A GuardedDataset allows us to handle multithreaded stateful processing.
In a multi-threaded environment such as a Servlet, multiple requests for the
same dataset may happen at the same time. This is a problem only when there is
some state that is being maintained. Caching strategies (eg the netcdf server
caches open netcdf files) may need to maintain state information in order to manage
resources efficiently.
All accesses to the DDS and DAS are made through the GuardedDataset.
Typically the server puts a mutex lock on the resource when getDDS() or getDAS()
is called. When the dataset processing is complete, release() is called, and
the server releases the mutex.
Example use:
public void doGetDAS(HttpServletRequest request,
HttpServletResponse response,
ReqState rs)
throws IOException, ServletException {
response.setContentType("text/plain");
response.setHeader("XDODS-Server", getServerVersion() );
response.setHeader("Content-Description", "dods-dds");
OutputStream Out = new BufferedOutputStream(response.getOutputStream());
GuardedDataset ds = null;
try {
ds = getDataset(rs);
DAS myDAS = ds.getDAS(); // server would lock here
myDAS.print(Out);
response.setStatus(response.SC_OK);
}
catch (DAP2Exception de){
dap2ExceptionHandler(de,response);
}
catch (ParseException pe) {
parseExceptionHandler(pe,response);
}
finally { // release lock if needed
if (ds != null) ds.release();
}
}
Its important that the DDS or DAS not be used after release() is called.
See opendap.servers.netcdf.NcDataset for example of implementing a locking
GuardedDataset.
If a server is not keeping state, it can simply pass the DDS and DAS without locking,
and implement a dummy release() method.