Add a compatibility interface for ohash.

This include's espie@'s wholesale src/lib/libc/ohash directory from OpenBSD
into compat_ohash.c (with a single copyright/license notice at the top)
and src/include/ohash.h as compat_ohash.h.
The ohash_int.h part of compat_ohash.c has been changed only in that ohash.h
points to compat_ohash.h.
Added HAVE_OHASH test (test-ohash.c) to Makefile.
In mandocdb.c and mansearch.c, check HAVE_OHASH test for inclusion.

6 files changed, 449 insertions, 5 deletions

diff --git a/Makefile b/Makefile
index 96832d5c..10221e4a 100644
--- a/Makefile
+++ b/Makefile
@@ -147,6 +147,7 @@ SRCS		 = Makefile \
 		   test-fgetln.c \
 		   test-getsubopt.c \
 		   test-mmap.c \
+		   test-ohash.c \
 		   test-strlcat.c \
 		   test-strlcpy.c \
 		   test-strptime.c \
@@ -187,6 +188,7 @@ LIBMANDOC_OBJS	 = $(LIBMAN_OBJS) \
 
 COMPAT_OBJS	 = compat_fgetln.o \
 		   compat_getsubopt.o \
+		   compat_ohash.o \
 		   compat_strlcat.o \
 		   compat_strlcpy.o
 
@@ -203,7 +205,7 @@ $(LIBMAN_OBJS): libman.h
 $(LIBMDOC_OBJS): libmdoc.h
 $(LIBROFF_OBJS): libroff.h
 $(LIBMANDOC_OBJS): mandoc.h mdoc.h man.h libmandoc.h config.h
-$(COMPAT_OBJS): config.h
+$(COMPAT_OBJS): config.h compat_ohash.h
 
