diff -r 387414a7afd8 -r d607a184925a src/cx/tree.h --- a/src/cx/tree.h Sun Jul 07 14:56:44 2024 +0200 +++ b/src/cx/tree.h Tue Aug 20 18:02:39 2024 +0200 @@ -321,15 +321,41 @@ * positive if one of the children might contain the data, * negative if neither the node, nor the children contains the data */ -typedef int (*cx_tree_search_func)(void const *node, void const *data); +typedef int (*cx_tree_search_data_func)(void const *node, void const *data); + +/** + * Function pointer for a search function. + * + * A function of this kind shall check if the specified \p node + * contains the same \p data as \p new_node or if one of the children might + * contain the data. + * + * The function should use the returned integer to indicate how close the + * match is, where a negative number means that it does not match at all. + * + * For example if a tree stores file path information, a node that is + * describing a parent directory of a filename that is searched, shall + * return a positive number to indicate that a child node might contain the + * searched item. On the other hand, if the node denotes a path that is not a + * prefix of the searched filename, the function would return -1 to indicate + * that the search does not need to be continued in that branch. + * + * @param node the node that is currently investigated + * @param new_node a new node with the information which is searched + * + * @return 0 if \p node contains the same data as \p new_node, + * positive if one of the children might contain the data, + * negative if neither the node, nor the children contains the data + */ +typedef int (*cx_tree_search_func)(void const *node, void const *new_node); /** * Searches for data in a tree. * * When the data cannot be found exactly, the search function might return a * closest result which might be a good starting point for adding a new node - * to the tree. + * to the tree (see also #cx_tree_add()). * * Depending on the tree structure it is not necessarily guaranteed that the * "closest" match is uniquely defined. This function will search for a node @@ -348,9 +374,42 @@ * contain any node that might be related to the searched data */ __attribute__((__nonnull__)) +int cx_tree_search_data( + void const *root, + void const *data, + cx_tree_search_data_func sfunc, + void **result, + ptrdiff_t loc_children, + ptrdiff_t loc_next +); + +/** + * Searches for a node in a tree. + * + * When no node with the same data can be found, the search function might + * return a closest result which might be a good starting point for adding the + * new node to the tree (see also #cx_tree_add()). + * + * Depending on the tree structure it is not necessarily guaranteed that the + * "closest" match is uniquely defined. This function will search for a node + * with the best match according to the \p sfunc (meaning: the return value of + * \p sfunc which is closest to zero). If that is also ambiguous, an arbitrary + * node matching the criteria is returned. + * + * @param root the root node + * @param node the node to search for + * @param sfunc the search function + * @param result where the result shall be stored + * @param loc_children offset in the node struct for the children linked list + * @param loc_next offset in the node struct for the next pointer + * @return zero if the node was found exactly, positive if a node was found that + * could contain the node (but doesn't right now), negative if the tree does not + * contain any node that might be related to the searched data + */ +__attribute__((__nonnull__)) int cx_tree_search( void const *root, - void const *data, + void const *node, cx_tree_search_func sfunc, void **result, ptrdiff_t loc_children, @@ -410,6 +469,192 @@ ptrdiff_t loc_next ); +/** + * Describes a function that creates a tree node from the specified data. + * The first argument points to the data the node shall contain and + * the second argument may be used for additional data (e.g. an allocator). + * Functions of this type shall either return a new pointer to a newly + * created node or \c NULL when allocation fails. + * + * \note the function may leave the node pointers in the struct uninitialized. + * The caller is responsible to set them according to the intended use case. + */ +typedef void *(*cx_tree_node_create_func)(void const *, void *); + +/** + * The local search depth for a new subtree when adding multiple elements. + * The default value is 3. + * This variable is used by #cx_tree_add_array() and #cx_tree_add_iter() to + * implement optimized insertion of multiple elements into a tree. + */ +extern unsigned int cx_tree_add_look_around_depth; + +/** + * Adds multiple elements efficiently to a tree. + * + * Once an element cannot be added to the tree, this function returns, leaving + * the iterator in a valid state pointing to the element that could not be + * added. + * Also, the pointer of the created node will be stored to \p failed. + * The integer returned by this function denotes the number of elements obtained + * from the \p iter that have been successfully processed. + * When all elements could be processed, a \c NULL pointer will be written to + * \p failed. + * + * The advantage of this function compared to multiple invocations of + * #cx_tree_add() is that the search for the insert locations is not always + * started from the root node. + * Instead, the function checks #cx_tree_add_look_around_depth many parent nodes + * of the current insert location before starting from the root node again. + * When the variable is set to zero, only the last found location is checked + * again. + * + * Refer to the documentation of #cx_tree_add() for more details. + * + * @param iter a pointer to an arbitrary iterator + * @param