Began making changes to do better handling of character sets.

[citadel.git] / webcit / tools.c

diff --git a/webcit/tools.c b/webcit/tools.c
index 9204a74cd06a646589f304169667d4e2a2fd8d95..0a0718381d7ae3d388ea1c6fdb7ac45b8d2d0669 100644 (file)
--- a/webcit/tools.c
+++ b/webcit/tools.c
@@ -3,6 +3,7 @@
  */
 /**
  * \defgroup MiscRout Miscellaneous routines 
+ * \ingroup tools
  */
 
 /*@{*/
@@ -44,7 +45,7 @@ char *safestrncpy(char *dest, const char *src, size_t n)
  */
 int num_tokens(char *source, char tok)
 {
-       int a;
+       int a = 0;
        int count = 1;
 
        if (source == NULL)
@@ -547,4 +548,71 @@ void CtdlMakeTempFileName(char *name, int len) {
 
 
 
+/*
+ * \brief      case-insensitive substring search
+ *
+ *             This uses the Boyer-Moore search algorithm and is therefore quite fast.
+ *             The code is roughly based on the strstr() replacement from 'tin' written
+ *             by Urs Jannsen.
+ *
+ * \param      text    String to be searched
+ * \param      pattern String to search for
+ */
+char *bmstrcasestr(char *text, char *pattern) {
+
+       register unsigned char *p, *t;
+       register int i, j, *delta;
+       register size_t p1;
+       int deltaspace[256];
+       size_t textlen;
+       size_t patlen;
+
+       textlen = strlen (text);
+       patlen = strlen (pattern);
+
+       /* algorithm fails if pattern is empty */
+       if ((p1 = patlen) == 0)
+               return (text);
+
+       /* code below fails (whenever i is unsigned) if pattern too long */
+       if (p1 > textlen)
+               return (NULL);
+
+       /* set up deltas */
+       delta = deltaspace;
+       for (i = 0; i <= 255; i++)
+               delta[i] = p1;
+       for (p = (unsigned char *) pattern, i = p1; --i > 0;)
+               delta[tolower(*p++)] = i;
+
+       /*
+        * From now on, we want patlen - 1.
+        * In the loop below, p points to the end of the pattern,
+        * t points to the end of the text to be tested against the
+        * pattern, and i counts the amount of text remaining, not
+        * including the part to be tested.
+        */
+       p1--;
+       p = (unsigned char *) pattern + p1;
+       t = (unsigned char *) text + p1;
+       i = textlen - patlen;
+       while(1) {
+               if (tolower(p[0]) == tolower(t[0])) {
+                       if (strncasecmp ((const char *)(p - p1), (const char *)(t - p1), p1) == 0) {
+                               return ((char *)t - p1);
+                       }
+               }
+               j = delta[tolower(t[0])];
+               if (i < j)
+                       break;
+               i -= j;
+               t += j;
+       }
+       return (NULL);
+}
+
+
+
+
+
 /*@}*/