[PATCH 1/2] Optimize indexing array parameters.

[PATCH 1/2] Optimize indexing array parameters.

[Daniel Shahaf]

% () { for 1 in $prefix/zsh/bin/zsh Src/zsh; do $1 -f -c 'a=( {1..1000000} ); repeat 3 time ( repeat 300 : $a[1]  )'; done }
( repeat 300; do; : $a[1]; done; )  1.68s user 0.01s system 98% cpu 1.718 total
( repeat 300; do; : $a[1]; done; )  1.69s user 0.01s system 99% cpu 1.710 total
( repeat 300; do; : $a[1]; done; )  1.69s user 0.01s system 99% cpu 1.714 total

( repeat 300; do; : $a[1]; done; )  0.00s user 0.01s system 72% cpu 0.022 total
( repeat 300; do; : $a[1]; done; )  0.00s user 0.01s system 72% cpu 0.022 total
( repeat 300; do; : $a[1]; done; )  0.01s user 0.01s system 69% cpu 0.023 total
---
 Src/params.c |  2 +-
 Src/subst.c  | 15 +++++++++++----
 Src/utils.c  | 40 ++++++++++++++++++++++++++++++++++++++++
 3 files changed, 52 insertions(+), 5 deletions(-)

diff --git a/Src/params.c b/Src/params.c
index e7a7365..33f177e 100644
--- a/Src/params.c
+++ b/Src/params.c
@@ -2026,7 +2026,7 @@ getstrvalue(Value v)
 	else {
 	    if (v->start < 0)
 		v->start += arrlen(ss);
-	    s = (v->start >= arrlen(ss) || v->start < 0) ?
+	    s = (arrlen_le(ss, v->start) || v->start < 0) ?
 		(char *) hcalloc(1) : ss[v->start];
 	}
 	return s;
diff --git a/Src/subst.c b/Src/subst.c
index 8e704b1..e3af156 100644
--- a/Src/subst.c
+++ b/Src/subst.c
@@ -2548,12 +2548,19 @@ paramsubst(LinkList l, LinkNode n, char **str, int qt, int pf_flags,
 		 * necessary joining of arrays until this point
 		 * to avoid the multsub() horror.
 		 */
-		int tmplen = arrlen(v->pm->gsu.a->getfn(v->pm));
 
-		if (v->start < 0)
+		/* arrlen() is expensive, so only compute it if needed. */
+		int tmplen = -1;
+
+		if (v->start < 0) {
+		    tmplen = arrlen(v->pm->gsu.a->getfn(v->pm));
 		    v->start += tmplen + ((v->flags & VALFLAG_INV) ? 1 : 0);
-		if (!(v->flags & VALFLAG_INV) &&
-		    (v->start >= tmplen || v->start < 0))
+		}
+		if (!(v->flags & VALFLAG_INV))
+		    if (v->start < 0 ||
+			(tmplen != -1
+			 ? v->start >= tmplen
+			 : arrlen_le(v->pm->gsu.a->getfn(v->pm), v->start)))
 		    vunset = 1;
 	    }
 	    if (!vunset) {
diff --git a/Src/utils.c b/Src/utils.c
index 95be1fb..95da960 100644
--- a/Src/utils.c
+++ b/Src/utils.c
@@ -2280,6 +2280,46 @@ arrlen(char **s)
     return count;
 }
 
+/* Return TRUE iff arrlen(s) >= lower_bound, but more efficiently. */
+
+/**/
+mod_export char
+arrlen_ge(char **s, unsigned lower_bound)
+{
+    while (lower_bound--)
+	if (!*s++)
+	    return 0 /* FALSE */;
+
+    return 1 /* TRUE */;
+}
+
+/* Return TRUE iff arrlen(s) > lower_bound, but more efficiently. */
+
+/**/
+mod_export char
+arrlen_gt(char **s, unsigned lower_bound)
+{
+    return arrlen_ge(s, 1+lower_bound);
+}
+
+/* Return TRUE iff arrlen(s) <= upper_bound, but more efficiently. */
+
+/**/
+mod_export char
+arrlen_le(char **s, unsigned upper_bound)
+{
+    return arrlen_lt(s, 1+upper_bound);
+}
+
+/* Return TRUE iff arrlen(s) < upper_bound, but more efficiently. */
+
+/**/
+mod_export char
+arrlen_lt(char **s, unsigned upper_bound)
+{
+    return !arrlen_ge(s, upper_bound);
+}
+
 /* Skip over a balanced pair of parenthesis. */
 
 /**/

[PATCH 2/2] Start using the new arrlen_ge() / arrlen_le() helpers.

