rbtree(S)

rbtree: rbt_delete, rbt_find, rbt_free, rbt_init, rbt_insert, rbt_walk -- manage balanced binary tree

Synopsis

   #include <rbtree.h>
   
   RBTree *rbt_init(size_t n,
        int (*cmp)(const void *, const void *),
        void *(*alloc)(size_t),
        void (*toss)(const void *));
   

   void rbt_walk(RBTree *rbtp, int order,
        void (*operate)(const void *));
   

   void rbt_free(RBTree *rbtp);
   

   void *rbt_find(RBTree *rbtp, const void *item);
   

   void *rbt_insert(RBTree *rbtp, const void *item);
   

   in rbt_delete(RBTree *rbtp, const void *item);

Description

These functions manage a ``balanced'' binary tree of arbitrary content. Each tree is created by calling rbt_init. The argument n gives the size of the data to be stored in each node, and the other three arguments are pointers to functions:

cmp: the comparison function used to order the tree
alloc: used to allocate node space
toss: used to deallocate node space.

These three function pointers can be null requesting the default functions, strcmp, malloc, and free respectively. The value returned is the handle for the just-created empty tree, which is the argument rbtp passed to the other functions. A tree is destroyed by calling rbt_free.

A tree is traversed by calling rbt_walk. The six different regular traversals are specified by the order argument. It is exactly one of RBT_PREORDER, RBT_INORDER, or RBT_POSTORDER, optionally combined (bitwise OR) with RBT_REVERSE. A ``preorder'' traversal visits the parent before either child, a ``postorder'' traversal visits the parent after any child, and an ``inorder'' traversal visits the parent between the two children. Unless RBT_REVERSE is specified, the children are visited in ascending order (as specified by the cmp argument). Visiting a node causes the function pointed to by the operate argument to be called, being passed a pointer to the start of that node's data.

For the remaining functions, item points to a particular data item to which the function is to be applied.

A data item can be searched for without modifying the tree by calling rbt_find. rbt_insert acts just like rbt_find except the data will be added to the tree if a match is not found. And, rbt_delete will remove the item from the tree if it was present.

Return values

rbt_init and rbt_insert return a null pointer for an allocation failure. Or, in other words, when the function pointed to by the alloc argument returns a null pointer. When rbt_insert returns a non-null pointer, it points to the start of the matching data.

rbt_find returns a null pointer if no match for the data is found; otherwise it returns a pointer to the start of the matching data.

rbt_delete returns nonzero when no match for the item is present in the tree; otherwise it returns zero having found a match and removed it from the tree.

References

malloc(S), string(S)

Usage

The following shows adding a collection of invocation parameter strings into a tree, and then printing them out, one per line, in descending order.

   #include <rbtree.h>
   #include <stdio.h>
   #include <string.h>
   
   static int compare(const void *p1, const void *p2) {
       return strcmp(*(const char **)p1, *(const char **)p2);
   }
   

   static void print(const void *p) {
       puts(*(const char **)p);
   }
   

   int main(int argc, char **argv) {
       const char *s;
       RBTree *rbtp;
   

       if ((rbtp = rbt_init(sizeof(const char *), &compare, 0, 0)) == 0)
   	return 2;
       while ((s = *++argv) != 0) {
   	if (rbt_insert(rbtp, &s) == 0)
   	    return 2;
       }
       rbt_walk(rbtp, RBT_INORDER|RBT_REVERSE, &print);
       rbt_free(rbtp);
       return 0;
   }

If the above is invoked with the three arguments ``aa cc bb'', it will output these three lines:

      cc
      bb
      aa

Algorithm

The tree is balanced using a ``Red Black'' data structure in which each binary node can be thought of as being all or a portion of a node in a balanced 2-3-4 tree. A balanced 2-3-4 tree always has the same distance from the root to any leaf. This is accomplished by adjusting the number of children of individual 2-3-4 tree nodes, since a 2-3-4 node is either a leaf (and thus has no children) or it has 2, 3 or 4 children.