 MANDOC_HTML_OBJS = eqn_html.o \
 		   html.o \
@@ -396,6 +398,10 @@ config.h: config.h.pre config.h.post
 	rm -f config.log
 	( cat config.h.pre; \
 	  echo; \
+	  if $(CC) $(CFLAGS) -Werror -o test-ohash test-ohash.c >> config.log 2>&1; then \
+		echo '#define HAVE_OHASH'; \
+		rm test-ohash; \
+	  fi; \
 	  if $(CC) $(CFLAGS) -Werror -o test-fgetln test-fgetln.c >> config.log 2>&1; then \
 		echo '#define HAVE_FGETLN'; \
 		rm test-fgetln; \
diff --git a/compat_ohash.c b/compat_ohash.c
new file mode 100644
index 00000000..802f458d
--- /dev/null
+++ b/compat_ohash.c
@@ -0,0 +1,337 @@
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#ifdef HAVE_OHASH
+
+int dummy;
+
+#else
+
+/*	$OpenBSD: ohash_int.h,v 1.3 2006/01/16 15:52:25 espie Exp $	*/
+
+#include <stddef.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include "compat_ohash.h"
+
+struct _ohash_record {
+	u_int32_t	hv;
+	const char 	*p;
+};
+
+#define DELETED		((const char *)h)
+#define NONE		(h->size)
+
+/* Don't bother changing the hash table if the change is small enough.  */
+#define MINSIZE		(1UL << 4)
+#define MINDELETED	4
+/* $OpenBSD: ohash_create_entry.c,v 1.2 2004/06/22 20:00:16 espie Exp $ */
+/* ex:ts=8 sw=4: 
+ */
+
+/* Copyright (c) 1999, 2004 Marc Espie <espie@openbsd.org>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+
+/* This handles the common case of variable length keys, where the 
+ * key is stored at the end of the record.
+ */
+void *
+ohash_create_entry(struct ohash_info *i, const char *start, const char **end)
+{
+	char *p;
+
+	if (!*end)
+		*end = start + strlen(start);
+	p = (i->alloc)(i->key_offset + (*end - start) + 1, i->data);
+	if (p) {
+	    memcpy(p+i->key_offset, start, *end-start);
+	    p[i->key_offset + (*end - start)] = '\0';
+	}
+	return (void *)p;
+}
+
+/* hash_delete only frees the hash structure. Use hash_first/hash_next
+ * to free entries as well.  */
+void 
+ohash_delete(struct ohash *h)
+{
+	(h->info.hfree)(h->t, sizeof(struct _ohash_record) * h->size,
+		h->info.data);
+#ifndef NDEBUG
+	h->t = NULL;
+#endif
+}
+
+static void ohash_resize(struct ohash *);
+
+static void 
+ohash_resize(struct ohash *h)
+{
+	struct _ohash_record *n;
+	unsigned int 	ns, j;
+	unsigned int	i, incr;
+
+	if (4 * h->deleted < h->total)
+		ns = h->size << 1;
+	else if (3 * h->deleted > 2 * h->total)
+		ns = h->size >> 1;
+	else
+		ns = h->size;
+	if (ns < MINSIZE)
+		ns = MINSIZE;
+#ifdef STATS_HASH
+	STAT_HASH_EXPAND++;
+	STAT_HASH_SIZE += ns - h->size;
+#endif
+	n = (h->info.halloc)(sizeof(struct _ohash_record) * ns, h->info.data);
+	if (!n)
+		return;
+
+	for (j = 0; j < h->size; j++) {
+		if (h->t[j].p != NULL && h->t[j].p != DELETED) {
+			i = h->t[j].hv % ns;
+			incr = ((h->t[j].hv % (ns - 2)) & ~1) + 1;
+			while (n[i].p != NULL) {
+				i += incr;
+				if (i >= ns)
+					i -= ns;
+		    	}
+			n[i].hv = h->t[j].hv;
+			n[i].p = h->t[j].p;
+		}
+	}
+	(h->info.hfree)(h->t, sizeof(struct _ohash_record) * h->size, 
+		h->info.data);
+	h->t = n;
+	h->size = ns;
+	h->total -= h->deleted;
+	h->deleted = 0;
+}
+
+void *
+ohash_remove(struct ohash *h, unsigned int i)
+{
+	void 		*result = (void *)h->t[i].p;
+
+	if (result == NULL || result == DELETED)
+		return NULL;
+
+#ifdef STATS_HASH
+	STAT_HASH_ENTRIES--;
+#endif
+	h->t[i].p = DELETED;
+	h->deleted++;
+	if (h->deleted >= MINDELETED && 4 * h->deleted > h->total)
+		ohash_resize(h);
+	return result;
+}
+
+void *
+ohash_find(struct ohash *h, unsigned int i)
+{
+	if (h->t[i].p == DELETED)
+		return NULL;
+	else
+		return (void *)h->t[i].p;
+}
+
+void *
+ohash_insert(struct ohash *h, unsigned int i, void *p)
+{
+#ifdef STATS_HASH
+	STAT_HASH_ENTRIES++;
+#endif
+	if (h->t[i].p == DELETED) {
+		h->deleted--;
+		h->t[i].p = p;
+	} else {
+		h->t[i].p = p;
+	/* Arbitrary resize boundary.  Tweak if not efficient enough.  */
+		if (++h->total * 4 > h->size * 3)
+			ohash_resize(h);
+	}
+    	return p;
+}
+
+unsigned int
+ohash_entries(struct ohash *h)
+{
+	return h->total - h->deleted;
+}
+
+void *
+ohash_first(struct ohash *h, unsigned int *pos)
+{
+	*pos = 0;
+	return ohash_next(h, pos);
+}
+	
+void *
+ohash_next(struct ohash *h, unsigned int *pos)
+{
+	for (; *pos < h->size; (*pos)++) 
+		if (h->t[*pos].p != DELETED && h->t[*pos].p != NULL) 
+			return (void *)h->t[(*pos)++].p;
+	return NULL;
+}
+
+void 
+ohash_init(struct ohash *h, unsigned int size, struct ohash_info *info)
+{
+	h->size = 1UL << size;
+	if (h->size < MINSIZE)
+		h->size = MINSIZE;
+#ifdef STATS_HASH
+	STAT_HASH_CREATION++;
+	STAT_HASH_SIZE += h->size;
+#endif
+	/* Copy info so that caller may free it.  */
+	h->info.key_offset = info->key_offset;
+	h->info.halloc = info->halloc;
+	h->info.hfree = info->hfree;
+	h->info.alloc = info->alloc;
+	h->info.data = info->data;
+	h->t = (h->info.halloc)(sizeof(struct _ohash_record) * h->size,
+	    h->info.data);
+	h->total = h->deleted = 0;
+}
+
+uint32_t
+ohash_interval(const char *s, const char **e)
+{
+	uint32_t k;
+
+	if (!*e)
+		*e = s + strlen(s);
+	if (s == *e)
