remove remaining references to deprecated XREP/XREQ names

only one line of actual code is changed.

these names have been removed from the next libzmq release,
and will presumably be removed from pyzmq someday.

IPython/frontend/consoleapp.py

@@ -1,356 +1,356 b''
 """ A minimal application base mixin for all ZMQ based IPython frontends.
 
 This is not a complete console app, as subprocess will not be able to receive
 input, there is no real readline support, among other limitations. This is a
 refactoring of what used to be the IPython/frontend/qt/console/qtconsoleapp.py
 
 Authors:
 
 * Evan Patterson
 * Min RK
 * Erik Tollerud
 * Fernando Perez
 * Bussonnier Matthias
 * Thomas Kluyver
 * Paul Ivanov
 
 """
 
 #-----------------------------------------------------------------------------
 # Imports
 #-----------------------------------------------------------------------------
 
 # stdlib imports
 import atexit
 import json
 import os
 import signal
 import sys
 import uuid
 
 
 # Local imports
 from IPython.config.application import boolean_flag
 from IPython.config.configurable import Configurable
 from IPython.core.profiledir import ProfileDir
 from IPython.lib.kernel import tunnel_to_kernel, find_connection_file, swallow_argv
 from IPython.zmq.blockingkernelmanager import BlockingKernelManager
 from IPython.utils.path import filefind
 from IPython.utils.py3compat import str_to_bytes
 from IPython.utils.traitlets import (
     Dict, List, Unicode, CUnicode, Int, CBool, Any
 )
 from IPython.zmq.ipkernel import (
     flags as ipkernel_flags,
     aliases as ipkernel_aliases,
     IPKernelApp
 )
 from IPython.zmq.session import Session, default_secure
 from IPython.zmq.zmqshell import ZMQInteractiveShell
 
 #-----------------------------------------------------------------------------
 # Network Constants
 #-----------------------------------------------------------------------------
 
 from IPython.utils.localinterfaces import LOCALHOST, LOCAL_IPS
 
 #-----------------------------------------------------------------------------
 # Globals
 #-----------------------------------------------------------------------------
 
 
 #-----------------------------------------------------------------------------
 # Aliases and Flags
 #-----------------------------------------------------------------------------
 
 flags = dict(ipkernel_flags)
 
 # the flags that are specific to the frontend
 # these must be scrubbed before being passed to the kernel,
 # or it will raise an error on unrecognized flags
 app_flags = {
     'existing' : ({'IPythonConsoleApp' : {'existing' : 'kernel*.json'}},
             "Connect to an existing kernel. If no argument specified, guess most recent"),
 }
 app_flags.update(boolean_flag(
     'confirm-exit', 'IPythonConsoleApp.confirm_exit',
     """Set to display confirmation dialog on exit. You can always use 'exit' or 'quit',
        to force a direct exit without any confirmation.
     """,
     """Don't prompt the user when exiting. This will terminate the kernel
        if it is owned by the frontend, and leave it alive if it is external.
     """
 ))
 flags.update(app_flags)
 
 aliases = dict(ipkernel_aliases)
 
 # also scrub aliases from the frontend
 app_aliases = dict(
     hb = 'IPythonConsoleApp.hb_port',
     shell = 'IPythonConsoleApp.shell_port',
     iopub = 'IPythonConsoleApp.iopub_port',
     stdin = 'IPythonConsoleApp.stdin_port',
     ip = 'IPythonConsoleApp.ip',
     existing = 'IPythonConsoleApp.existing',
     f = 'IPythonConsoleApp.connection_file',
 
 
     ssh = 'IPythonConsoleApp.sshserver',
 )
 aliases.update(app_aliases)
 
 #-----------------------------------------------------------------------------
 # Classes
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # IPythonConsole
 #-----------------------------------------------------------------------------
 
 classes = [IPKernelApp, ZMQInteractiveShell, ProfileDir, Session]
 
 try:
     from IPython.zmq.pylab.backend_inline import InlineBackend
 except ImportError:
     pass
 else:
     classes.append(InlineBackend)
 
 class IPythonConsoleApp(Configurable):
     name = 'ipython-console-mixin'
     default_config_file_name='ipython_config.py'
 
     description = """
         The IPython Mixin Console.
         
         This class contains the common portions of console client (QtConsole,
         ZMQ-based terminal console, etc).  It is not a full console, in that
         launched terminal subprocesses will not be able to accept input.
         
         The Console using this mixing supports various extra features beyond
         the single-process Terminal IPython shell, such as connecting to
         existing kernel, via:
         
             ipython <appname> --existing
         
         as well as tunnel via SSH
         
     """
 
     classes = classes
     flags = Dict(flags)
     aliases = Dict(aliases)
     kernel_manager_class = BlockingKernelManager
 
     kernel_argv = List(Unicode)
     # frontend flags&aliases to be stripped when building kernel_argv
     frontend_flags = Any(app_flags)
     frontend_aliases = Any(app_aliases)
 
     # create requested profiles by default, if they don't exist:
     auto_create = CBool(True)
     # connection info:
     ip = Unicode(LOCALHOST, config=True,
         help="""Set the kernel\'s IP address [default localhost].
         If the IP address is something other than localhost, then
         Consoles on other machines will be able to connect
         to the Kernel, so be careful!"""
     )
     
     sshserver = Unicode('', config=True,
         help="""The SSH server to use to connect to the kernel.""")
     sshkey = Unicode('', config=True,
         help="""Path to the ssh key to use for logging in to the ssh server.""")
     
