summaryrefslogtreecommitdiffstats
path: root/src/registry_url.c
diff options
context:
space:
mode:
Diffstat (limited to '')
-rw-r--r--src/registry_url.c85
1 files changed, 85 insertions, 0 deletions
diff --git a/src/registry_url.c b/src/registry_url.c
new file mode 100644
index 00000000..52d36a89
--- /dev/null
+++ b/src/registry_url.c
@@ -0,0 +1,85 @@
+#include "registry_internals.h"
+
+// ----------------------------------------------------------------------------
+// REGISTRY_URL
+
+int registry_url_compare(void *a, void *b) {
+ if(((REGISTRY_URL *)a)->hash < ((REGISTRY_URL *)b)->hash) return -1;
+ else if(((REGISTRY_URL *)a)->hash > ((REGISTRY_URL *)b)->hash) return 1;
+ else return strcmp(((REGISTRY_URL *)a)->url, ((REGISTRY_URL *)b)->url);
+}
+
+inline REGISTRY_URL *registry_url_index_add(REGISTRY_URL *u) {
+ return (REGISTRY_URL *)avl_insert(&(registry.registry_urls_root_index), (avl *)(u));
+}
+
+inline REGISTRY_URL *registry_url_index_del(REGISTRY_URL *u) {
+ return (REGISTRY_URL *)avl_remove(&(registry.registry_urls_root_index), (avl *)(u));
+}
+
+REGISTRY_URL *registry_url_get(const char *url, size_t urllen) {
+ // protection from too big URLs
+ if(urllen > registry.max_url_length)
+ urllen = registry.max_url_length;
+
+ debug(D_REGISTRY, "Registry: registry_url_get('%s', %zu)", url, urllen);
+
+ char buf[sizeof(REGISTRY_URL) + urllen]; // no need for +1, 1 is already in REGISTRY_URL
+ REGISTRY_URL *n = (REGISTRY_URL *)&buf[0];
+ n->len = (uint16_t)urllen;
+ strncpyz(n->url, url, n->len);
+ n->hash = simple_hash(n->url);
+
+ REGISTRY_URL *u = (REGISTRY_URL *)avl_search(&(registry.registry_urls_root_index), (avl *)n);
+ if(!u) {
+ debug(D_REGISTRY, "Registry: registry_url_get('%s', %zu): allocating %zu bytes", url, urllen, sizeof(REGISTRY_URL) + urllen);
+ u = callocz(1, sizeof(REGISTRY_URL) + urllen); // no need for +1, 1 is already in REGISTRY_URL
+
+ // a simple strcpy() should do the job
+ // but I prefer to be safe, since the caller specified urllen
+ u->len = (uint16_t)urllen;
+ strncpyz(u->url, url, u->len);
+ u->links = 0;
+ u->hash = simple_hash(u->url);
+
+ registry.urls_memory += sizeof(REGISTRY_URL) + urllen; // no need for +1, 1 is already in REGISTRY_URL
+
+ debug(D_REGISTRY, "Registry: registry_url_get('%s'): indexing it", url);
+ n = registry_url_index_add(u);
+ if(n != u) {
+ error("INTERNAL ERROR: registry_url_get(): url '%s' already exists in the registry as '%s'", u->url, n->url);
+ free(u);
+ u = n;
+ }
+ else
+ registry.urls_count++;
+ }
+
+ return u;
+}
+
+void registry_url_link(REGISTRY_URL *u) {
+ u->links++;
+ debug(D_REGISTRY, "Registry: registry_url_link('%s'): URL has now %u links", u->url, u->links);
+}
+
+void registry_url_unlink(REGISTRY_URL *u) {
+ u->links--;
+ if(!u->links) {
+ debug(D_REGISTRY, "Registry: registry_url_unlink('%s'): No more links for this URL", u->url);
+ REGISTRY_URL *n = registry_url_index_del(u);
+ if(!n) {
+ error("INTERNAL ERROR: registry_url_unlink('%s'): cannot find url in index", u->url);
+ }
+ else {
+ if(n != u) {
+ error("INTERNAL ERROR: registry_url_unlink('%s'): deleted different url '%s'", u->url, n->url);
+ }
+
+ registry.urls_memory -= sizeof(REGISTRY_URL) + n->len; // no need for +1, 1 is already in REGISTRY_URL
+ freez(n);
+ }
+ }
+ else
+ debug(D_REGISTRY, "Registry: registry_url_unlink('%s'): URL has %u links left", u->url, u->links);
+}