/*
- * Copyright (C) 1994, Index Data I/S
+ * Copyright (C) 1994-1996, Index Data I/S
* All rights reserved.
* Sebastian Hammer, Adam Dickmeiss
*
* $Log: insert.c,v $
- * Revision 1.8 1994-09-16 12:35:01 adam
+ * Revision 1.17 1996-05-14 15:49:09 adam
+ * Bug fix: In function split_page. In rare cases variable best_indxp was
+ * referenced.
+ *
+ * Revision 1.16 1996/02/02 13:43:50 adam
+ * The public functions simply use char instead of Dict_char to represent
+ * search strings. Dict_char is used internally only.
+ *
+ * Revision 1.15 1996/02/01 20:39:59 adam
+ * Bug fix: insert didn't work on 8-bit characters due to unsigned char
+ * compares in dict_strcmp (strcmp) and signed Dict_char. Dict_char is
+ * unsigned now.
+ *
+ * Revision 1.14 1995/12/07 11:48:56 adam
+ * Insert operation obeys DICT_type = 1 (slack in page).
+ * Function dict_open exists if page size or magic aren't right.
+ *
+ * Revision 1.13 1995/11/28 09:06:37 adam
+ * Fixed potential dangling pointer.
+ *
+ * Revision 1.12 1995/09/06 10:34:44 adam
+ * Memcpy in clean_page edited to satisfy checkergcc.
+ *
+ * Revision 1.11 1995/09/04 12:33:31 adam
+ * Various cleanup. YAZ util used instead.
+ *
+ * Revision 1.10 1994/10/05 12:16:48 adam
+ * Pagesize is a resource now.
+ *
+ * Revision 1.9 1994/09/16 15:39:13 adam
+ * Initial code of lookup - not tested yet.
+ *
+ * Revision 1.8 1994/09/16 12:35:01 adam
* New version of split_page which use clean_page for splitting.
*
* Revision 1.7 1994/09/12 08:06:42 adam
void *subp;
char *info_here;
Dict_ptr subptr;
- int i, need;
- short *indxp, *best_indxp;
+ int i, j;
+ short *indxp, *best_indxp = NULL;
Dict_char best_char = 0;
Dict_char prev_char = 0;
int best_no = -1, no_current = 1;
- indxp = (short*) ((char*) p+DICT_PAGESIZE-sizeof(short));
+ /* determine splitting char... */
+ indxp = (short*) ((char*) p+DICT_pagesize(dict)-sizeof(short));
for (i = DICT_nodir (p); --i >= 0; --indxp)
{
if (*indxp > 0) /* tail string here! */
{ /* first entry met */
best_char = prev_char = dc;
best_no = 1;
+ best_indxp = indxp;
}
else if (prev_char == dc)
{ /* same char prefix. update */
if (best_no < 0) /* we didn't find any tail string entry at all! */
return -1;
+ j = best_indxp - (short*) p;
subptr = new_page (dict, ptr, &subp);
/* scan entries to see if there is a string with */
/* length 1. info_here indicates if such entry exist */
info_here = NULL;
- for (indxp=best_indxp, i=0; i<best_no; i++, indxp++)
+ for (i=0; i<best_no; i++, j++)
{
- char *info;
+ char *info, *info1;
int slen;
-
- assert (*indxp > 0);
+ Dict_char dc;
- info = (char*) p + *indxp; /* entry start */
- assert (*info == best_char);
- slen = dict_strlen(info);
+
+ info = (char*) p + ((short*) p)[j];
+ /* entry start */
+ memcpy (&dc, info, sizeof(dc));
+ assert (dc == best_char);
+ slen = dict_strlen((Dict_char*) info);
assert (slen > 0);
if (slen == 1)
assert (!info_here);
info_here = info+(slen+1)*sizeof(Dict_char);
}
- }
- /* calculate the amount of bytes needed for this entry when */
- /* transformed to a sub entry */
- need = sizeof(Dict_char)+sizeof(Dict_ptr)+1;
- if (info_here)
- need += *info_here;
-
- indxp = best_indxp;
- /* now loop on all entries with string length > 1 i.e. all */
- /* those entries which contribute to a sub page */