     hb_port = Int(0, config=True,
         help="set the heartbeat port [default: random]")
     shell_port = Int(0, config=True,
-        help="set the shell (XREP) port [default: random]")
+        help="set the shell (ROUTER) port [default: random]")
     iopub_port = Int(0, config=True,
         help="set the iopub (PUB) port [default: random]")
     stdin_port = Int(0, config=True,
-        help="set the stdin (XREQ) port [default: random]")
+        help="set the stdin (DEALER) port [default: random]")
     connection_file = Unicode('', config=True,
         help="""JSON file in which to store connection info [default: kernel-<pid>.json]
 
         This file will contain the IP, ports, and authentication key needed to connect
         clients to this kernel. By default, this file will be created in the security-dir
         of the current profile, but can be specified by absolute path.
         """)
     def _connection_file_default(self):
         return 'kernel-%i.json' % os.getpid()
 
     existing = CUnicode('', config=True,
         help="""Connect to an already running kernel""")
 
     confirm_exit = CBool(True, config=True,
         help="""
         Set to display confirmation dialog on exit. You can always use 'exit' or 'quit',
         to force a direct exit without any confirmation.""",
     )
 
 
     def build_kernel_argv(self, argv=None):
         """build argv to be passed to kernel subprocess"""
         if argv is None:
             argv = sys.argv[1:]
         self.kernel_argv = swallow_argv(argv, self.frontend_aliases, self.frontend_flags)
         # kernel should inherit default config file from frontend
         self.kernel_argv.append("--KernelApp.parent_appname='%s'"%self.name)
     
     def init_connection_file(self):
         """find the connection file, and load the info if found.
         
         The current working directory and the current profile's security
         directory will be searched for the file if it is not given by
         absolute path.
         
         When attempting to connect to an existing kernel and the `--existing`
         argument does not match an existing file, it will be interpreted as a
         fileglob, and the matching file in the current profile's security dir
         with the latest access time will be used.
         
         After this method is called, self.connection_file contains the *full path*
         to the connection file, never just its name.
         """
         if self.existing:
             try:
                 cf = find_connection_file(self.existing)
             except Exception:
                 self.log.critical("Could not find existing kernel connection file %s", self.existing)
                 self.exit(1)
             self.log.info("Connecting to existing kernel: %s" % cf)
             self.connection_file = cf
         else:
             # not existing, check if we are going to write the file
             # and ensure that self.connection_file is a full path, not just the shortname
             try:
                 cf = find_connection_file(self.connection_file)
             except Exception:
                 # file might not exist
                 if self.connection_file == os.path.basename(self.connection_file):
                     # just shortname, put it in security dir
                     cf = os.path.join(self.profile_dir.security_dir, self.connection_file)
                 else:
                     cf = self.connection_file
                 self.connection_file = cf
         
         # should load_connection_file only be used for existing?
         # as it is now, this allows reusing ports if an existing
         # file is requested
         try:
             self.load_connection_file()
         except Exception:
             self.log.error("Failed to load connection file: %r", self.connection_file, exc_info=True)
             self.exit(1)
     
     def load_connection_file(self):
         """load ip/port/hmac config from JSON connection file"""
         # this is identical to KernelApp.load_connection_file
         # perhaps it can be centralized somewhere?
         try:
             fname = filefind(self.connection_file, ['.', self.profile_dir.security_dir])
         except IOError:
             self.log.debug("Connection File not found: %s", self.connection_file)
             return
         self.log.debug(u"Loading connection file %s", fname)
         with open(fname) as f:
             s = f.read()
         cfg = json.loads(s)
         if self.ip == LOCALHOST and 'ip' in cfg:
             # not overridden by config or cl_args
             self.ip = cfg['ip']
         for channel in ('hb', 'shell', 'iopub', 'stdin'):
             name = channel + '_port'
             if getattr(self, name) == 0 and name in cfg:
                 # not overridden by config or cl_args
                 setattr(self, name, cfg[name])
         if 'key' in cfg:
             self.config.Session.key = str_to_bytes(cfg['key'])
     
     def init_ssh(self):
         """set up ssh tunnels, if needed."""
         if not self.sshserver and not self.sshkey:
             return
         
         if self.sshkey and not self.sshserver:
             # specifying just the key implies that we are connecting directly
             self.sshserver = self.ip
             self.ip = LOCALHOST
         
         # build connection dict for tunnels:
         info = dict(ip=self.ip,
                     shell_port=self.shell_port,
                     iopub_port=self.iopub_port,
                     stdin_port=self.stdin_port,
                     hb_port=self.hb_port
         )
         
         self.log.info("Forwarding connections to %s via %s"%(self.ip, self.sshserver))
         
         # tunnels return a new set of ports, which will be on localhost:
         self.ip = LOCALHOST
         try:
             newports = tunnel_to_kernel(info, self.sshserver, self.sshkey)
         except:
             # even catch KeyboardInterrupt
             self.log.error("Could not setup tunnels", exc_info=True)
             self.exit(1)
         
         self.shell_port, self.iopub_port, self.stdin_port, self.hb_port = newports
         
         cf = self.connection_file
         base,ext = os.path.splitext(cf)
         base = os.path.basename(base)
         self.connection_file = os.path.basename(base)+'-ssh'+ext
         self.log.critical("To connect another client via this tunnel, use:")
         self.log.critical("--existing %s" % self.connection_file)
     
     def _new_connection_file(self):
         cf = ''
         while not cf:
             # we don't need a 128b id to distinguish kernels, use more readable
             # 48b node segment (12 hex chars).  Users running more than 32k simultaneous
             # kernels can subclass.
             ident = str(uuid.uuid4()).split('-')[-1]
             cf = os.path.join(self.profile_dir.security_dir, 'kernel-%s.json' % ident)
             # only keep if it's actually new.  Protect against unlikely collision
             # in 48b random search space
             cf = cf if not os.path.exists(cf) else ''
         return cf
 
     def init_kernel_manager(self):
         # Don't let Qt or ZMQ swallow KeyboardInterupts.
         signal.signal(signal.SIGINT, signal.SIG_DFL)
 
