistribution information. a33 2 #include "config.h" d35 1 a35 1 static const char sccsid[] = "@@(#)ex_subst.c 10.30 (Berkeley) 5/16/96"; d46 1 d50 1 d53 6 a58 2 #include "../common/common.h" #include "../vi/vi.h" d63 4 a66 6 static int re_conv __P((SCR *, char **, int *)); static int re_cscope_conv __P((SCR *, char **, int *)); static int re_sub __P((SCR *, char *, char **, size_t *, size_t *, regmatch_t [10])); static int re_tag_conv __P((SCR *, char **, int *)); static int s __P((SCR *, EXCMD *, char *, regex_t *, u_int)); d69 1 a69 1 * ex_s -- a72 2 * * PUBLIC: int ex_s __P((SCR *, EXCMD *)); d75 1 a75 1 ex_s(sp, cmdp) d77 2 a78 1 EXCMD *cmdp; d80 1 a80 1 regex_t *re; d83 1 a83 1 int delim; a96 2 if (cmdp->argc == 0) goto subagain; d103 1 a103 2 subagain: return (ex_subagain(sp, cmdp)); d105 3 a107 2 if (isalnum(delim) || delim == '\\') return (s(sp, cmdp, p, &sp->subre_c, SUB_MUSTSETR)); d115 1 a115 1 d160 2 a161 2 if (sp->re == NULL) { ex_emsg(sp, NULL, EXM_NOPREVRE); d164 9 d174 2 a175 3 /* Compile the RE if necessary. */ if (!F_ISSET(sp, SC_RE_SEARCH) && re_compile(sp, sp->re, NULL, NULL, &sp->re_c, RE_C_SEARCH)) d177 13 a189 2 flags = 0; } else { d191 2 d194 2 a195 4 * Compile the RE. Historic practice is that substitutes set * the search direction as well as both substitute and search * RE's. We compile the RE twice, as we don't want to bother * ref counting the pattern string and (opaque) structure. d197 7 a203 7 if (re_compile(sp, ptrn, &sp->re, &sp->re_len, &sp->re_c, RE_C_SEARCH)) return (1); if (re_compile(sp, ptrn, &sp->subre, &sp->subre_len, &sp->subre_c, RE_C_SUBST)) return (1); a204 1 sp->searchdir = FORWARD; a205 1 re = &sp->re_c; d212 1 a212 1 * re_sub(). d235 1 a235 1 free(sp->repl); d285 1 a285 1 return (s(sp, cmdp, p, re, flags)); a292 2 * * PUBLIC: int ex_subagain __P((SCR *, EXCMD *)); d295 1 a295 1 ex_subagain(sp, cmdp) d297 2 a298 1 EXCMD *cmdp; d300 2 a301 2 if (sp->subre == NULL) { ex_emsg(sp, NULL, EXM_NOPREVRE); d304 1 a304 5 if (!F_ISSET(sp, SC_RE_SUBST) && re_compile(sp, sp->subre, NULL, NULL, &sp->subre_c, RE_C_SUBST)) return (1); return (s(sp, cmdp, cmdp->argc ? cmdp->argv[0]->bp : NULL, &sp->subre_c, 0)); a311 2 * * PUBLIC: int ex_subtilde __P((SCR *, EXCMD *)); d314 1 a314 1 ex_subtilde(sp, cmdp) d316 2 a317 1 EXCMD *cmdp; d319 2 a320 2 if (sp->re == NULL) { ex_emsg(sp, NULL, EXM_NOPREVRE); d323 1 a323 5 if (!F_ISSET(sp, SC_RE_SEARCH) && re_compile(sp, sp->re, NULL, NULL, &sp->re_c, RE_C_SEARCH)) return (1); return (s(sp, cmdp, cmdp->argc ? cmdp->argv[0]->bp : NULL, &sp->re_c, 0)); a326 5 * s -- * Do the substitution. This stuff is *really* tricky. There are lots of * special cases, and general nastiness. Don't mess with it unless you're * pretty confident. * d331 1 a331 1 * when the replacement is done. Don't change it unless you're *damned* d371 6 d378 1 a378 1 s(sp, cmdp, s, re, flags) d380 2 a381 1 EXCMD *cmdp; a385 1 EVENT ev; d387 1 a387 2 TEXTH tiq; recno_t elno, lno, slno; d389 1 a389 2 size_t blen, cnt, last, lbclen, lblen, len, llen; size_t