- best_indxp = NULL;
- for (i=0; i<best_no; i++, indxp++)
- {
- char *info, *info1;
- int slen;
-
- assert (*indxp > 0);
-
- info = (char*) p + *indxp; /* entry start */
- assert (*info == best_char);
- slen = dict_strlen(info);
-
- if (slen > 1)
+ else
{
info1 = info+(1+slen)*sizeof(Dict_char); /* info start */
-
- if (need <= (1+slen)*sizeof(Dict_char) + 1 + *info1)
- best_indxp = indxp; /* space for entry */
- dict_ins (dict, info+sizeof(Dict_char), subptr, *info1, info1+1);
+ dict_ins (dict, (Dict_char*) (info+sizeof(Dict_char)),
+ subptr, *info1, info1+1);
dict_bf_readp (dict->dbf, ptr, &p);
}
}
-#if 1
/* now clean the page ... */
clean_page (dict, ptr, p, &best_char, subptr, info_here);
return 0;
-#endif
- if (best_indxp)
- { /* there was a hole big enough for a sub entry */
- char *info = (char*) p + *best_indxp;
- short *indxp1;
-
- *--indxp = - *best_indxp;
- DICT_type(p) = 1;
- DICT_nodir (p) -= (best_no-1);
- indxp1 = (short*)((char*)p+DICT_PAGESIZE-DICT_nodir(p)*sizeof(short));
- while (indxp != indxp1)
- {
- --indxp;
- *indxp = indxp[1-best_no];
- }
- memcpy (info, &subptr, sizeof(Dict_ptr)); /* store subptr */
- info += sizeof(Dict_ptr);
- memcpy (info, &best_char, sizeof(Dict_char)); /* store sub char */
- info += sizeof(Dict_char);
- if (info_here)
- memcpy (info, info_here, *info_here+1); /* with information */
- else
- *info = 0; /* without info */
-#if CHECK
- best_indxp = NULL;
- prev_char = 0;
- indxp = (short*) ((char*) p+DICT_PAGESIZE-sizeof(short));
- for (i = DICT_nodir (p); --i >= 0; --indxp)
- {
- if (*indxp > 0) /* tail string here! */
- {
- Dict_char dc;
-
- memcpy (&dc, (char*) p + *indxp, sizeof(dc));
- assert (dc != best_char);
- assert (dc >= prev_char);
- prev_char = dc;
- }
- else
- {
- Dict_char dc;
- memcpy (&dc, (char*)p - *indxp+sizeof(Dict_ptr),
- sizeof(dc));
- assert (dc > prev_char);
- if (dc == best_char)
- {
- assert (best_indxp == NULL);
- best_indxp = indxp;
- }
- prev_char = dc;
- }
- }
- assert (best_indxp);
-#endif
- }
- else
- {
- short *indxp1, *indxp2;
- assert (0);
- DICT_type(p) = 1;
- DICT_nodir(p) -= best_no;
- indxp2 = indxp;
- indxp1 = (short*)((char*) p+DICT_PAGESIZE-DICT_nodir(p)*sizeof(short));
- do
- {
- --indxp2;
- indxp2[0] = indxp2[-best_no];
- } while (indxp2 != indxp1);
- }
- return 0;
}
static void clean_page (Dict dict, Dict_ptr ptr, void *p, Dict_char *out,
- Dict_ptr subptr, char *userinfo)
+ Dict_ptr subptr, char *userinfo)
{
char *np = xmalloc (dict->head.page_size);
int i, slen, no = 0;
short *indxp1, *indxp2;
char *info1, *info2;
- indxp1 = (short*) ((char*) p+DICT_PAGESIZE-sizeof(short));
- indxp2 = (short*) ((char*) np+DICT_PAGESIZE);
+ indxp1 = (short*) ((char*) p+DICT_pagesize(dict)-sizeof(short));
+ indxp2 = (short*) ((char*) np+DICT_pagesize(dict));
info2 = (char*) np + DICT_infoffset;
for (i = DICT_nodir (p); --i >= 0; --indxp1)
{
continue;
}
*--indxp2 = info2 - np;
- slen = (dict_strlen(info1)+1)*sizeof(Dict_char);
+ slen = (dict_strlen((Dict_char*) info1)+1)*sizeof(Dict_char);
memcpy (info2, info1, slen);
info1 += slen;
info2 += slen;
info2 += slen;
++no;
}
+#if 1
+ memcpy ((char*)p+DICT_infoffset,
+ (char*)np+DICT_infoffset,
+ info2 - ((char*)np+DICT_infoffset));
+ memcpy ((char*)p + ((char*)indxp2 - (char*)np),
+ indxp2,