         # Create a KernelManager and start a kernel.
         self.kernel_manager = self.kernel_manager_class(
                                 ip=self.ip,
                                 shell_port=self.shell_port,
                                 iopub_port=self.iopub_port,
                                 stdin_port=self.stdin_port,
                                 hb_port=self.hb_port,
                                 connection_file=self.connection_file,
                                 config=self.config,
         )
         # start the kernel
         if not self.existing:
             self.kernel_manager.start_kernel(extra_arguments=self.kernel_argv)
         elif self.sshserver:
             # ssh, write new connection file
             self.kernel_manager.write_connection_file()
         atexit.register(self.kernel_manager.cleanup_connection_file)
         self.kernel_manager.start_channels()
 
 
     def initialize(self, argv=None):
         """
         Classes which mix this class in should call:
                IPythonConsoleApp.initialize(self,argv)
         """
         self.init_connection_file()
         default_secure(self.config)
         self.init_ssh()
         self.init_kernel_manager()
 

IPython/frontend/html/notebook/kernelmanager.py

@@ -1,324 +1,324 b''
 """A kernel manager for multiple kernels.
 
 Authors:
 
 * Brian Granger
 """
 
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2008-2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Imports
 #-----------------------------------------------------------------------------
 
 import os
 import signal
 import sys
 import uuid
 
 import zmq
 from zmq.eventloop.zmqstream import ZMQStream
 
 from tornado import web
 
 from IPython.config.configurable import LoggingConfigurable
 from IPython.utils.importstring import import_item
 from IPython.utils.traitlets import (
     Instance, Dict, List, Unicode, Float, Integer, Any, DottedObjectName,
 )
 #-----------------------------------------------------------------------------
 # Classes
 #-----------------------------------------------------------------------------
 
 class DuplicateKernelError(Exception):
     pass
 
 
 class MultiKernelManager(LoggingConfigurable):
     """A class for managing multiple kernels."""
     
     kernel_manager_class = DottedObjectName(
         "IPython.zmq.kernelmanager.KernelManager", config=True,
         help="""The kernel manager class.  This is configurable to allow
         subclassing of the KernelManager for customized behavior.
         """
     )
     def _kernel_manager_class_changed(self, name, old, new):
         self.kernel_manager_factory = import_item(new)
     
     kernel_manager_factory = Any(help="this is kernel_manager_class after import")
     def _kernel_manager_factory_default(self):
         return import_item(self.kernel_manager_class)
     
     context = Instance('zmq.Context')
     def _context_default(self):
         return zmq.Context.instance()
     
     connection_dir = Unicode('')
 
     _kernels = Dict()
 
     @property
     def kernel_ids(self):
         """Return a list of the kernel ids of the active kernels."""
         return self._kernels.keys()
 
     def __len__(self):
         """Return the number of running kernels."""
         return len(self.kernel_ids)
 
     def __contains__(self, kernel_id):
         if kernel_id in self.kernel_ids:
             return True
         else:
             return False
 
     def start_kernel(self, **kwargs):
         """Start a new kernel."""
         kernel_id = unicode(uuid.uuid4())
         # use base KernelManager for each Kernel
         km = self.kernel_manager_factory(connection_file=os.path.join(
                     self.connection_dir, "kernel-%s.json" % kernel_id),
                     config=self.config,
         )
         km.start_kernel(**kwargs)
         self._kernels[kernel_id] = km
         return kernel_id
 
     def kill_kernel(self, kernel_id):
         """Kill a kernel by its kernel uuid.
 
         Parameters
         ==========
         kernel_id : uuid
             The id of the kernel to kill.
         """
         self.get_kernel(kernel_id).kill_kernel()
         del self._kernels[kernel_id]
 
     def interrupt_kernel(self, kernel_id):
         """Interrupt (SIGINT) the kernel by its uuid.
 
         Parameters
         ==========
         kernel_id : uuid
             The id of the kernel to interrupt.
         """
         return self.get_kernel(kernel_id).interrupt_kernel()
 
     def signal_kernel(self, kernel_id, signum):
         """ Sends a signal to the kernel by its uuid.
 
         Note that since only SIGTERM is supported on Windows, this function
         is only useful on Unix systems.
 
         Parameters
         ==========
         kernel_id : uuid
             The id of the kernel to signal.
         """
         return self.get_kernel(kernel_id).signal_kernel(signum)
 
     def get_kernel(self, kernel_id):
         """Get the single KernelManager object for a kernel by its uuid.
 
         Parameters
         ==========
         kernel_id : uuid
             The id of the kernel.
         """
         km = self._kernels.get(kernel_id)
         if km is not None:
             return km
         else:
             raise KeyError("Kernel with id not found: %s" % kernel_id)
 
     def get_kernel_ports(self, kernel_id):
         """Return a dictionary of ports for a kernel.
 
         Parameters
         ==========
         kernel_id : uuid
             The id of the kernel.
 
         Returns
         =======
         port_dict : dict
             A dict of key, value pairs where the keys are the names
             (stdin_port,iopub_port,shell_port) and the values are the
             integer port numbers for those channels.
         """
         # this will raise a KeyError if not found:
         km = self.get_kernel(kernel_id)
         return dict(shell_port=km.shell_port,
                     iopub_port=km.iopub_port,
                     stdin_port=km.stdin_port,
                     hb_port=km.hb_port,
         )
 
     def get_kernel_ip(self, kernel_id):
         """Return ip address for a kernel.
 
         Parameters
         ==========
         kernel_id : uuid
             The id of the kernel.
 
