> On Mar 9, 2016, at 9:55 AM, Emmanuel Lécharny <[email protected]> wrote:
>
> The real question is : do we have to protect #7 at all ? graph is not a
> global variable, so there is no concurrent access issue here. Worse case
> scenario : two threads call loadGraph() with the same ContextID, but
> here, that mean we should check the conncury in loadgraph, not in
> getGraph().
So let’s dissect that function:
private static SimpleDirectedGraph<String, Relationship> loadGraph( String
contextId )
{
Hier inHier = new Hier( Hier.Type.ROLE );
inHier.setContextId( contextId );
LOG.info( "loadGraph initializing ROLE context [{}]",
inHier.getContextId() );
List<Graphable> descendants = null;
try
{
1 descendants = roleP.getAllDescendants( inHier.getContextId() );
}
catch ( SecurityException se )
{
LOG.info( "loadGraph caught SecurityException={}", se );
}
2 Hier hier = HierUtil.loadHier( contextId, descendants );
SimpleDirectedGraph<String, Relationship> graph;
3 graph = HierUtil.buildGraph( hier );
4 roleCache.put( getKey( contextId ), graph );
return graph;
}
The entire method needs to be synchronized as 1 - 4 need to be performed as a
unit of work.
>
> On Mar 9, 2016, at 9:55 AM, Emmanuel Lécharny <[email protected]> wrote:
>
>
> That of course assume roleCache itself to be thread-safe.
Yes
