sv_cmp and friends

   Message-Id: <20020304023103.A14140@math.ohio-state.edu>
p4raw-link: @14577 on //depot/perl: 0ad5258ff3f3328f321188cbb4fcd6a74b365431

p4raw-id: //depot/perl@14985

3 files changed, 56 insertions, 59 deletions

diff --git a/embed.fnc b/embed.fnc
index a94654f123..2853071e33 100644
--- a/embed.fnc
+++ b/embed.fnc
@@ -426,6 +426,7 @@ Ap	|void	|markstack_grow
 #if defined(USE_LOCALE_COLLATE)
 p	|char*	|mem_collxfrm	|const char* s|STRLEN len|STRLEN* xlen
 #endif
+Apd	|int	|memcmp_byte_utf8	|char *sbyte|STRLEN lbyte|char *sutf|STRLEN lutf
 Afp	|SV*	|mess		|const char* pat|...
 Ap	|SV*	|vmess		|const char* pat|va_list* args
 p	|void	|qerror		|SV* err
diff --git a/sv.c b/sv.c
index 27150d6706..f893fa6bd8 100644
--- a/sv.c
+++ b/sv.c
@@ -5349,7 +5349,6 @@ Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
     char *pv2;
     STRLEN cur2;
     I32  eq     = 0;
-    char *tpv   = Nullch;
 
     if (!sv1) {
 	pv1 = "";
@@ -5365,35 +5364,13 @@ Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
     else
 	pv2 = SvPV(sv2, cur2);
 
-    /* do not utf8ize the comparands as a side-effect */
-    if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
-	bool is_utf8 = TRUE;
-        /* UTF-8ness differs */
-
-	if (SvUTF8(sv1)) {
-	    /* sv1 is the UTF-8 one , If is equal it must be downgrade-able */
-	    char *pv = (char*)bytes_from_utf8((U8*)pv1, &cur1, &is_utf8);
-	    if (pv != pv1)
-		pv1 = tpv = pv;
-	}
-	else {
-	    /* sv2 is the UTF-8 one , If is equal it must be downgrade-able */
-	    char *pv = (char *)bytes_from_utf8((U8*)pv2, &cur2, &is_utf8);
-	    if (pv != pv2)
-		pv2 = tpv = pv;
-	}
-	if (is_utf8) {
-	    /* Downgrade not possible - cannot be eq */
-	    return FALSE;
-	}
-    }
-
-    if (cur1 == cur2)
-	eq = memEQ(pv1, pv2, cur1);
+    if (SvUTF8(sv1) == SvUTF8(sv2) || IN_BYTES)
+	eq = (cur1 == cur2) && memEQ(pv1, pv2, cur1);
+    else if (SvUTF8(sv1)) /* do not utf8ize the comparands as a side-effect */
+	eq = !memcmp_byte_utf8(pv2, cur2, pv1, cur1);
+    else
+	eq = !memcmp_byte_utf8(pv1, cur1, pv2, cur2);
 	
-    if (tpv != Nullch)
-	Safefree(tpv);
-
     return eq;
 }
 
@@ -5413,9 +5390,7 @@ Perl_sv_cmp(pTHX_ register SV *sv1, register SV *sv2)
 {
     STRLEN cur1, cur2;
     char *pv1, *pv2;
-    I32  cmp;
-    bool pv1tmp = FALSE;
-    bool pv2tmp = FALSE;
+    I32  retval;
 
     if (!sv1) {
 	pv1 = "";
@@ -5431,40 +5406,28 @@ Perl_sv_cmp(pTHX_ register SV *sv1, register SV *sv2)
     else
 	pv2 = SvPV(sv2, cur2);
 
-    /* do not utf8ize the comparands as a side-effect */
-    if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
-	if (SvUTF8(sv1)) {
-	    pv2 = (char*)bytes_to_utf8((U8*)pv2, &cur2);
-	    pv2tmp = TRUE;
-	}
-	else {
-	    pv1 = (char*)bytes_to_utf8((U8*)pv1, &cur1);
-	    pv1tmp = TRUE;
-	}
-    }
-
     if (!cur1) {
-	cmp = cur2 ? -1 : 0;
+	return cur2 ? -1 : 0;
     } else if (!cur2) {
-	cmp = 1;
-    } else {
-	I32 retval = memcmp((void*)pv1, (void*)pv2, cur1 < cur2 ? cur1 : cur2);
+	return 1;
+    } else if (SvUTF8(sv1) == SvUTF8(sv2) || IN_BYTES) {
+	retval = memcmp((void*)pv1, (void*)pv2, cur1 < cur2 ? cur1 : cur2);
 
 	if (retval) {
-	    cmp = retval < 0 ? -1 : 1;
+	    return retval < 0 ? -1 : 1;
 	} else if (cur1 == cur2) {
-	    cmp = 0;
-        } else {
-	    cmp = cur1 < cur2 ? -1 : 1;
+	    return 0;
+	} else {
+	    return cur1 < cur2 ? -1 : 1;
 	}
-    }
-
-    if (pv1tmp)
-	Safefree(pv1);
-    if (pv2tmp)
-	Safefree(pv2);
+    } else if (SvUTF8(sv1)) /* do not utf8ize the comparands as a side-effect */
+	retval = -memcmp_byte_utf8(pv2, cur2, pv1, cur1);
+    else
+	retval = memcmp_byte_utf8(pv1, cur1, pv2, cur2);
 
-    return cmp;
+    if (retval)				/* CURs taken into account already */
+	return retval < 0 ? -1 : 1;
+    return 0;
 }
 
 /*
diff --git a/util.c b/util.c
index a9f9ade8f6..4dc86764a7 100644
--- a/util.c
+++ b/util.c
@@ -4346,5 +4346,38 @@ Perl_sv_nounlocking(pTHX_ SV *sv)
 {
 }
 
+/*
+=for apidoc memcmp_byte_utf8
+
+Similar to memcmp(), but the first string is with bytes, the second
+with utf8.  Takes into account that the lengths may be different.
 
+=cut
+*/
 
+int
+Perl_memcmp_byte_utf8(pTHX_ char *sb, STRLEN lbyte, char *su, STRLEN lutf)
+{
+    U8 *sbyte = (U8*)sb;
+    U8 *sutf  = (U8*)su;
+    U8 *ebyte = sbyte + lbyte;
+    U8 *eutf  = sutf  + lutf;
+
+    while (sbyte < ebyte) {
+	if (sutf >= eutf)
+	    return 1;			/* utf one shorter */
+	if (*sbyte < 128) {
+	    if (*sbyte != *sutf)
+		return *sbyte - *sutf;
+	    sbyte++; sutf++;	/* CONTINUE */
+	} else if ((*sutf & 0x3F) == (*sbyte >> 6)) { /* byte 0xFF: 0xC3 BF */
+	    if ((sutf[1] & 0x3F) != (*sbyte & 0x3F))
+		return (*sbyte & 0x3F) - (*sutf & 0x3F);
+	    sbyte++, sutf += 2;	/* CONTINUE */
+	} else
+	    return (*sbyte >> 6) - (*sutf & 0x3F);
+    }
+    if (sutf >= eutf)
+	return 0;
+    return -1;				/* byte one shorter */
+}