         Returns
         =======
         ip : str
             The ip address of the kernel.
         """
         return self.get_kernel(kernel_id).ip
 
     def create_connected_stream(self, ip, port, socket_type):
         sock = self.context.socket(socket_type)
         addr = "tcp://%s:%i" % (ip, port)
         self.log.info("Connecting to: %s" % addr)
         sock.connect(addr)
         return ZMQStream(sock)
 
     def create_iopub_stream(self, kernel_id):
         ip = self.get_kernel_ip(kernel_id)
         ports = self.get_kernel_ports(kernel_id)
         iopub_stream = self.create_connected_stream(ip, ports['iopub_port'], zmq.SUB)
         iopub_stream.socket.setsockopt(zmq.SUBSCRIBE, b'')
         return iopub_stream
 
     def create_shell_stream(self, kernel_id):
         ip = self.get_kernel_ip(kernel_id)
         ports = self.get_kernel_ports(kernel_id)
-        shell_stream = self.create_connected_stream(ip, ports['shell_port'], zmq.XREQ)
+        shell_stream = self.create_connected_stream(ip, ports['shell_port'], zmq.DEALER)
         return shell_stream
 
     def create_hb_stream(self, kernel_id):
         ip = self.get_kernel_ip(kernel_id)
         ports = self.get_kernel_ports(kernel_id)
         hb_stream = self.create_connected_stream(ip, ports['hb_port'], zmq.REQ)
         return hb_stream
 
 
 class MappingKernelManager(MultiKernelManager):
     """A KernelManager that handles notebok mapping and HTTP error handling"""
 
     kernel_argv = List(Unicode)
     
     time_to_dead = Float(3.0, config=True, help="""Kernel heartbeat interval in seconds.""")
     first_beat = Float(5.0, config=True, help="Delay (in seconds) before sending first heartbeat.")
     
     max_msg_size = Integer(65536, config=True, help="""
         The max raw message size accepted from the browser
         over a WebSocket connection.
     """)
 
     _notebook_mapping = Dict()
 
     #-------------------------------------------------------------------------
     # Methods for managing kernels and sessions
     #-------------------------------------------------------------------------
 
     def kernel_for_notebook(self, notebook_id):
         """Return the kernel_id for a notebook_id or None."""
         return self._notebook_mapping.get(notebook_id)
 
     def set_kernel_for_notebook(self, notebook_id, kernel_id):
         """Associate a notebook with a kernel."""
         if notebook_id is not None:
             self._notebook_mapping[notebook_id] = kernel_id
 
     def notebook_for_kernel(self, kernel_id):
         """Return the notebook_id for a kernel_id or None."""
         notebook_ids = [k for k, v in self._notebook_mapping.iteritems() if v == kernel_id]
         if len(notebook_ids) == 1:
             return notebook_ids[0]
         else:
             return None
 
     def delete_mapping_for_kernel(self, kernel_id):
         """Remove the kernel/notebook mapping for kernel_id."""
         notebook_id = self.notebook_for_kernel(kernel_id)
         if notebook_id is not None:
             del self._notebook_mapping[notebook_id]
 
     def start_kernel(self, notebook_id=None):
         """Start a kernel for a notebok an return its kernel_id.
 
         Parameters
         ----------
         notebook_id : uuid
             The uuid of the notebook to associate the new kernel with. If this
             is not None, this kernel will be persistent whenever the notebook
             requests a kernel.
         """
         kernel_id = self.kernel_for_notebook(notebook_id)
         if kernel_id is None:
             kwargs = dict()
             kwargs['extra_arguments'] = self.kernel_argv
             kernel_id = super(MappingKernelManager, self).start_kernel(**kwargs)
             self.set_kernel_for_notebook(notebook_id, kernel_id)
             self.log.info("Kernel started: %s" % kernel_id)
             self.log.debug("Kernel args: %r" % kwargs)
         else:
             self.log.info("Using existing kernel: %s" % kernel_id)
         return kernel_id
 
     def kill_kernel(self, kernel_id):
         """Kill a kernel and remove its notebook association."""
         self._check_kernel_id(kernel_id)
         super(MappingKernelManager, self).kill_kernel(kernel_id)
         self.delete_mapping_for_kernel(kernel_id)
         self.log.info("Kernel killed: %s" % kernel_id)
 
     def interrupt_kernel(self, kernel_id):
         """Interrupt a kernel."""
         self._check_kernel_id(kernel_id)
         super(MappingKernelManager, self).interrupt_kernel(kernel_id)
         self.log.info("Kernel interrupted: %s" % kernel_id)
 
     def restart_kernel(self, kernel_id):
         """Restart a kernel while keeping clients connected."""
         self._check_kernel_id(kernel_id)
         km = self.get_kernel(kernel_id)
         km.restart_kernel(now=True)
         self.log.info("Kernel restarted: %s" % kernel_id)
         return kernel_id
         
         # the following remains, in case the KM restart machinery is
         # somehow unacceptable
         # Get the notebook_id to preserve the kernel/notebook association.
         notebook_id = self.notebook_for_kernel(kernel_id)
         # Create the new kernel first so we can move the clients over.
         new_kernel_id = self.start_kernel()
         # Now kill the old kernel.
         self.kill_kernel(kernel_id)
         # Now save the new kernel/notebook association. We have to save it
         # after the old kernel is killed as that will delete the mapping.
         self.set_kernel_for_notebook(notebook_id, new_kernel_id)
         self.log.info("Kernel restarted: %s" % new_kernel_id)
         return new_kernel_id
 
     def create_iopub_stream(self, kernel_id):
         """Create a new iopub stream."""
         self._check_kernel_id(kernel_id)
         return super(MappingKernelManager, self).create_iopub_stream(kernel_id)
 
     def create_shell_stream(self, kernel_id):
         """Create a new shell stream."""
         self._check_kernel_id(kernel_id)
         return super(MappingKernelManager, self).create_shell_stream(kernel_id)
 
     def create_hb_stream(self, kernel_id):
         """Create a new hb stream."""
         self._check_kernel_id(kernel_id)
         return super(MappingKernelManager, self).create_hb_stream(kernel_id)
 
     def _check_kernel_id(self, kernel_id):
         """Check a that a kernel_id exists and raise 404 if not."""
         if kernel_id not in self:
             raise web.HTTPError(404, u'Kernel does not exist: %s' % kernel_id)
 

IPython/frontend/html/notebook/zmqhttp.py

@@ -1,96 +1,96 b''
 """Unfinished code for ZMQ/HTTP bridging. We use WebSockets instead.
 
