In 4.2 during object_store refactoring, we have removed that assumption.
Thanks
-min
On 7/8/13 7:44 PM, "Kelven Yang" <kelven.y...@citrix.com> wrote:
>When I ran a build from latest 4.1 branch, I found an issue and I'm
>wondering why it is not breaking 4.1.1 builds from others, here is some
>of details
>
>
>public class ArrayTypeAdaptor<T> implements JsonDeserializer<T[]>,
>JsonSerializer<T[]> {
>
>
> protected Gson _gson = null;
>
>
>
> private static final String s_pkg =
>Command.class.getPackage().getName() + ".";
>
>
> public ArrayTypeAdaptor() {
>
> }
>
>
> public void initGson(Gson gson) {
>
> _gson = gson;
>
> }
>
>
> @Override
>
> public JsonElement serialize(T[] src, Type typeOfSrc,
>JsonSerializationContext context) {
>
> JsonArray array = new JsonArray();
>
> for (T cmd : src) {
>
> JsonObject obj = new JsonObject();
>
> obj.add(cmd.getClass().getName().substring(s_pkg.length()),
>_gson.toJsonTree(cmd));
>
> array.add(obj);
>
> }
>
>
> return array;
>
> }
>
>
> @Override
>
> @SuppressWarnings("unchecked")
>
> public T[] deserialize(JsonElement json, Type typeOfT,
>JsonDeserializationContext context)
>
> throws JsonParseException {
>
> JsonArray array = json.getAsJsonArray();
>
> Iterator<JsonElement> it = array.iterator();
>
> ArrayList<T> cmds = new ArrayList<T>();
>
> while (it.hasNext()) {
>
> JsonObject element = (JsonObject)it.next();
>
> Map.Entry<String, JsonElement> entry =
>element.entrySet().iterator().next();
>
>
> String name = s_pkg + entry.getKey();
>
> Class<?> clazz;
>
> try {
>
> clazz = Class.forName(name);
>
> } catch (ClassNotFoundException e) {
>
> throw new CloudRuntimeException("can't find " + name);
>
> }
>
> T cmd = (T)_gson.fromJson(entry.getValue(), clazz);
>
> cmds.add(cmd);
>
> }
>
> Class<?> type = ((Class<?>)typeOfT).getComponentType();
>
> T[] ts = (T[])Array.newInstance(type, cmds.size());
>
> return cmds.toArray(ts);
>
> }
>
>}
>
>
>Inside ArrayTypeAdapter, it retrieves the package info from where
>ArrayTypeAdaatper.class locates at run time. and it assumes that all
>array element object classes are also from the same java package (up to
>the package detail path above the class name). Apparently we have
>relocated many of classes, for example, ArrayTypeAdatper is currently at
>com.cloud.agent.transport and agent cmd classes are mostly put under
>com.cloud.agent.api now.
>
>
>I'm going to remove this hard-coded assumption about the package location
>of ArrayTypeAdatper class and related agent command classes. Anyone has a
>different idea?
>
>
>Kelven
>
>