+ ((char*) np+DICT_pagesize(dict)) - (char*)indxp2);
+#else
memcpy ((char*)p+DICT_infoffset, (char*)np+DICT_infoffset,
- DICT_PAGESIZE-DICT_infoffset);
+ DICT_pagesize(dict)-DICT_infoffset);
+#endif
DICT_size(p) = info2 - np;
DICT_type(p) = 0;
DICT_nodir(p) = no;
xfree (np);
+ dict_bf_touch (dict->dbf, ptr);
}
static int dict_ins (Dict dict, const Dict_char *str,
Dict_ptr back_ptr, int userlen, void *userinfo)
{
- int hi, lo, mid, i, slen, cmp = 1;
+ int hi, lo, mid, slen, cmp = 1;
Dict_ptr ptr = back_ptr;
short *indxp;
char *info;
mid = lo = 0;
hi = DICT_nodir(p)-1;
- indxp = (short*) ((char*) p+DICT_PAGESIZE-sizeof(short));
+ indxp = (short*) ((char*) p+DICT_pagesize(dict)-sizeof(short));
while (lo <= hi)
{
mid = (lo+hi)/2;
cmp = dict_strcmp((Dict_char*) info, str);
if (!cmp)
{
- info += (dict_strlen(info)+1)*sizeof(Dict_char);
+ info += (dict_strlen((Dict_char*) info)+1)*sizeof(Dict_char);
/* consider change of userinfo length... */
if (*info == userlen)
{
}
if (DICT_size(p)+sizeof(Dict_char)+sizeof(Dict_ptr)+
userlen >=
- DICT_PAGESIZE - (1+DICT_nodir(p))*sizeof(short))
+ DICT_pagesize(dict) - (1+DICT_nodir(p))*sizeof(short))
{
if (DICT_type(p) == 1)
{
clean_page (dict, ptr, p, NULL, 0, NULL);
- dict_bf_touch (dict->dbf, ptr);
return dict_ins (dict, str-1, ptr,
userlen, userinfo);
}
if (split_page (dict, ptr, p))
{
- log (LOG_FATAL, "Unable to split page %d\n", ptr);
+ logf (LOG_FATAL, "Unable to split page %d\n", ptr);
abort ();
}
return dict_ins (dict, str-1, ptr, userlen, userinfo);
--indxp;
slen = (dict_strlen(str)+1)*sizeof(Dict_char);
if (DICT_size(p)+slen+userlen >=
- DICT_PAGESIZE - (1+DICT_nodir(p))*sizeof(short)) /* overflow? */
+ DICT_pagesize(dict) - (1+DICT_nodir(p))*sizeof(short)) /* overflow? */
{
- if (DICT_type(p) == 1)
+ if (DICT_type(p))
{
clean_page (dict, ptr, p, NULL, 0, NULL);
- dict_bf_touch (dict->dbf, ptr);
return dict_ins (dict, str, ptr, userlen, userinfo);
}
- i = 0;
- do
- {
- assert (i <= 1);
- if (split_page (dict, ptr, p))
- {
- log (LOG_FATAL, "Unable to split page %d\n", ptr);
- abort ();
- }
- if (DICT_size(p)+slen+userlen <
- DICT_PAGESIZE - (1+DICT_nodir(p))*sizeof(short))
- break;
- i++;
- clean_page (dict, ptr, p, NULL, 0, NULL);
- } while (DICT_size(p)+slen+userlen > DICT_PAGESIZE -
- (1+DICT_nodir(p))*sizeof(short));
+ split_page (dict, ptr, p);
return dict_ins (dict, str, ptr, userlen, userinfo);
}
if (cmp)
{
short *indxp1;
(DICT_nodir(p))++;
- indxp1 = (short*)((char*) p + DICT_PAGESIZE
+ indxp1 = (short*)((char*) p + DICT_pagesize(dict)
- DICT_nodir(p)*sizeof(short));
for (; indxp1 != indxp; indxp1++)
indxp1[0] = indxp1[1];
#if CHECK
- indxp1 = (short*) ((char*) p+DICT_PAGESIZE-sizeof(short));
+ indxp1 = (short*) ((char*) p+DICT_pagesize(dict)-sizeof(short));
for (i = DICT_nodir (p); --i >= 0; --indxp1)
{
if (*indxp1 < 0)
return 1;
}
-int dict_insert (Dict dict, const Dict_char *str, int userlen, void *userinfo)
+int dict_insert (Dict dict, const char *str, int userlen, void *userinfo)
{
assert (dict->head.last > 0);
if (dict->head.last == 1)
- return dict_ins (dict, str, 0, userlen, userinfo);
+ return dict_ins (dict, (const Dict_char *) str, 0, userlen, userinfo);
else
- return dict_ins (dict, str, 1, userlen, userinfo);
+ return dict_ins (dict, (const Dict_char *) str, 1, userlen, userinfo);
}