 Authors:
 
 * Brian Granger
 """
 
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2008-2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Imports
 #-----------------------------------------------------------------------------
 
 import json
 import logging
 
 from tornado import web
 
 #-----------------------------------------------------------------------------
 # Code
 #-----------------------------------------------------------------------------
 
 class ZMQHandler(web.RequestHandler):
 
     def get_stream(self):
         """Get the ZMQStream for this request."""
         raise NotImplementedError('Implement get_stream() in a subclass.')
 
     def _save_method_args(self, *args, **kwargs):
         """Save the args and kwargs to get/post/put/delete for future use.
 
         These arguments are not saved in the request or handler objects, but
         are often needed by methods such as get_stream().
         """ 
         self._method_args = args
         self._method_kwargs = kwargs
 
     def _handle_msgs(self, msg):
         msgs = [msg]
         stream = self.get_stream()
         stream.on_recv(lambda m: msgs.append(json.loads(m)))
         stream.flush()
         stream.stop_on_recv()
         logging.info("Reply: %r" % msgs)
         self.write(json.dumps(msgs))
         self.finish()
 
 
 class ZMQPubHandler(ZMQHandler):
 
     SUPPORTED_METHODS = ("POST",)
 
     def post(self, *args, **kwargs):
         self._save_method_args(*args, **kwargs)
         try:
             msg = json.loads(self.request.body)
         except:
             self.send_error(status_code=415)
         else:
             logging.info("Request: %r" % msg)
             self.get_stream().send_json(msg)
 
 
 class ZMQSubHandler(ZMQHandler):
 
     SUPPORTED_METHODS = ("GET",)
 
     @web.asynchronous
     def get(self, *args, **kwargs):
         self._save_method_args(*args, **kwargs)
         self.get_stream().on_recv(self._handle_msgs)
 
 
-class ZMQXReqHandler(ZMQHandler):
+class ZMQDealerHandler(ZMQHandler):
 
     SUPPORTED_METHODS = ("POST",)
 
     @web.asynchronous
     def post(self, *args, **kwargs):
         self._save_method_args(*args, **kwargs)
         logging.info("request: %r" % self.request)
         try:
             msg = json.loads(self.request.body)
         except:
             self.send_error(status_code=415)
         else:
             logging.info("Reply: %r" % msg)
             stream = self.get_stream()
             stream.send_json(msg)
             stream.on_recv(self._handle_msgs)
 

IPython/parallel/apps/ipcontrollerapp.py

@@ -1,491 +1,491 b''
 #!/usr/bin/env python
 # encoding: utf-8
 """
 The IPython controller application.
 
 Authors:
 
 * Brian Granger
 * MinRK
 
 """
 
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2008-2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Imports
 #-----------------------------------------------------------------------------
 
 from __future__ import with_statement
 
 import json
 import os
 import socket
 import stat
 import sys
 
 from multiprocessing import Process
 from signal import signal, SIGINT, SIGABRT, SIGTERM
 
 import zmq
 from zmq.devices import ProcessMonitoredQueue
 from zmq.log.handlers import PUBHandler
 
 from IPython.core.profiledir import ProfileDir
 
 from IPython.parallel.apps.baseapp import (
     BaseParallelApplication,
     base_aliases,
     base_flags,
     catch_config_error,
 )
 from IPython.utils.importstring import import_item
 from IPython.utils.traitlets import Instance, Unicode, Bool, List, Dict, TraitError
 
 from IPython.zmq.session import (
     Session, session_aliases, session_flags, default_secure
 )
 
 from IPython.parallel.controller.heartmonitor import HeartMonitor
 from IPython.parallel.controller.hub import HubFactory
 from IPython.parallel.controller.scheduler import TaskScheduler,launch_scheduler
 from IPython.parallel.controller.sqlitedb import SQLiteDB
 
 from IPython.parallel.util import split_url, disambiguate_url
 
 # conditional import of MongoDB backend class
 
 try:
     from IPython.parallel.controller.mongodb import MongoDB
 except ImportError:
     maybe_mongo = []
 else:
     maybe_mongo = [MongoDB]
 
 
 #-----------------------------------------------------------------------------
 # Module level variables
 #-----------------------------------------------------------------------------
 
 
 #: The default config file name for this application
 default_config_file_name = u'ipcontroller_config.py'
 
 
 _description = """Start the IPython controller for parallel computing.
 
