Backport More efficiently load child beans for HippoItem

Currently, there is code that looks like:

public <T extends HippoBean> List<T> getLinkedBeans(String relPath, Class<T> beanMappingClass) {

// snip
NodeIterator nodes = relNode.getNodes();
while (nodes.hasNext()) {
Node child = nodes.nextNode();
if (child == null)

Object bean;
try {
bean = this.objectConverter.getObject(child);
if (child.getName().equals(nodeName)) {

this is horribly inefficient as relNode.getNodes() could already check the node name pattern.