sfunc a search function + * @param cfunc a node creation function + * @param cdata optional additional data + * @param root the root node of the tree + * @param failed location where the pointer to a failed node shall be stored + * @param loc_parent offset in the node struct for the parent pointer + * @param loc_children offset in the node struct for the children linked list + * @param loc_last_child optional offset in the node struct for the pointer to + * the last child in the linked list (negative if there is no such pointer) + * @param loc_prev offset in the node struct for the prev pointer + * @param loc_next offset in the node struct for the next pointer + * @return the number of nodes created and added + * @see cx_tree_add() + */ +__attribute__((__nonnull__(1, 2, 3, 5, 6))) +size_t cx_tree_add_iter( + struct cx_iterator_base_s *iter, + cx_tree_search_func sfunc, + cx_tree_node_create_func cfunc, + void *cdata, + void **failed, + void *root, + ptrdiff_t loc_parent, + ptrdiff_t loc_children, + ptrdiff_t loc_last_child, + ptrdiff_t loc_prev, + ptrdiff_t loc_next +); + +/** + * Adds multiple elements efficiently to a tree. + * + * Once an element cannot be added to the tree, this function returns, storing + * the pointer of the created node to \p failed. + * The integer returned by this function denotes the number of elements from + * the \p src array that have been successfully processed. + * When all elements could be processed, a \c NULL pointer will be written to + * \p failed. + * + * The advantage of this function compared to multiple invocations of + * #cx_tree_add() is that the search for the insert locations is not always + * started from the root node. + * Instead, the function checks #cx_tree_add_look_around_depth many parent nodes + * of the current insert location before starting from the root node again. + * When the variable is set to zero, only the last found location is checked + * again. + * + * Refer to the documentation of #cx_tree_add() for more details. + * + * @param src a pointer to the source data array + * @param num the number of elements in the \p src array + * @param elem_size the size of each element in the \p src array + * @param sfunc a search function + * @param cfunc a node creation function + * @param cdata optional additional data + * @param failed location where the pointer to a failed node shall be stored + * @param root the root node of the tree + * @param loc_parent offset in the node struct for the parent pointer + * @param loc_children offset in the node struct for the children linked list + * @param loc_last_child optional offset in the node struct for the pointer to + * the last child in the linked list (negative if there is no such pointer) + * @param loc_prev offset in the node struct for the prev pointer + * @param loc_next offset in the node struct for the next pointer + * @return the number of array elements successfully processed + * @see cx_tree_add() + */ +__attribute__((__nonnull__(1, 4, 5, 7, 8))) +size_t cx_tree_add_array( + void const *src, + size_t num, + size_t elem_size, + cx_tree_search_func sfunc, + cx_tree_node_create_func cfunc, + void *cdata, + void **failed, + void *root, + ptrdiff_t loc_parent, + ptrdiff_t loc_children, + ptrdiff_t loc_last_child, + ptrdiff_t loc_prev, + ptrdiff_t loc_next +); + +/** + * Adds data to a tree. + * + * An adequate location where to add the new tree node is searched with the + * specified \p sfunc. + * + * When a location is found, the \p cfunc will be invoked with \p cdata. + * + * The node returned by \p cfunc will be linked into the tree. + * When \p sfunc returned a positive integer, the new node will be linked as a + * child. The other children (now siblings of the new node) are then checked + * with \p sfunc, whether they could be children of the new node and re-linked + * accordingly. + * + * When \p sfunc returned zero and the found node has a parent, the new + * node will be added as sibling - otherwise, the new node will be added + * as a child. + * + * When \p sfunc returned a negative value, the new node will not be added to + * the tree and this function returns a non-zero value. + * The caller should check if \p cnode contains a node pointer and deal with the + * node that could not be added. + * + * This function also returns a non-zero value when \p cfunc tries to allocate + * a new node but fails to do so. In that case, the pointer stored to \p cnode + * will be \c NULL. + * + * Multiple elements can be added more efficiently with + * #cx_tree_add_array() or #cx_tree_add_iter(). + * + * @param src a pointer to the data + * @param sfunc a search function + * @param cfunc a node creation function + * @param cdata optional additional data + * @param cnode the location where a pointer to the new node is stored + * @param root the root node of the tree + * @param loc_parent offset in the node struct for the parent pointer + * @param loc_children offset in the node struct for the children linked list + * @param loc_last_child optional offset in the node struct for the pointer to + * the last child in the linked list (negative if there is no such pointer) + * @param loc_prev offset in the node struct for the prev pointer + * @param loc_next offset in the node struct for the next pointer + * @return zero when a new node was created and added to the tree, + * non-zero otherwise + */ +__attribute__((__nonnull__(1, 2, 3, 5, 6))) +int cx_tree_add( + void const *src, + cx_tree_search_func sfunc, + cx_tree_node_create_func cfunc, + void *cdata, + void **cnode, + void *root, + ptrdiff_t loc_parent, + ptrdiff_t loc_children, + ptrdiff_t loc_last_child, + ptrdiff_t loc_prev, + ptrdiff_t loc_next +); + #ifdef __cplusplus } // extern "C" #endif