 The IPython controller provides a gateway between the IPython engines and
 clients. The controller needs to be started before the engines and can be
 configured using command line options or using a cluster directory. Cluster
 directories contain config, log and security files and are usually located in
 your ipython directory and named as "profile_name". See the `profile`
 and `profile-dir` options for details.
 """
 
 _examples = """
 ipcontroller --ip=192.168.0.1 --port=1000  # listen on ip, port for engines
 ipcontroller --scheme=pure  # use the pure zeromq scheduler
 """
 
 
 #-----------------------------------------------------------------------------
 # The main application
 #-----------------------------------------------------------------------------
 flags = {}
 flags.update(base_flags)
 flags.update({
     'usethreads' : ( {'IPControllerApp' : {'use_threads' : True}},
                     'Use threads instead of processes for the schedulers'),
     'sqlitedb' : ({'HubFactory' : {'db_class' : 'IPython.parallel.controller.sqlitedb.SQLiteDB'}},
                     'use the SQLiteDB backend'),
     'mongodb' : ({'HubFactory' : {'db_class' : 'IPython.parallel.controller.mongodb.MongoDB'}},
                     'use the MongoDB backend'),
     'dictdb' : ({'HubFactory' : {'db_class' : 'IPython.parallel.controller.dictdb.DictDB'}},
                     'use the in-memory DictDB backend'),
     'nodb' : ({'HubFactory' : {'db_class' : 'IPython.parallel.controller.dictdb.NoDB'}},
                     """use dummy DB backend, which doesn't store any information.
                     
                     This can be used to prevent growth of the memory footprint of the Hub
                     in cases where its record-keeping is not required.  Requesting results
                     of tasks submitted by other clients, db_queries, and task resubmission
                     will not be available."""),
     'reuse' : ({'IPControllerApp' : {'reuse_files' : True}},
                     'reuse existing json connection files')
 })
 
 flags.update(session_flags)
 
 aliases = dict(
     ssh = 'IPControllerApp.ssh_server',
     enginessh = 'IPControllerApp.engine_ssh_server',
     location = 'IPControllerApp.location',
 
     url = 'HubFactory.url',
     ip = 'HubFactory.ip',
     transport = 'HubFactory.transport',
     port = 'HubFactory.regport',
 
     ping = 'HeartMonitor.period',
 
     scheme = 'TaskScheduler.scheme_name',
     hwm = 'TaskScheduler.hwm',
 )
 aliases.update(base_aliases)
 aliases.update(session_aliases)
 
 
 class IPControllerApp(BaseParallelApplication):
 
     name = u'ipcontroller'
     description = _description
     examples = _examples
     config_file_name = Unicode(default_config_file_name)
     classes = [ProfileDir, Session, HubFactory, TaskScheduler, HeartMonitor, SQLiteDB] + maybe_mongo
     
     # change default to True
     auto_create = Bool(True, config=True,
         help="""Whether to create profile dir if it doesn't exist.""")
     
     reuse_files = Bool(False, config=True,
         help="""Whether to reuse existing json connection files.
         If False, connection files will be removed on a clean exit.
         """
     )
     ssh_server = Unicode(u'', config=True,
         help="""ssh url for clients to use when connecting to the Controller
         processes. It should be of the form: [user@]server[:port]. The
         Controller's listening addresses must be accessible from the ssh server""",
     )
     engine_ssh_server = Unicode(u'', config=True,
         help="""ssh url for engines to use when connecting to the Controller
         processes. It should be of the form: [user@]server[:port]. The
         Controller's listening addresses must be accessible from the ssh server""",
     )
     location = Unicode(u'', config=True,
         help="""The external IP or domain name of the Controller, used for disambiguating
         engine and client connections.""",
     )
     import_statements = List([], config=True,
         help="import statements to be run at startup.  Necessary in some environments"
     )
 
     use_threads = Bool(False, config=True,
         help='Use threads instead of processes for the schedulers',
     )
 
     engine_json_file = Unicode('ipcontroller-engine.json', config=True,
         help="JSON filename where engine connection info will be stored.")
     client_json_file = Unicode('ipcontroller-client.json', config=True,
         help="JSON filename where client connection info will be stored.")
     
     def _cluster_id_changed(self, name, old, new):
         super(IPControllerApp, self)._cluster_id_changed(name, old, new)
         self.engine_json_file = "%s-engine.json" % self.name
         self.client_json_file = "%s-client.json" % self.name
 
 
     # internal
     children = List()
     mq_class = Unicode('zmq.devices.ProcessMonitoredQueue')
 
     def _use_threads_changed(self, name, old, new):
         self.mq_class = 'zmq.devices.%sMonitoredQueue'%('Thread' if new else 'Process')
     
     write_connection_files = Bool(True,
         help="""Whether to write connection files to disk.
         True in all cases other than runs with `reuse_files=True` *after the first*
         """
     )
 
     aliases = Dict(aliases)
     flags = Dict(flags)
     
 
     def save_connection_dict(self, fname, cdict):
         """save a connection dict to json file."""
         c = self.config
         url = cdict['url']
         location = cdict['location']
         if not location:
             try:
                 proto,ip,port = split_url(url)
             except AssertionError:
                 pass
             else:
                 try:
                     location = socket.gethostbyname_ex(socket.gethostname())[2][-1]
                 except (socket.gaierror, IndexError):
                     self.log.warn("Could not identify this machine's IP, assuming 127.0.0.1."
                     " You may need to specify '--location=<external_ip_address>' to help"
                     " IPython decide when to connect via loopback.")
                     location = '127.0.0.1'
             cdict['location'] = location
         fname = os.path.join(self.profile_dir.security_dir, fname)
         self.log.info("writing connection info to %s", fname)
         with open(fname, 'w') as f:
             f.write(json.dumps(cdict, indent=2))
         os.chmod(fname, stat.S_IRUSR|stat.S_IWUSR)
     
