Make the 'y' (translate) command aware of multibyte characters.

5 files changed, 145 insertions, 17 deletions

diff --git a/usr.bin/sed/compile.c b/usr.bin/sed/compile.c
index f0a6c9b..afeaa01 100644
--- a/usr.bin/sed/compile.c
+++ b/usr.bin/sed/compile.c
@@ -47,12 +47,14 @@ static const char sccsid[] = "@(#)compile.c	8.1 (Berkeley) 6/6/93";
 
 #include <ctype.h>
 #include <err.h>
+#include <errno.h>
 #include <fcntl.h>
 #include <limits.h>
 #include <regex.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+#include <wchar.h>
 
 #include "defs.h"
 #include "extern.h"
@@ -73,7 +75,7 @@ static char	 *compile_flags(char *, struct s_subst *);
 static char	 *compile_re(char *, regex_t **);
 static char	 *compile_subst(char *, struct s_subst *);
 static char	 *compile_text(void);
-static char	 *compile_tr(char *, char **);
+static char	 *compile_tr(char *, struct s_tr **);
 static struct s_command
 		**compile_stream(struct s_command **);
 static char	 *duptoeol(char *, const char *);
@@ -337,7 +339,7 @@ nonsel:		/* Now parse the command */
 			break;
 		case TR:			/* y */
 			p++;
-			p = compile_tr(p, (char **)&cmd->u.y);
+			p = compile_tr(p, &cmd->u.y);
 			EATSPACE();
 			if (*p == ';') {
 				p++;
@@ -619,12 +621,20 @@ compile_flags(char *p, struct s_subst *s)
  * Compile a translation set of strings into a lookup table.
  */
 static char *
-compile_tr(char *p, char **transtab)
+compile_tr(char *p, struct s_tr **py)
 {
+	struct s_tr *y;
 	int i;
-	char *lt, *op, *np;
+	const char *op, *np;
 	char old[_POSIX2_LINE_MAX + 1];
 	char new[_POSIX2_LINE_MAX + 1];
+	size_t oclen, oldlen, nclen, newlen;
+	mbstate_t mbs1, mbs2;
+
+	if ((*py = y = malloc(sizeof(*y))) == NULL)
+		err(1, NULL);
+	y->multis = NULL;
+	y->nmultis = 0;
 
 	if (*p == '\0' || *p == '\\')
 		errx(1,
@@ -639,17 +649,63 @@ compile_tr(char *p, char **transtab)
 		errx(1, "%lu: %s: unterminated transform target string",
 				linenum, fname);
 	EATSPACE();
-	if (strlen(new) != strlen(old))
+	op = old;
+	oldlen = mbsrtowcs(NULL, &op, 0, NULL);
+	if (oldlen == (size_t)-1)
+		err(1, NULL);
+	np = new;
+	newlen = mbsrtowcs(NULL, &np, 0, NULL);
+	if (newlen == (size_t)-1)
+		err(1, NULL);
+	if (newlen != oldlen)
 		errx(1, "%lu: %s: transform strings are not the same length",
 				linenum, fname);
-	/* We assume characters are 8 bits */
-	if ((lt = malloc(UCHAR_MAX)) == NULL)
-		err(1, "malloc");
-	for (i = 0; i <= UCHAR_MAX; i++)
-		lt[i] = (char)i;
-	for (op = old, np = new; *op; op++, np++)
-		lt[(u_char)*op] = *np;
-	*transtab = lt;
+	if (MB_CUR_MAX == 1) {
+		/*
+		 * The single-byte encoding case is easy: generate a
+		 * lookup table.
+		 */
+		for (i = 0; i <= UCHAR_MAX; i++)
+			y->bytetab[i] = (char)i;
+		for (; *op; op++, np++)
+			y->bytetab[(u_char)*op] = *np;
+	} else {
+		/*
+		 * Multi-byte encoding case: generate a lookup table as
+		 * above, but only for single-byte characters. The first
+		 * bytes of multi-byte characters have their lookup table
+		 * entries set to 0, which causes do_tr() to search through
+		 * an auxiliary vector of multi-byte mappings.
+		 */
+		memset(&mbs1, 0, sizeof(mbs1));
+		memset(&mbs2, 0, sizeof(mbs2));
+		for (i = 0; i <= UCHAR_MAX; i++)
+			y->bytetab[i] = (btowc(i) != WEOF) ? i : 0;
+		while (*op != '\0') {
+			oclen = mbrlen(op, MB_LEN_MAX, &mbs1);
+			if (oclen == (size_t)-1 || oclen == (size_t)-2)
+				errc(1, EILSEQ, NULL);
+			nclen = mbrlen(np, MB_LEN_MAX, &mbs2);
+			if (nclen == (size_t)-1 || nclen == (size_t)-2)
+				errc(1, EILSEQ, NULL);
+			if (oclen == 1 && nclen == 1)
+				y->bytetab[(u_char)*op] = *np;
+			else {
+				y->bytetab[(u_char)*op] = 0;
+				y->multis = realloc(y->multis,
+				    (y->nmultis + 1) * sizeof(*y->multis));
+				if (y->multis == NULL)
+					err(1, NULL);
+				i = y->nmultis++;
+				y->multis[i].fromlen = oclen;
+				memcpy(y->multis[i].from, op, oclen);
+				y->multis[i].tolen = nclen;
+				memcpy(y->multis[i].to, np, nclen);
+			}
+			op += oclen;
+			np += nclen;
+		}
+	}
 	return (p);
 }
 
diff --git a/usr.bin/sed/defs.h b/usr.bin/sed/defs.h
index 0e77014..bc487d9 100644
--- a/usr.bin/sed/defs.h
+++ b/usr.bin/sed/defs.h
@@ -35,6 +35,7 @@
  * SUCH DAMAGE.
  *
  *	@(#)defs.h	8.1 (Berkeley) 6/6/93
+ * $FreeBSD$
  */
 
