rizin/librz/util/intervaltree.c
Florian Märkl f430f28c02
Make rz_interval_tree_insert return the node (#6613)
There are APIs for which the node is needed, so it makes sense to return
it directly on insertion instead of only the boolean success state.
2026-07-11 14:08:35 +02:00

300 lines
9.7 KiB
C

// SPDX-FileCopyrightText: 2019 thestr4ng3r <info@florianmaerkl.de>
// SPDX-License-Identifier: LGPL-3.0-only
#include <rz_util/rz_intervaltree.h>
#include <rz_util/rz_assert.h>
#define unwrap(rbnode) ((rbnode) ? container_of(rbnode, RzIntervalNode, node) : NULL)
static void node_max(RBNode *node) {
RzIntervalNode *intervalnode = unwrap(node);
intervalnode->max_end = intervalnode->end;
int i;
for (i = 0; i < 2; i++) {
if (node->child[i]) {
ut64 end = unwrap(node->child[i])->max_end;
if (end > intervalnode->max_end) {
intervalnode->max_end = end;
}
}
}
}
static int cmp(const void *incoming, const RBNode *in_tree, void *user) {
ut64 incoming_start = *(ut64 *)incoming;
ut64 other_start = container_of(in_tree, const RzIntervalNode, node)->start;
if (incoming_start < other_start) {
return -1;
}
if (incoming_start > other_start) {
return 1;
}
return 0;
}
// like cmp, but handles searches for an exact RzIntervalNode * in the tree instead of only comparing the start values
static int cmp_exact_node(const void *incoming, const RBNode *in_tree, void *user) {
RzIntervalNode *incoming_node = (RzIntervalNode *)incoming;
const RzIntervalNode *node = container_of(in_tree, const RzIntervalNode, node);
if (node == incoming_node) {
return 0;
}
if (incoming_node->start < node->start) {
return -1;
}
if (incoming_node->start > node->start) {
return 1;
}
// Here we have the same start value, but a different pointer.
// This means we need to guide the caller into the direction where the actual node is.
// Since we have nothing to compare anymore, we have to iterate through all the same-start children to find the correct path.
RBIter *path_cache = user;
if (!path_cache->len) {
RBNode *cur = (RBNode *)&node->node;
// go down to the leftmost child that has the same start
while (cur) {
path_cache->path[path_cache->len++] = cur;
if (incoming_node->start <= unwrap(cur)->start) {
cur = cur->child[0];
} else {
cur = cur->child[1];
}
}
// iterate through all children with the same start and stop when the pointer is identical
// The RBIter works a bit different than normal here. We store each node in the path, including right-descended ones
// because we want to get the full path in the end.
while (rz_rbtree_iter_has(path_cache)) {
RzIntervalNode *intervalnode = rz_rbtree_iter_get(path_cache, RzIntervalNode, node);
if (intervalnode == incoming_node || intervalnode->start > incoming_node->start) {
break;
}
// rz_rbtree_iter_next does not work here
RBNode *rbnode = &intervalnode->node;
if (rbnode->child[1]) {
// next node after the current is always the leftmost in the right branch
for (rbnode = rbnode->child[1]; rbnode; rbnode = rbnode->child[0]) {
path_cache->path[path_cache->len++] = rbnode;
}
} else {
// if there is no right branch, go up
do {
rbnode = path_cache->path[--path_cache->len];
} while (path_cache->len && path_cache->path[path_cache->len - 1]->child[1] == rbnode);
}
}
}
RBNode *next_child = NULL;
// Go through the path to find the next node one step down
size_t i;
for (i = 0; i < path_cache->len - 1; i++) {
if (unwrap(path_cache->path[i]) == node) {
next_child = path_cache->path[i + 1];
break;
}
}
// Determine the direction from the next child node
return (next_child && node->node.child[0] == next_child) ? -1 : 1;
}
RZ_API void rz_interval_tree_init(RzIntervalTree *tree, RzIntervalNodeFree free) {
tree->root = NULL;
tree->free = free;
}
static void interval_node_free(RBNode *node, void *user) {
RzIntervalNode *ragenode /* >:-O */ = unwrap(node);
if (user) {
((RContRBFree)user)(ragenode->data);
}
free(ragenode);
}
RZ_API void rz_interval_tree_fini(RzIntervalTree *tree) {
if (!tree || !tree->root) {
return;
}
rz_rbtree_free(&tree->root->node, interval_node_free, tree->free);
}
/**
* \brief Insert an element into the interval tree
* \param start Lowest value covered by the element
* \param end May be the inclusive or exclusive end of the interval. This is determined only by how it is queried later.
* \return The newly created node or NULL if the insertion failed.