     def load_config_from_json(self):
         """load config from existing json connector files."""
         c = self.config
         self.log.debug("loading config from JSON")
         # load from engine config
         fname = os.path.join(self.profile_dir.security_dir, self.engine_json_file)
         self.log.info("loading connection info from %s", fname)
         with open(fname) as f:
             cfg = json.loads(f.read())
         key = cfg['exec_key']
         # json gives unicode, Session.key wants bytes
         c.Session.key = key.encode('ascii')
         xport,addr = cfg['url'].split('://')
         c.HubFactory.engine_transport = xport
         ip,ports = addr.split(':')
         c.HubFactory.engine_ip = ip
         c.HubFactory.regport = int(ports)
         self.location = cfg['location']
         if not self.engine_ssh_server:
             self.engine_ssh_server = cfg['ssh']
         # load client config
         fname = os.path.join(self.profile_dir.security_dir, self.client_json_file)
         self.log.info("loading connection info from %s", fname)
         with open(fname) as f:
             cfg = json.loads(f.read())
         assert key == cfg['exec_key'], "exec_key mismatch between engine and client keys"
         xport,addr = cfg['url'].split('://')
         c.HubFactory.client_transport = xport
         ip,ports = addr.split(':')
         c.HubFactory.client_ip = ip
         if not self.ssh_server:
             self.ssh_server = cfg['ssh']
         assert int(ports) == c.HubFactory.regport, "regport mismatch"
     
     def cleanup_connection_files(self):
         if self.reuse_files:
             self.log.debug("leaving JSON connection files for reuse")
             return
         self.log.debug("cleaning up JSON connection files")
         for f in (self.client_json_file, self.engine_json_file):
             f = os.path.join(self.profile_dir.security_dir, f)
             try:
                 os.remove(f)
             except Exception as e:
                 self.log.error("Failed to cleanup connection file: %s", e)
             else:
                 self.log.debug(u"removed %s", f)
     
     def load_secondary_config(self):
         """secondary config, loading from JSON and setting defaults"""
         if self.reuse_files:
             try:
                 self.load_config_from_json()
             except (AssertionError,IOError) as e:
                 self.log.error("Could not load config from JSON: %s" % e)
             else:
                 # successfully loaded config from JSON, and reuse=True
                 # no need to wite back the same file
                 self.write_connection_files = False
                 
         # switch Session.key default to secure
         default_secure(self.config)
         self.log.debug("Config changed")
         self.log.debug(repr(self.config))
         
     def init_hub(self):
         c = self.config
         
         self.do_import_statements()
         
         try:
             self.factory = HubFactory(config=c, log=self.log)
             # self.start_logging()
             self.factory.init_hub()
         except TraitError:
             raise
         except Exception:
             self.log.error("Couldn't construct the Controller", exc_info=True)
             self.exit(1)
         
         if self.write_connection_files:
             # save to new json config files
             f = self.factory
             cdict = {'exec_key' : f.session.key.decode('ascii'),
                     'ssh' : self.ssh_server,
                     'url' : "%s://%s:%s"%(f.client_transport, f.client_ip, f.regport),
                     'location' : self.location
                     }
             self.save_connection_dict(self.client_json_file, cdict)
             edict = cdict
             edict['url']="%s://%s:%s"%((f.client_transport, f.client_ip, f.regport))
             edict['ssh'] = self.engine_ssh_server
             self.save_connection_dict(self.engine_json_file, edict)
 
     def init_schedulers(self):
         children = self.children
         mq = import_item(str(self.mq_class))
         
         hub = self.factory
         # disambiguate url, in case of *
         monitor_url = disambiguate_url(hub.monitor_url)
         # maybe_inproc = 'inproc://monitor' if self.use_threads else monitor_url
         # IOPub relay (in a Process)
         q = mq(zmq.PUB, zmq.SUB, zmq.PUB, b'N/A',b'iopub')
         q.bind_in(hub.client_info['iopub'])
         q.bind_out(hub.engine_info['iopub'])
         q.setsockopt_out(zmq.SUBSCRIBE, b'')
         q.connect_mon(monitor_url)
         q.daemon=True
         children.append(q)
 
         # Multiplexer Queue (in a Process)
         q = mq(zmq.ROUTER, zmq.ROUTER, zmq.PUB, b'in', b'out')
         q.bind_in(hub.client_info['mux'])
         q.setsockopt_in(zmq.IDENTITY, b'mux')
         q.bind_out(hub.engine_info['mux'])
         q.connect_mon(monitor_url)
         q.daemon=True
         children.append(q)
 
         # Control Queue (in a Process)
         q = mq(zmq.ROUTER, zmq.ROUTER, zmq.PUB, b'incontrol', b'outcontrol')
         q.bind_in(hub.client_info['control'])
         q.setsockopt_in(zmq.IDENTITY, b'control')
         q.bind_out(hub.engine_info['control'])
         q.connect_mon(monitor_url)
         q.daemon=True
         children.append(q)
         try:
             scheme = self.config.TaskScheduler.scheme_name
         except AttributeError:
             scheme = TaskScheduler.scheme_name.get_default_value()
         # Task Queue (in a Process)
         if scheme == 'pure':
-            self.log.warn("task::using pure XREQ Task scheduler")
+            self.log.warn("task::using pure DEALER Task scheduler")
             q = mq(zmq.ROUTER, zmq.DEALER, zmq.PUB, b'intask', b'outtask')
             # q.setsockopt_out(zmq.HWM, hub.hwm)
             q.bind_in(hub.client_info['task'][1])
             q.setsockopt_in(zmq.IDENTITY, b'task')
             q.bind_out(hub.engine_info['task'])
             q.connect_mon(monitor_url)
             q.daemon=True
             children.append(q)
         elif scheme == 'none':
             self.log.warn("task::using no Task scheduler")
 
         else:
             self.log.info("task::using Python %s Task scheduler"%scheme)
             sargs = (hub.client_info['task'][1], hub.engine_info['task'],
                                 monitor_url, disambiguate_url(hub.client_info['notification']))
             kwargs = dict(logname='scheduler', loglevel=self.log_level,
                             log_url = self.log_url, config=dict(self.config))
             if 'Process' in self.mq_class:
                 # run the Python scheduler in a Process
                 q = Process(target=launch_scheduler, args=sargs, kwargs=kwargs)
                 q.daemon=True
                 children.append(q)
             else:
                 # single-threaded Controller
                 kwargs['in_thread'] = True
                 launch_scheduler(*sargs, **kwargs)
 