offset, saved_offset, scno; a394 5 NEEDFILE(sp, cmdp); slno = sp->lno; scno = sp->cno; d401 1 a401 1 * specified substitute/replacement patterns (see ex_s()). a422 2 if (s == NULL) goto noargs; d444 1 a444 1 msgq(sp, M_ERR, "153|Count overflow"); d446 1 a446 1 msgq(sp, M_ERR, "154|Count underflow"); a456 3 if (!db_exist(sp, cmdp->addr2.lno) && db_last(sp, &cmdp->addr2.lno)) return (1); a462 6 /* Ex text structure initialization. */ if (F_ISSET(sp, SC_EX)) { memset(&tiq, 0, sizeof(TEXTH)); CIRCLEQ_INIT(&tiq); } d476 1 a476 1 "155|Regular expression specified; r flag meaningless"); d479 3 a481 2 if (!F_ISSET(sp, SC_RE_SEARCH)) { ex_emsg(sp, NULL, EXM_NOPREVRE); d485 1 a485 1 re = &sp->re_c; d492 1 a492 1 usage: ex_emsg(sp, cmdp->cmd->usage, EXM_USAGE); d496 1 a496 1 noargs: if (F_ISSET(sp, SC_VI) && sp->c_suffix && (lflag || nflag || pflag)) { d498 1 a498 1 "156|The #, l and p flags may not be combined with the c flag in vi mode"); d517 3 a519 1 if (INTERRUPTED(sp)) d521 1 d524 4 a527 2 if (db_get(sp, lno, DBG_FATAL, &s, &llen)) goto err; d588 1 a588 1 goto err; d631 5 d639 1 a639 5 to.cno = match[0].rm_eo + offset; /* * Both ex and vi have to correct for a change before * the first character in the line. */ d642 1 a642 9 if (F_ISSET(sp, SC_VI)) { /* * Only vi has to correct for a change after * the last character in the line. * * XXX * It would be nice to change the vi code so * that we could display a cursor past EOL. */ a646 29 sp->lno = from.lno; sp->cno = from.cno; if (vs_refresh(sp, 1)) goto err; vs_update(sp, msg_cat(sp, "169|Confirm change? [n]", NULL), NULL); if (v_event_get(sp, &ev, 0, 0)) goto err; switch (ev.e_event) { case E_CHARACTER: break; case E_EOF: case E_ERR: case E_INTERRUPT: goto lquit; default: v_event_err(sp, &ev); goto lquit; } } else { if (ex_print(sp, cmdp, &from, &to, 0) || ex_scprint(sp, &from, &to)) goto lquit; if (ex_txt(sp, &tiq, 0, TXT_CR)) goto err; ev.e_c = tiq.cqh_first->lb[0]; d648 2 a649 3 switch (ev.e_c) { case CH_YES: d651 1 a651 2 default: case CH_NO: d655 7 a661 4 case CH_QUIT: /* Set the quit/interrupted flags. */ lquit: quit = 1; F_SET(sp->gp, G_INTERRUPTED); d664 2 a665 2 * Resolve any changes, then return to (and * exit from) the main loop. a670 7 /* * Set the cursor to the last position changed, converting * from 1-based to 0-based. */ sp->lno = lno; sp->cno = match[0].rm_so; d676 2 a677 2 if (re_sub(sp, s + offset, &lb, &lbclen, &lblen, match)) goto err; d713 1 a713 1 if (db_insert(sp, lno, d715 1 a715 1 goto err; d725 6 a730 4 if (db_set(sp, lno, lb + last, lbclen)) goto err; if (db_get(sp, lno, DBG_FATAL, &s, &llen)) goto err; d781 1 a781 1 if (db_insert(sp, d783 1 a783 1 goto err; d792 2 a793 2 if (db_set(sp, lno, lb + last, lbclen)) goto err; d811 1 a811 1 (void)ex_print(sp, cmdp, &from, &to, E_C_LIST); d813 1 a813 1 (void)ex_print(sp, cmdp, &from, &to, E_C_HASH); d815 1 a815 1 (void)ex_print(sp, cmdp, &from, &to, E_C_PRINT); d817 10 d831 1 a831 7 * Historically, vi attempted to leave the cursor at the same place if * the substitution was done at the current cursor