+		k = 0;
+	else
+		k = *s++;
+	while (s != *e)
+		k =  ((k << 2) | (k >> 30)) ^ *s++;
+	return k;
+}
+
+unsigned int
+ohash_lookup_interval(struct ohash *h, const char *start, const char *end, 
+    uint32_t hv)
+{
+	unsigned int 	i, incr;
+	unsigned int	empty;
+	
+#ifdef STATS_HASH
+	STAT_HASH_LOOKUP++;
+#endif
+	empty = NONE;
+	i = hv % h->size;
+	incr = ((hv % (h->size-2)) & ~1) + 1;
+	while (h->t[i].p != NULL) {
+#ifdef STATS_HASH
+		STAT_HASH_LENGTH++;
+#endif
+		if (h->t[i].p == DELETED) {
+			if (empty == NONE)
+				empty = i;
+		} else if (h->t[i].hv == hv && 
+		    strncmp(h->t[i].p+h->info.key_offset, start, 
+		    	end - start) == 0 &&
+		    (h->t[i].p+h->info.key_offset)[end-start] == '\0') {
+		    	if (empty != NONE) {
+				h->t[empty].hv = hv;
+				h->t[empty].p = h->t[i].p;
+				h->t[i].p = DELETED;
+				return empty;
+			} else {
+#ifdef STATS_HASH
+				STAT_HASH_POSITIVE++;
+#endif
+				return i;
+			}
+		}
+		i += incr;
+		if (i >= h->size) 
+			i -= h->size;
+	}
+
+	/* Found an empty position.  */
+	if (empty != NONE) 
+		i = empty;
+	h->t[i].hv = hv;
+	return i;
+}
+
+unsigned int
+ohash_lookup_memory(struct ohash *h, const char *k, size_t size, uint32_t hv)
+{
+	unsigned int	i, incr;
+	unsigned int	empty;
+	
+#ifdef STATS_HASH
+	STAT_HASH_LOOKUP++;
+#endif
+	empty = NONE;
+	i = hv % h->size;
+	incr = ((hv % (h->size-2)) & ~1) + 1;
+	while (h->t[i].p != NULL) {
+#ifdef STATS_HASH
+		STAT_HASH_LENGTH++;
+#endif
+		if (h->t[i].p == DELETED) {
+			if (empty == NONE)
+				empty = i;
+		} else if (h->t[i].hv == hv && 
+		    memcmp(h->t[i].p+h->info.key_offset, k, size) == 0) {
+		    	if (empty != NONE) {
+				h->t[empty].hv = hv;
+				h->t[empty].p = h->t[i].p;
+				h->t[i].p = DELETED;
+				return empty;
+			} else {
+#ifdef STATS_HASH
+				STAT_HASH_POSITIVE++;
+#endif
+			}	return i;
+		}
+		i += incr;
+		if (i >= h->size) 
+			i -= h->size;
+	}
+
+	/* Found an empty position.  */
+	if (empty != NONE) 
+		i = empty;
+	h->t[i].hv = hv;
+	return i;
+}
+
+unsigned int
+ohash_qlookup(struct ohash *h, const char *s)
+{
+	const char *e = NULL;
+	return ohash_qlookupi(h, s, &e);
+}
+
+unsigned int
+ohash_qlookupi(struct ohash *h, const char *s, const char **e)
+{
+	u_int32_t hv;
+
+	hv = ohash_interval(s, e);
+	return ohash_lookup_interval(h, s, *e, hv);
+}
+
+#endif /*!HAVE_OHASH*/
diff --git a/compat_ohash.h b/compat_ohash.h
new file mode 100644
index 00000000..85cf9114
--- /dev/null
+++ b/compat_ohash.h
@@ -0,0 +1,73 @@
+#ifndef OHASH_H
+#define OHASH_H
+/* $OpenBSD: ohash.h,v 1.9 2006/01/16 15:52:25 espie Exp $ */
+/* ex:ts=8 sw=4: 
+ */
+
+/* Copyright (c) 1999, 2004 Marc Espie <espie@openbsd.org>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+
+/* Open hashing support. 
+ * Open hashing was chosen because it is much lighter than other hash
+ * techniques, and more efficient in most cases.
+ */
+
+struct ohash_info {
+	ptrdiff_t key_offset;
+	void *data;	/* user data */
+	void *(*halloc)(size_t, void *);
+	void (*hfree)(void *, size_t, void *);
+	void *(*alloc)(size_t, void *);
+};
+
+struct _ohash_record;
+
+struct ohash {
+	struct _ohash_record 	*t;
+	struct ohash_info 	info;
+	unsigned int 		size;
+	unsigned int 		total;
+	unsigned int 		deleted;
+};
+
+/* For this to be tweakable, we use small primitives, and leave part of the
+ * logic to the client application.  e.g., hashing is left to the client
+ * application.  We also provide a simple table entry lookup that yields
+ * a hashing table index (opaque) to be used in find/insert/remove.
+ * The keys are stored at a known position in the client data.
+ */
+__BEGIN_DECLS
+void ohash_init(struct ohash *, unsigned, struct ohash_info *);
+void ohash_delete(struct ohash *);
+
+unsigned int ohash_lookup_interval(struct ohash *, const char *,
+	    const char *, uint32_t);
+unsigned int ohash_lookup_memory(struct ohash *, const char *,
+	    size_t, uint32_t)
+		__attribute__ ((__bounded__(__string__,2,3)));
+void *ohash_find(struct ohash *, unsigned int);
+void *ohash_remove(struct ohash *, unsigned int);
+void *ohash_insert(struct ohash *, unsigned int, void *);
+void *ohash_first(struct ohash *, unsigned int *);
+void *ohash_next(struct ohash *, unsigned int *);
+unsigned int ohash_entries(struct ohash *);
+
+void *ohash_create_entry(struct ohash_info *, const char *, const char **);
+u_int32_t ohash_interval(const char *, const char **);
+
+unsigned int ohash_qlookupi(struct ohash *, const char *, const char **);
+unsigned int ohash_qlookup(struct ohash *, const char *);
+__END_DECLS
+#endif
diff --git a/mandocdb.c b/mandocdb.c
index b3730139..78aefad4 100644
--- a/mandocdb.c
+++ b/mandocdb.c
@@ -1,4 +1,4 @@
-/*	$Id: mandocdb.c,v 1.52 2012/06/08 15:06:28 kristaps Exp $ */
+/*	$Id: mandocdb.c,v 1.53 2012/06/09 11:00:13 kristaps Exp $ */
 /*
  * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
  * Copyright (c) 2011, 2012 Ingo Schwarze <schwarze@openbsd.org>
@@ -34,7 +34,11 @@
 #include <string.h>
 #include <unistd.h>
 