*/
RZ_API RZ_NULLABLE RzIntervalNode *rz_interval_tree_insert(RzIntervalTree *tree, ut64 start, ut64 end, void *data) {
rz_return_val_if_fail(tree && end >= start, false);
RzIntervalNode *node = RZ_NEW0(RzIntervalNode);
if (!node) {
return false;
}
node->start = start;
node->end = end;
node->data = data;
RBNode *root = tree->root ? &tree->root->node : NULL;
bool r = rz_rbtree_aug_insert(&root, &start, &node->node, cmp, NULL, node_max);
tree->root = unwrap(root);
if (!r) {
free(node);
return NULL;
}
return node;
}
RZ_API bool rz_interval_tree_delete(RzIntervalTree *tree, RzIntervalNode *node, bool free) {
RBNode *root = &tree->root->node;
RBIter path_cache = { 0 };
bool r = rz_rbtree_aug_delete(&root, node, cmp_exact_node, &path_cache, interval_node_free, free ? tree->free : NULL, node_max);
tree->root = unwrap(root);
return r;
}
RZ_API bool rz_interval_tree_resize(RzIntervalTree *tree, RzIntervalNode *node, ut64 new_start, ut64 new_end) {
rz_return_val_if_fail(new_end >= new_start, false);
if (node->start != new_start) {
// Start change means the tree needs a different structure
void *data = node->data;
if (!rz_interval_tree_delete(tree, node, false)) {
return false;
}
return rz_interval_tree_insert(tree, new_start, new_end, data) != NULL;
}
if (node->end != new_end) {
// Only end change just needs the updated augmented max value to be propagated upwards
node->end = new_end;
RBIter path_cache = { 0 };
return rz_rbtree_aug_update_sum(&tree->root->node, node, &node->node, cmp_exact_node, &path_cache, node_max);
}
// no change
return true;
}
// This must always return the topmost node that matches start!
// Otherwise rz_interval_tree_first_at will break!!!
RZ_API RzIntervalNode *rz_interval_tree_node_at(RzIntervalTree *tree, ut64 start) {
RzIntervalNode *node = tree->root;
while (node) {
if (start < node->start) {
node = unwrap(node->node.child[0]);
} else if (start > node->start) {
node = unwrap(node->node.child[1]);
} else {
return node;
}
}
return NULL;
}
RZ_API RBIter rz_interval_tree_first_at(RzIntervalTree *tree, ut64 start) {
RBIter it = { 0 };
// Find the topmost node matching start so we have a sub-tree with all entries that we want to find.
RzIntervalNode *top_intervalnode = rz_interval_tree_node_at(tree, start);
if (!top_intervalnode) {
return it;
}
// If there are more nodes with the same key, they can be in both children.
RBNode *node = &top_intervalnode->node;
while (node) {
if (start <= unwrap(node)->start) {
it.path[it.len++] = node;
node = node->child[0];
} else {
node = node->child[1];
}
}
return it;
}
RZ_API RzIntervalNode *rz_interval_tree_node_at_data(RzIntervalTree *tree, ut64 start, void *data) {
RBIter it = rz_interval_tree_first_at(tree, start);
while (rz_rbtree_iter_has(&it)) {
RzIntervalNode *intervalnode = rz_rbtree_iter_get(&it, RzIntervalNode, node);
if (intervalnode->start != start) {
break;
}
if (intervalnode->data == data) {
return intervalnode;
}
rz_rbtree_iter_next(&it);
}
return NULL;
}
RZ_API bool rz_interval_tree_all_at(RzIntervalTree *tree, ut64 start, RzIntervalIterCb cb, void *user) {
RBIter it = rz_interval_tree_first_at(tree, start);
bool ret = true;
while (rz_rbtree_iter_has(&it)) {
RzIntervalNode *intervalnode = rz_rbtree_iter_get(&it, RzIntervalNode, node);
if (intervalnode->start != start) {
break;
}
ret = cb(intervalnode, user);
if (!ret) {
break;
}
rz_rbtree_iter_next(&it);
}
return ret;
}
RZ_API bool rz_interval_node_all_in(RzIntervalNode *node, ut64 value, bool end_inclusive, RzIntervalIterCb cb, void *user) {
while (node && value < node->start) {
// less than the current node, but might still be contained further down
node = unwrap(node->node.child[0]);
}
if (!node) {
return true;
}
if (end_inclusive ? value > node->max_end : value >= node->max_end) {
return true;
}
if (end_inclusive ? value <= node->end : value < node->end) {
if (!cb(node, user)) {
return false;
}
}
// This can be done more efficiently by building the stack manually
bool ret = rz_interval_node_all_in(unwrap(node->node.child[0]), value, end_inclusive, cb, user);
if (!ret) {
return false;
}
return rz_interval_node_all_in(unwrap(node->node.child[1]), value, end_inclusive, cb, user);
}
RZ_API bool rz_interval_tree_all_in(RzIntervalTree *tree, ut64 value, bool end_inclusive, RzIntervalIterCb cb, void *user) {
// all in! 🂡
return rz_interval_node_all_in(tree->root, value, end_inclusive, cb, user);
}
static bool rz_interval_node_all_intersect(RzIntervalNode *node, ut64 start, ut64 end, bool end_inclusive, RzIntervalIterCb cb, void *user) {
rz_return_val_if_fail(end >= start, true);
while (node && (end_inclusive ? end < node->start : end <= node->start)) {
// less than the current node, but might still be contained further down
node = unwrap(node->node.child[0]);
}
if (!node) {
return true;
}
if (end_inclusive ? start > node->max_end : start >= node->max_end) {
return true;
}
if (end_inclusive ? start <= node->end : start < node->end) {
if (!cb(node, user)) {
return false;
}
}
// This can be done more efficiently by building the stack manually
if (!rz_interval_node_all_intersect(unwrap(node->node.child[0]), start, end, end_inclusive, cb, user)) {
return false;
}
return rz_interval_node_all_intersect(unwrap(node->node.child[1]), start, end, end_inclusive, cb, user);
}
RZ_API bool rz_interval_tree_all_intersect(RzIntervalTree *tree, ut64 start, ut64 end, bool end_inclusive, RzIntervalIterCb cb, void *user) {
return rz_interval_node_all_intersect(tree->root, start, end, end_inclusive, cb, user);
}