position. Otherwise * it moved it to the first non-blank of the last line changed. There * were some problems: for example, :s/$/foo/ with the cursor on the * last character of the line left the cursor on the last character, or * the & command with multiple occurrences of the matching string in the * line usually left the cursor in a fairly random position. d833 3 a835 5 * We try to do the same thing, with the exception that if the user is * doing substitution with confirmation, we move to the last line about * which the user was consulted, as opposed to the last line that they * actually changed. This prevents a screen flash if the user doesn't * change many of the possible lines. d837 1 a837 1 if (!sp->c_suffix && (sp->lno != slno || sp->cno != scno)) { d839 1 a839 1 (void)nonblank(sp, sp->lno, &sp->cno); a845 1 rval = 0; d847 2 a848 4 if (!F_ISSET(sp, SC_EX_GLOBAL)) { msgq(sp, M_ERR, "157|No match found"); goto err; } d850 1 a850 1 F_SET(cmdp, E_AUTOPRINT); d852 1 d854 1 a854 1 err: rval = 1; d865 1 a865 404 * re_compile -- * Compile the RE. * * PUBLIC: int re_compile __P((SCR *, * PUBLIC: char *, char **, size_t *, regex_t *, u_int)); */ int re_compile(sp, ptrn, ptrnp, lenp, rep, flags) SCR *sp; char *ptrn, **ptrnp; size_t *lenp; regex_t *rep; u_int flags; { size_t len; int reflags, replaced, rval; char *p; /* Set RE flags. */ reflags = 0; if (!LF_ISSET(RE_C_CSCOPE | RE_C_TAG)) { if (O_ISSET(sp, O_EXTENDED)) reflags |= REG_EXTENDED; if (O_ISSET(sp, O_IGNORECASE)) reflags |= REG_ICASE; if (O_ISSET(sp, O_ICLOWER)) { for (p = ptrn; *p != '\0'; ++p) if (isupper(*p)) break; if (*p == '\0') reflags |= REG_ICASE; } } /* If we're replacing a saved value, clear the old one. */ if (LF_ISSET(RE_C_SEARCH) && F_ISSET(sp, SC_RE_SEARCH)) { regfree(&sp->re_c); F_CLR(sp, SC_RE_SEARCH); } if (LF_ISSET(RE_C_SUBST) && F_ISSET(sp, SC_RE_SUBST)) { regfree(&sp->subre_c); F_CLR(sp, SC_RE_SUBST); } /* * If we're saving the string, it's a pattern we haven't seen before, * so convert the vi-style RE's to POSIX 1003.2 RE's. Save a copy for * later recompilation. Free any previously saved value. */ replaced = 0; if (ptrnp != NULL) { if (*ptrnp != NULL) { free(*ptrnp); *ptrnp = NULL; } if (LF_ISSET(RE_C_CSCOPE)) { if (re_cscope_conv(sp, &ptrn, &replaced)) return (1); /* * XXX * Currently, the match-any- expression used in * re_cscope_conv() requires extended RE's. This may * not be right or safe. */ reflags |= REG_EXTENDED; } else if (LF_ISSET(RE_C_TAG)) { if (re_tag_conv(sp, &ptrn, &replaced)) return (1); } else if (re_conv(sp, &ptrn, &replaced)) return (1); len = strlen(ptrn); if (lenp != NULL) *lenp = len; if ((*ptrnp = v_strdup(sp, ptrn, len)) == NULL) return (1); } rval = regcomp(rep, ptrn, reflags); /* Free up any allocated memory. */ if (replaced) FREE_SPACE(sp, ptrn, 0); if (rval) { if (!LF_ISSET(RE_C_SILENT)) re_error(sp, rval, rep); return (1); } if (LF_ISSET(RE_C_SEARCH)) F_SET(sp, SC_RE_SEARCH); if (LF_ISSET(RE_C_SUBST)) F_SET(sp, SC_RE_SUBST); return (0); } /* * re_conv -- * Convert vi's regular expressions into something that the * the POSIX 1003.2 RE functions can handle. * * There are three conversions we make to make vi's RE's (specifically * the global, search, and substitute patterns) work with POSIX RE's. * * 1: If O_MAGIC is not set, strip backslashes from the magic character * set (.