+#ifdef HAVE_OHASH
 #include <ohash.h>
+#else
+#include "compat_ohash.h"
+#endif
 #include <sqlite3.h>
 
 #include "mdoc.h"
@@ -327,6 +331,7 @@ main(int argc, char *argv[])
 {
 	char		  cwd[MAXPATHLEN];
 	int		  ch, rc, fd, i;
+	unsigned int	  index;
 	size_t		  j, sz;
 	const char	 *dir;
 	struct str	 *s;
@@ -521,8 +526,8 @@ out:
 	manpath_free(&dirs);
 	mchars_free(mc);
 	mparse_free(mp);
-	for (s = ohash_first(&strings, &ch);
-			NULL != s; s = ohash_next(&strings, &ch)) {
+	for (s = ohash_first(&strings, &index);
+			NULL != s; s = ohash_next(&strings, &index)) {
 		if (s->utf8 != s->key)
 			free(s->utf8);
 		free(s);
diff --git a/mansearch.c b/mansearch.c
index e3502b99..70e6ad05 100644
--- a/mansearch.c
+++ b/mansearch.c
@@ -1,4 +1,4 @@
-/*	$Id: mansearch.c,v 1.3 2012/06/08 15:06:28 kristaps Exp $ */
+/*	$Id: mansearch.c,v 1.4 2012/06/09 11:00:13 kristaps Exp $ */
 /*
  * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
  *
@@ -30,7 +30,11 @@
 #include <string.h>
 #include <unistd.h>
 
+#ifdef HAVE_OHASH
 #include <ohash.h>
+#else
+#include "compat_ohash.h"
+#endif
 #include <sqlite3.h>
 
 #include "mandoc.h"
diff --git a/test-ohash.c b/test-ohash.c
new file mode 100644
index 00000000..72b3e3dd
--- /dev/null
+++ b/test-ohash.c
@@ -0,0 +1,19 @@
+#include <stdint.h>
+#include <stddef.h>
+#include <stdlib.h>
+#include <ohash.h>
+
+void *xalloc(size_t sz, void *arg) { return(calloc(sz,1)); }
+void xfree(void *p, size_t sz, void *arg) { free(p); }
+
+int
+main(void)
+{
+	struct ohash h;
+	struct ohash_info i;
+	i.halloc = i.alloc = xalloc;
+	i.free = free;
+	ohash_init(&h, 2, &i);
+	ohash_delete(&h);
+	return 0;
+}