 /*
@@ -71,6 +72,19 @@ struct s_subst {
 	char *new;				/* Replacement text */
 };
 
+/*
+ * Translate command.
+ */
+struct s_tr {
+	unsigned char bytetab[256];
+	struct trmulti {
+		int fromlen;
+		char from[MB_LEN_MAX];
+		int tolen;
+		char to[MB_LEN_MAX];
+	} *multis;
+	int nmultis;
+};
 
 /*
  * An internally compiled command.
@@ -84,7 +98,7 @@ struct s_command {
 	union {
 		struct s_command *c;		/* Command(s) for b t { */
 		struct s_subst *s;		/* Substitute command */
-		u_char *y;			/* Replace command array */
+		struct s_tr *y;			/* Replace command array */
 		int fd;				/* File descriptor for w */
 	} u;
 	char code;				/* Command code */
diff --git a/usr.bin/sed/main.c b/usr.bin/sed/main.c
index bb5ffec..670298e 100644
--- a/usr.bin/sed/main.c
+++ b/usr.bin/sed/main.c
@@ -56,6 +56,7 @@ static const char sccsid[] = "@(#)main.c	8.2 (Berkeley) 1/3/94";
 #include <err.h>
 #include <errno.h>
 #include <fcntl.h>
+#include <limits.h>
 #include <locale.h>
 #include <regex.h>
 #include <stddef.h>
diff --git a/usr.bin/sed/misc.c b/usr.bin/sed/misc.c
index 5b47f3d..91fff53 100644
--- a/usr.bin/sed/misc.c
+++ b/usr.bin/sed/misc.c
@@ -45,6 +45,7 @@ static const char sccsid[] = "@(#)misc.c	8.1 (Berkeley) 6/6/93";
 #include <sys/types.h>
 
 #include <err.h>
+#include <limits.h>
 #include <regex.h>
 #include <stdio.h>
 #include <stdlib.h>
diff --git a/usr.bin/sed/process.c b/usr.bin/sed/process.c
index 0d63994..1858b65 100644
--- a/usr.bin/sed/process.c
+++ b/usr.bin/sed/process.c
@@ -63,7 +63,7 @@ static const char sccsid[] = "@(#)process.c	8.6 (Berkeley) 4/20/94";
 #include "defs.h"
 #include "extern.h"
 
-static SPACE HS, PS, SS;
+static SPACE HS, PS, SS, YS;
 #define	pd		PS.deleted
 #define	ps		PS.space
 #define	psl		PS.len
@@ -71,6 +71,7 @@ static SPACE HS, PS, SS;
 #define	hsl		HS.len
 
 static __inline int	 applies(struct s_command *);
+static void		 do_tr(struct s_tr *);
 static void		 flush_appends(void);
 static void		 lputs(char *, size_t);
 static __inline int	 regexec_e(regex_t *, const char *, int, int, size_t);
@@ -97,6 +98,7 @@ process(void)
 	SPACE tspace;
 	size_t len, oldpsl = 0;
 	char *p;
+	char nc;
 
 	p = NULL;
 
@@ -247,8 +249,7 @@ redirect:
 			case 'y':
 				if (pd || psl == 0)
 					break;
-				for (p = ps, len = psl; len--; ++p)
-					*p = cp->u.y[(unsigned char)*p];
+				do_tr(cp->u.y);
 				break;
 			case ':':
 			case '}':
@@ -426,6 +427,61 @@ substitute(struct s_command *cp)
 }
 
 /*
+ * do_tr --
+ *	Perform translation ('y' command) in the pattern space.
+ */
+static void
+do_tr(struct s_tr *y)
+{
+	SPACE tmp;
+	char c, *p;
+	size_t clen, left;
+	int i;
+
+	if (MB_CUR_MAX == 1) {
+		/*
+		 * Single-byte encoding: perform in-place translation
+		 * of the pattern space.
+		 */
+		for (p = ps; p < &ps[psl]; p++)
+			*p = y->bytetab[(u_char)*p];
+	} else {
+		/*
+		 * Multi-byte encoding: perform translation into the
+		 * translation space, then swap the translation and
+		 * pattern spaces.
+		 */
+		/* Clean translation space. */
+		YS.len = 0;
+		for (p = ps, left = psl; left > 0; p += clen, left -= clen) {
+			if ((c = y->bytetab[(u_char)*p]) != '\0') {
+				cspace(&YS, &c, 1, APPEND);
+				clen = 1;
+				continue;
+			}
+			for (i = 0; i < y->nmultis; i++)
+				if (left >= y->multis[i].fromlen &&
+				    memcmp(p, y->multis[i].from,
+				    y->multis[i].fromlen) == 0)
+					break;
+			if (i < y->nmultis) {
+				cspace(&YS, y->multis[i].to,
+				    y->multis[i].tolen, APPEND);
+				clen = y->multis[i].fromlen;
+			} else {
+				cspace(&YS, p, 1, APPEND);
+				clen = 1;
+			}
+		}
+		/* Swap the translation space and the pattern space. */
+		tmp = PS;
+		PS = YS;
+		YS = tmp;
+		YS.space = YS.back;
+	}
+}
+
+/*
  * Flush append requests.  Always called before reading a line,
  * therefore it also resets the substitution done (sdone) flag.
  */