[*~) that have them, and add them to the ones that don't. * 2: If O_MAGIC is not set, the string "\~" is replaced with the text * from the last substitute command's replacement string. If O_MAGIC * is set, it's the string "~". * 3: The pattern \ does "word" searches, convert it to use the * new RE escapes. * * !!!/XXX * This doesn't exactly match the historic behavior of vi because we do * the ~ substitution before calling the RE engine, so magic characters * in the replacement string will be expanded by the RE engine, and they * weren't historically. It's a bug. */ static int re_conv(sp, ptrnp, replacedp) SCR *sp; char **ptrnp; int *replacedp; { size_t blen, needlen; int magic; char *bp, *p, *t; /* * First pass through, we figure out how much space we'll need. * We do it in two passes, on the grounds that most of the time * the user is doing a search and won't have magic characters. * That way we can skip the malloc and memmove's. */ for (p = *ptrnp, magic = 0, needlen = 0; *p != '\0'; ++p) switch (*p) { case '\\': switch (*++p) { case '<': magic = 1; needlen += sizeof(RE_WSTART); break; case '>': magic = 1; needlen += sizeof(RE_WSTOP); break; case '~': if (!O_ISSET(sp, O_MAGIC)) { magic = 1; needlen += sp->repl_len; } break; case '.': case '[': case '*': if (!O_ISSET(sp, O_MAGIC)) { magic = 1; needlen += 1; } break; default: needlen += 2; } break; case '~': if (O_ISSET(sp, O_MAGIC)) { magic = 1; needlen += sp->repl_len; } break; case '.': case '[': case '*': if (!O_ISSET(sp, O_MAGIC)) { magic = 1; needlen += 2; } break; default: needlen += 1; break; } if (!magic) { *replacedp = 0; return (0); } /* * Get enough memory to hold the final pattern. * * XXX * It's nul-terminated, for now. */ GET_SPACE_RET(sp, bp, blen, needlen + 1); for (p = *ptrnp, t = bp; *p != '\0'; ++p) switch (*p) { case '\\': switch (*++p) { case '<': memmove(t, RE_WSTART, sizeof(RE_WSTART) - 1); t += sizeof(RE_WSTART) - 1; break; case '>': memmove(t, RE_WSTOP, sizeof(RE_WSTOP) - 1); t += sizeof(RE_WSTOP) - 1; break; case '~': if (O_ISSET(sp, O_MAGIC)) *t++ = '~'; else { memmove(t, sp->repl, sp->repl_len); t += sp->repl_len; } break; case '.': case '[': case '*': if (O_ISSET(sp, O_MAGIC)) *t++ = '\\'; *t++ = *p; break; default: *t++ = '\\'; *t++ = *p; } break; case '~': if (O_ISSET(sp, O_MAGIC)) { memmove(t, sp->repl, sp->repl_len); t += sp->repl_len; } else *t++ = '~'; break; case '.': case '[': case '*': if (!O_ISSET(sp, O_MAGIC)) *t++ = '\\'; *t++ = *p; break; default: *t++ = *p; break; } *t = '\0'; *ptrnp = bp; *replacedp = 1; return (0); } /* * re_tag_conv -- * Convert a tags search path into something that the POSIX * 1003.2 RE functions can handle. */ static int re_tag_conv(sp, ptrnp, replacedp) SCR *sp; char **ptrnp; int *replacedp; { size_t