src/base/AsyncJobs.dox

   1 /**
   2 \defgroup AsyncJobs Asynchronous Jobs
   3 \ingroup Components
   4
   5 \section Terminology Terminology
   6
   7 - \b Job: an AsyncJob object.
   8 - \b Creator: the code creating the job. Usually the Initiator.
   9 - \b Start: the act of calling AsyncStart with a job pointer.
  10 - \b Initiator: the code starting the job. Usually the Creator.
  11
  12 \section Life Typical life cycle
  13
  14 -# Creator creates and initializes a job.
  15 -# Initiator starts the job. If Initiator expects
  16 to communicate with the started job, then it stores the job pointer
  17 returned by AsyncStart.
  18 -# The job's start() method is called. The method usually schedules
  19 some I/O or registers to receive some other callbacks.
  20 -# The job runs and does what it is supposed to do. This usually involves
  21 scheduling I/O, setting timeouts, receiving Comm or Store callbacks, and
  22 then notifying Initiator of the final result.
  23 -# The job reaches its goal or encounters an error condition.
  24 -# The swanSong() method is called.
  25 -# The job object is destroyed.
  26
  27
  28 If you want to do something before starting the job, do it in the constructor
  29 or some custom method that the job creator will call _before_ calling
  30 AsyncStart():
  31
  32     std::auto_ptr<MyJob> job(new MyJob(...)); // sync/blocking
  33     job->prepare(...); // sync/blocking
  34     job->prepareSomethingElse(...); // sync/blocking
  35     AsyncStart(job.release()); // non-blocking
  36
  37 If you do not need complex preparations, it is better to do this instead:
  38
  39     AsyncStart(new MyJob(...));
  40
  41 Keep in mind that you have no async debugging, cleanup, and protections until
  42 you call AsyncStart with a job pointer.
  43
  44
  45 \section Rules Basic rules
  46
  47 - To start a job, use AsyncStart. Do not start the same job more than once.
  48
  49 - Never call start() directly. Treat this method as main() in C/C++.
  50
  51 - Never call swanSong() directly. If you are outside an AsyncCall
  52 handler, and want to kill the job, then call deleteThis(). If you are
  53 inside an AsyncCall handler, you have several options for job termination:
  54
  55   -# Call mustStop(reason) for errors that require further processing in
  56 the same method(s) chain, below/after the mustStop() call. Efficient.
  57
  58   -# Throw (via Must or directly) for errors that do not require further
  59 processing in the same method(s) chain, below/after the mustStop() call.
  60 Inefficient but simple and allows exiting from deeply nested method calls.
  61
  62   -# Otherwise, just finish the call. Your doneAll() should return true
  63 and the job will terminate successfully.
  64
  65 swanSong() will be called automatically in all of these cases when the
  66 job is being terminated. It is a general cleanup method, like a
  67 destructor. The only difference is that a destructor must not throw.
  68
  69
  70 - Do not assume swanSong() is called in some perfectly nice job state.
  71 The job code or the code it calls may throw at any time after start()
  72 was called. The entry may be gone, the Abort may have been called, the
  73 fd may have been closed, etc.
  74
  75
  76 - Never call deleteThis() in contexts other than those documented above.
  77 It is a hack for the old-style code. You can avoid it and other
  78 old-style special precautions altogether if you convert sync calls into
  79 async ones. This is especially easy for old-style calls that have only
  80 one parameter ("data") or two simple parameters.
  81
  82
  83 - In swanSong, always call swanSong() of the parent, after you are done
  84 cleaning up your job. It does not matter whether the [current] parent
  85 swanSong() does nothing.
  86
  87 - You must implement start() and doneAll() methods. These methods may be
  88 marked as pure virtual in future releases.
  89
  90 - In doneAll(), always call doneAll() of the parent. If the parent is not
  91 done, you are not done. It does not matter whether the [current] parent
  92 doneAll() always returns true.
  93
  94 - If a job does not have a doneAll() method implemented, it is probably
  95 buggy. Any job must know what it wants to accomplish. Please note that
  96 doneAll() is for defining the successful termination goal/condition.
  97 Errors are handled by mustStop() or throw, as discussed above.
  98
  99 Similarly, if your doneAll() implementation looks like "return isDone;",
 100 you are doing it wrong. Compute the condition directly rather than
 101 expecting various job methods to maintain some isDone variable correctly.
 102
 103
 104 - If a job does Comm I/O, it probably needs a Comm closing handler.
 105
 106 - If a job stores a StoreEntry, it probably needs an entry Abort handler.
 107
 108 - Ask yourself what the user will see/experience when the job throws,
 109 which could happen as early as in the start() method (technically, it
 110 can happen even earlier, during job creation and initialization). Are you
 111 OK with that?
 112
 113
 114 */