[Daniel Shahaf]

---
 Src/Modules/terminfo.c | 2 +-
 Src/Modules/zutil.c    | 4 ++--
 Src/builtin.c          | 6 +++---
 Src/params.c           | 4 ++--
 Src/prompt.c           | 6 +++---
 Src/utils.c            | 2 +-
 6 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/Src/Modules/terminfo.c b/Src/Modules/terminfo.c
index e0439af..bbd3258 100644
--- a/Src/Modules/terminfo.c
+++ b/Src/Modules/terminfo.c
@@ -99,7 +99,7 @@ bin_echoti(char *name, char **argv, UNUSED(Options ops), UNUSED(int func))
 	return 1;
     }
     /* check that the number of arguments provided is not too high */
-    if (arrlen(argv) > 9) {
+    if (arrlen_gt(argv, 9)) {
         zwarnnam(name, "too many arguments");
         return 1;
     }
diff --git a/Src/Modules/zutil.c b/Src/Modules/zutil.c
index 477575c..d95c0c2 100644
--- a/Src/Modules/zutil.c
+++ b/Src/Modules/zutil.c
@@ -472,7 +472,7 @@ bin_zstyle(char *nam, char **args, UNUSED(Options ops), UNUSED(int func))
 	Patprog prog;
 	char *pat;
 
-	if (arrlen(args) < 2) {
+	if (arrlen_lt(args, 2)) {
 	    zwarnnam(nam, "not enough arguments");
 	    return 1;
 	}
@@ -491,7 +491,7 @@ bin_zstyle(char *nam, char **args, UNUSED(Options ops), UNUSED(int func))
 	Style s;
 	char *context, *stylename;
 
-	switch (arrlen(args)) {
+	switch (arrlen_ge(args, 3) ? 3 : arrlen(args)) {
 	case 2:
 	    context = args[0];
 	    stylename = args[1];
diff --git a/Src/builtin.c b/Src/builtin.c
index bfb9e69..fb14b2e 100644
--- a/Src/builtin.c
+++ b/Src/builtin.c
@@ -3406,7 +3406,7 @@ bin_unset(char *name, char **argv, Options ops, int func)
 		    } else {
 			/* start is after the element for reverse index */
 			int start = vbuf.start - !!(vbuf.flags & VALFLAG_INV);
-			if (start < arrlen(vbuf.pm->u.arr)) {
+			if (arrlen_gt(vbuf.pm->u.arr, start)) {
 			    char *arr[2];
 			    arr[0] = "";
 			    arr[1] = 0;
@@ -5026,7 +5026,7 @@ bin_shift(char *name, char **argv, Options ops, UNUSED(int func))
     if (*argv) {
         for (; *argv; argv++)
             if ((s = getaparam(*argv))) {
-                if (num > arrlen(s)) {
+                if (arrlen_lt(s, num)) {
 		    zwarnnam(name, "shift count must be <= $#");
 		    ret++;
 		    continue;
@@ -5095,7 +5095,7 @@ bin_getopts(UNUSED(char *name), char **argv, UNUSED(Options ops), UNUSED(int fun
 	zoptind = 1;
 	optcind = 0;
     }
-    if(zoptind > arrlen(args))
+    if (arrlen_lt(args, zoptind))
 	/* no more options */
 	return 1;
 
diff --git a/Src/params.c b/Src/params.c
index 33f177e..0eda784 100644
--- a/Src/params.c
+++ b/Src/params.c
@@ -2225,13 +2225,13 @@ getarrvalue(Value v)
 	v->start += arrlen(s);
     if (v->end < 0)
 	v->end += arrlen(s) + 1;
-    if (v->start > arrlen(s) || v->start < 0)
+    if (arrlen_lt(s, v->start) || v->start < 0)
 	s = arrdup(nular);
     else
 	s = arrdup(s + v->start);
     if (v->end <= v->start)
 	s[0] = NULL;
-    else if (v->end - v->start <= arrlen(s))
+    else if (arrlen_ge(s, v->end - v->start))
 	s[v->end - v->start] = NULL;
     return s;
 }
diff --git a/Src/prompt.c b/Src/prompt.c
index bb27453..d4f3898 100644
--- a/Src/prompt.c
+++ b/Src/prompt.c
@@ -395,11 +395,11 @@ putpromptchar(int doprint, int endchar, unsigned int *txtchangep)
 			test = 1;
 		    break;
 		case 'v':
-		    if (arrlen(psvar) >= arg)
+		    if (arrlen_ge(psvar, arg))
 			test = 1;
 		    break;
 		case 'V':
-		    if (arrlen(psvar) >= arg) {
+		    if (arrlen_ge(psvar, arg)) {
 			if (*psvar[(arg ? arg : 1) - 1])
 			    test = 1;
 		    }
@@ -736,7 +736,7 @@ putpromptchar(int doprint, int endchar, unsigned int *txtchangep)
 		    arg = 1;
 		else if (arg < 0)
 		    arg += arrlen(psvar) + 1;
-		if (arg > 0 && arrlen(psvar) >= arg)
+		if (arg > 0 && arrlen_ge(psvar, arg))
 		    stradd(psvar[arg - 1]);
 		break;
 	    case 'E':
diff --git a/Src/utils.c b/Src/utils.c
index 95da960..0a5954f 100644
--- a/Src/utils.c
+++ b/Src/utils.c
@@ -1157,7 +1157,7 @@ finddir(char *s)
     scanhashtable(nameddirtab, 0, 0, 0, finddir_scan, 0);
 
     ares = subst_string_by_hook("zsh_directory_name", "d", finddir_full);
-    if (ares && arrlen(ares) >= 2 &&
+    if (ares && arrlen_ge(ares, 2) &&
 	(len = (int)zstrtol(ares[1], NULL, 10)) > finddir_best) {
 	/* better duplicate this string since it's come from REPLY */
 	finddir_last = (Nameddir)hcalloc(sizeof(struct nameddir));