blen, len; int lastdollar; char *bp, *p, *t; *replacedp = 0; len = strlen(p = *ptrnp); /* Max memory usage is 2 times the length of the string. */ GET_SPACE_RET(sp, bp, blen, len * 2); t = bp; /* The last character is a '/' or '?', we just strip it. */ if (p[len - 1] == '/' || p[len - 1] == '?') p[len - 1] = '\0'; /* The next-to-last character is a '$', and it's magic. */ if (p[len - 2] == '$') { lastdollar = 1; p[len - 2] = '\0'; } else lastdollar = 0; /* The first character is a '/' or '?', we just strip it. */ if (p[0] == '/' || p[0] == '?') ++p; /* The second character is a '^', and it's magic. */ if (p[0] == '^') *t++ = *p++; /* * Escape every other magic character we can find, stripping the * backslashes ctags inserts to escape the search delimiter * characters. */ while (p[0]) { /* Ctags escapes the search delimiter characters. */ if (p[0] == '\\' && (p[1] == '/' || p[1] == '?')) ++p; else if (strchr("^.[]$*", p[0])) *t++ = '\\'; *t++ = *p++; } if (lastdollar) *t++ = '$'; *t++ = '\0'; *ptrnp = bp; *replacedp = 1; return (0); } /* * re_cscope_conv -- * Convert a cscope search path into something that the POSIX * 1003.2 RE functions can handle. */ static int re_cscope_conv(sp, ptrnp, replacedp) SCR *sp; char **ptrnp; int *replacedp; { size_t blen, len, nspaces; char *bp, *p, *re; /* * Each space in the source line printed by cscope represents an * arbitrary sequence of spaces, tabs, and comments. */ #define CSCOPE_RE_SPACE "([ \t]|/\\*([^*]|\\*/)*\\*/)*" for (nspaces = 0, p = *ptrnp; *p != '\0'; ++p) if (*p == ' ') ++nspaces; /* * Allocate plenty of space: * the string, plus potential escaping characters * nspaces + 2 copies of CSCOPE_RE_SPACE * ^, $, nul terminator characters */ len = (p - *ptrnp) * 2 + (nspaces + 2) * sizeof(CSCOPE_RE_SPACE) + 3; GET_SPACE_RET(sp, bp, blen, len); p = bp; *p++ = '^'; memcpy(p, CSCOPE_RE_SPACE, sizeof(CSCOPE_RE_SPACE) - 1); p += sizeof(CSCOPE_RE_SPACE) - 1; for (re = *ptrnp; *re != '\0'; ++re) if (*re == ' ') { memcpy(p, CSCOPE_RE_SPACE, sizeof(CSCOPE_RE_SPACE) - 1); p += sizeof(CSCOPE_RE_SPACE) - 1; } else { if (strchr("\\^.[]$*", *re)) *p++ = '\\'; *p++ = *re; } memcpy(p, CSCOPE_RE_SPACE, sizeof(CSCOPE_RE_SPACE) - 1); p += sizeof(CSCOPE_RE_SPACE) - 1; *p++ = '$'; *p = '\0'; *ptrnp = bp; *replacedp = 1; return (0); } /* * re_error -- * Report a regular expression error. * * PUBLIC: void re_error __P((SCR *, int, regex_t *)); */ void re_error(sp, errcode, preg) SCR *sp; int errcode; regex_t *preg; { size_t s; char *oe; s = regerror(errcode, preg, "", 0); if ((oe = malloc(s)) == NULL) msgq(sp, M_SYSERR, NULL); else { (void)regerror(errcode, preg, oe, s); msgq(sp, M_ERR, "RE error: %s", oe); free(oe); } } /* * re_sub -- d868 2 a869 2 static int re_sub(sp, ip, lbp, lbclenp, lblenp, match) d907 1 a907 1 #define OUTCH(ch) { \ d966 1 a966 1 OUTCH(*t); d994 1 a994 1 OUTCH(ch); @ 1.1.1.6 log @import of nvi 1.79 @ text @d13 1 a13 1 static const char sccsid[] = "@@(#)ex_subst.c 10.37 (Berkeley) 9/15/96"; d35 2 a36 2 static int re_conv __P((SCR *, char **, size_t *, int *)); static int re_cscope_conv __P((SCR *, char **, size_t *, int *)); d39 1 a39 1 static int re_tag_conv __P((SCR *, char **, size_t *, int *)); d94 1 a94 1 * Get the pattern string, toss escaping characters. d106 1 a106 1 * Only toss an escaping character if it escapes a delimiter. d109 1 a109 1 * escaping a single escaping character is removed, but that's d142 3 a144 3 /* Re-compile the RE if necessary. */ if (!F_ISSET(sp, SC_RE_SEARCH) && re_compile(sp, sp->re, sp->re_len, NULL, NULL, &sp->re_c, RE_C_SEARCH)) d155 2 a156 2 if (re_compile(sp, ptrn, t - ptrn, &sp->re, &sp->re_len, &sp->re_c, RE_C_SEARCH)) d158 2 a159 2 if (re_compile(sp, ptrn, t - ptrn, &sp->subre, &sp->subre_len, &sp->subre_c, RE_C_SUBST)) d182 1 a182 1 * Only toss an escaping character if it escapes a delimiter or d225 1 a225 1 memcpy(t, sp->repl, sp->repl_len); d241 1 a241 1 memcpy(sp->repl, bp, len); d265 2 a266 2 if (!F_ISSET(sp, SC_RE_SUBST) && re_compile(sp, sp->subre, sp->subre_len, NULL, NULL, &sp->subre_c, RE_C_SUBST)) d289 2 a290 2 if (!F_ISSET(sp, SC_RE_SEARCH) && re_compile(sp, sp->re, sp->re_len, NULL, NULL, &sp->re_c, RE_C_SEARCH)) d330 1 a330 1 memcpy(lb + lbclen, l, len); \ d519 1 a519 1 memcpy(bp, s, llen); d753 1 a753 1 memcpy(bp, s, llen); d890 1 a890 1 * PUBLIC: char *, size_t, char **, size_t *, regex_t *, u_int)); d893 1 a893 1 re_compile(sp, ptrn, plen, ptrnp, lenp, rep, flags) d896 1 a896 1 size_t plen, *lenp; d912 1 a912 1 for (p = ptrn, len = plen; len > 0; ++p, --len) d915 1 a915 1 if (len == 0) d935 1 d937 4 d942 1 a942 1 if (re_cscope_conv(sp, &ptrn, &plen, &replaced)) d952 1 a952 1 if (re_tag_conv(sp, &ptrn, &plen, &replaced)) d955 1 a955 1 if (re_conv(sp, &ptrn, &plen, &replaced)) d957 1 a957 6 /* Discard previous pattern. */ if (*ptrnp != NULL) { free(*ptrnp); *ptrnp = NULL; } d959 4 a962 18 *lenp = plen; /* * Copy the string into allocated memory. * * XXX * Regcomp isn't 8-bit clean, so the pattern is nul-terminated * for now. There's just no other solution. */ MALLOC(sp, *ptrnp, char *, plen + 1); if (*ptrnp != NULL) { memcpy(*ptrnp, ptrn, plen); (*ptrnp)[plen] = '\0'; } /* Free up conversion-routine-allocated memory. */ if (replaced) FREE_SPACE(sp, ptrn, 0); d964 1 a964 2 if (*ptrnp == NULL) return (1); d966 3 a968 2 ptrn = *ptrnp; } d970 1 a970 6 /* * XXX * Regcomp isn't 8-bit clean, so we just lost if the pattern * contained a nul. Bummer! */ if ((rval = regcomp(rep, ptrn, /* plen, */ reflags)) != 0) { d1007 1 a1007 1 re_conv(sp, ptrnp, plenp, replacedp) a1009 1 size_t *plenp; d1012 1 a1012 1 size_t blen, len, needlen; d1020 1 a1020 1 * That way we can skip most of the memory allocation and copies. d1022 1 a1022 2 magic = 0; for (p = *ptrnp, len = *plenp, needlen = 0; len > 0; ++p, --len) d1025 11 a1035 4 if (len > 1) { --len; switch (*++p) { case '<': d1037 7 a1043 3 needlen += sizeof(RE_WSTART); break; case '>': d1045 1 a1045 18 needlen += sizeof(RE_WSTOP); break; case '~': if (!O_ISSET(sp, O_MAGIC)) { magic = 1; needlen += sp->repl_len; } break; case '.': case '[': case '*': if (!O_ISSET(sp, O_MAGIC)) { magic = 1; needlen += 1; } break; default: needlen += 2; d1047 4 a1050 2 } else needlen += 1; d1076 7 a1082 3 /* Get enough memory to hold the final pattern. */ *replacedp = 1; GET_SPACE_RET(sp, bp, blen, needlen); d1084 1 a1084 1 for (p = *ptrnp, len = *plenp, t = bp; len > 0; ++p, --len) d1087 21 a1107 30 if (len > 1) { --len; switch (*++p) { case '<': memcpy(t, RE_WSTART, sizeof(RE_WSTART) - 1); t += sizeof(RE_WSTART) - 1; break; case '>': memcpy(t, RE_WSTOP, sizeof(RE_WSTOP) - 1); t += sizeof(RE_WSTOP) - 1; break; case '~': if (O_ISSET(sp, O_MAGIC)) *t++ = '~'; else { memcpy(t, sp->repl, sp->repl_len); t += sp->repl_len; } break; case '.': case '[': case '*': if (O_ISSET(sp, O_MAGIC)) *t++ = '\\'; *t++ = *p; break; default: d1109 3 a1111 3 *t++ = *p; } } else d1113 2 d1118 1 a1118 1 memcpy(t, sp->repl, sp->repl_len); d1134 1 d1137 1 a1137 1 *plenp = t - bp; d1147 1 a1147 1 re_tag_conv(sp, ptrnp, plenp, replacedp) a1149 1 size_t *plenp; d1156 3 a1158 1 len = *plenp; a1160 1 *replacedp = 1; a1162 1 p = *ptrnp; d1165 6 a1170 7 /* If the last character is a '/' or '?', we just strip it. */ if (len > 0 && (p[len - 1] == '/' || p[len - 1] == '?')) --len; /* If the next-to-last or last character is a '$', it's magic. */ if (len > 0 && p[len - 1] == '$') { --len; d1172 1 d1176 2 a1177 2 /* If the first character is a '/' or '?', we just strip it. */ if (len > 0 && (p[0] == '/' || p[0] == '?')) { a1178 2 --len; } d1180 2 a1181 2 /* If the first or second character is a '^', it's magic. */ if (p[0] == '^') { a1182 2 --len; } d1185 2 a1186 2 * Escape every other magic character we can find, meanwhile stripping * the backslashes ctags inserts when escaping the search delimiter d1189 3 a1191 2 for (; len > 0; --len) { if (p[0] == '\\' && (p[1] == '/' || p[1] == '?')) { d1193 1 a1193 2 --len; } else if (strchr("^.[]$*", p[0])) d1199 1 d1202 1 a1202 1 *plenp = t - bp; d1212 1 a1212 1 re_cscope_conv(sp, ptrnp, plenp, replacedp) a1214 1 size_t *plenp; d1218 1 a1218 1 char *bp, *p, *t; d1225 1 a1225 1 for (nspaces = 0, p = *ptrnp, len = *plenp; len > 0; ++p, --len) d1231 3 a1233 3 * the string, plus potential escaping characters; * nspaces + 2 copies of CSCOPE_RE_SPACE; * ^, $, nul terminator characters. a1234 1 *replacedp = 1; d1238 9 a1246 11 p = *ptrnp; t = bp; *t++ = '^'; memcpy(t, CSCOPE_RE_SPACE, sizeof(CSCOPE_RE_SPACE) - 1); t += sizeof(CSCOPE_RE_SPACE) - 1; for (len = *plenp; len > 0; ++p, --len) if (*p == ' ') { memcpy(t, CSCOPE_RE_SPACE, sizeof(CSCOPE_RE_SPACE) - 1); t += sizeof(CSCOPE_RE_SPACE) - 1; d1248 3 a1250 3 if (strchr("\\^.[]$*+?()|{}", *p)) *t++ = '\\'; *t++ = *p; d1253 4 a1256 3 memcpy(t, CSCOPE_RE_SPACE, sizeof(CSCOPE_RE_SPACE) - 1); t += sizeof(CSCOPE_RE_SPACE) - 1; *t++ = '$'; d1259 1 a1259 1 *plenp = t - bp; d1329 1 a1329 1 * all escaping characters. This (hopefully) matches historic practice. d1331 1 a1331 1 #define OUTCH(ch, nltrans) { \ d1334 1 a1334 1 if (nltrans && (__value == K_CR || __value == K_NL)) { \ d1390 1 a1390 1 OUTCH(*t, 0); d1418 1 a1418 1 OUTCH(ch, 1); @