     def terminate_children(self):
         child_procs = []
         for child in self.children:
             if isinstance(child, ProcessMonitoredQueue):
                 child_procs.append(child.launcher)
             elif isinstance(child, Process):
                 child_procs.append(child)
         if child_procs:
             self.log.critical("terminating children...")
             for child in child_procs:
                 try:
                     child.terminate()
                 except OSError:
                     # already dead
                     pass
     
     def handle_signal(self, sig, frame):
         self.log.critical("Received signal %i, shutting down", sig)
         self.terminate_children()
         self.loop.stop()
     
     def init_signal(self):
         for sig in (SIGINT, SIGABRT, SIGTERM):
             signal(sig, self.handle_signal)
 
     def do_import_statements(self):
         statements = self.import_statements
         for s in statements:
             try:
                 self.log.msg("Executing statement: '%s'" % s)
                 exec s in globals(), locals()
             except:
                 self.log.msg("Error running statement: %s" % s)
 
     def forward_logging(self):
         if self.log_url:
             self.log.info("Forwarding logging to %s"%self.log_url)
             context = zmq.Context.instance()
             lsock = context.socket(zmq.PUB)
             lsock.connect(self.log_url)
             handler = PUBHandler(lsock)
             handler.root_topic = 'controller'
             handler.setLevel(self.log_level)
             self.log.addHandler(handler)
     
     @catch_config_error
     def initialize(self, argv=None):
         super(IPControllerApp, self).initialize(argv)
         self.forward_logging()
         self.load_secondary_config()
         self.init_hub()
         self.init_schedulers()
     
     def start(self):
         # Start the subprocesses:
         self.factory.start()
         # children must be started before signals are setup,
         # otherwise signal-handling will fire multiple times
         for child in self.children:
             child.start()
         self.init_signal()
 
         self.write_pid_file(overwrite=True)
 
         try:
             self.factory.loop.start()
         except KeyboardInterrupt:
             self.log.critical("Interrupted, Exiting...\n")
         finally:
             self.cleanup_connection_files()
             
 
 
 def launch_new_instance():
     """Create and run the IPython controller"""
     if sys.platform == 'win32':
         # make sure we don't get called from a multiprocessing subprocess
         # this can result in infinite Controllers being started on Windows
         # which doesn't have a proper fork, so multiprocessing is wonky
         
         # this only comes up when IPython has been installed using vanilla
         # setuptools, and *not* distribute.
         import multiprocessing
         p = multiprocessing.current_process()
         # the main process has name 'MainProcess'
         # subprocesses will have names like 'Process-1'
         if p.name != 'MainProcess':
             # we are a subprocess, don't start another Controller!
             return
     app = IPControllerApp.instance()
     app.initialize()
     app.start()
 
 
 if __name__ == '__main__':
     launch_new_instance()

IPython/parallel/controller/heartmonitor.py

@@ -1,182 +1,182 b''
 #!/usr/bin/env python
 """
-A multi-heart Heartbeat system using PUB and XREP sockets. pings are sent out on the PUB,
-and hearts are tracked based on their XREQ identities.
+A multi-heart Heartbeat system using PUB and ROUTER sockets. pings are sent out on the PUB,
+and hearts are tracked based on their DEALER identities.
 
 Authors:
 
 * Min RK
 """
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2010-2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 from __future__ import print_function
 import time
 import uuid
 
 import zmq
 from zmq.devices import ThreadDevice
 from zmq.eventloop import ioloop, zmqstream
 
 from IPython.config.configurable import LoggingConfigurable
 from IPython.utils.py3compat import str_to_bytes
 from IPython.utils.traitlets import Set, Instance, CFloat, Integer
 
 from IPython.parallel.util import log_errors
 
 class Heart(object):
     """A basic heart object for responding to a HeartMonitor.
     This is a simple wrapper with defaults for the most common
     Device model for responding to heartbeats.
 
     It simply builds a threadsafe zmq.FORWARDER Device, defaulting to using
-    SUB/XREQ for in/out.
+    SUB/DEALER for in/out.
 
-    You can specify the XREQ's IDENTITY via the optional heart_id argument."""
+    You can specify the DEALER's IDENTITY via the optional heart_id argument."""
     device=None
     id=None
     def __init__(self, in_addr, out_addr, in_type=zmq.SUB, out_type=zmq.DEALER, heart_id=None):
         self.device = ThreadDevice(zmq.FORWARDER, in_type, out_type)
         # do not allow the device to share global Context.instance,
         # which is the default behavior in pyzmq > 2.1.10
         self.device.context_factory = zmq.Context
         
         self.device.daemon=True
         self.device.connect_in(in_addr)
         self.device.connect_out(out_addr)
         if in_type == zmq.SUB:
             self.device.setsockopt_in(zmq.SUBSCRIBE, b"")
         if heart_id is None:
             heart_id = uuid.uuid4().bytes
         self.device.setsockopt_out(zmq.IDENTITY, heart_id)
         self.id = heart_id
 
     def start(self):
         return self.device.start()
 
 
 class HeartMonitor(LoggingConfigurable):
     """A basic HeartMonitor class
     pingstream: a PUB stream
-    pongstream: an XREP stream
+    pongstream: an ROUTER stream
     period: the period of the heartbeat in milliseconds"""
 
     period = Integer(3000, config=True,
         help='The frequency at which the Hub pings the engines for heartbeats '
         '(in ms)',
     )
 
     pingstream=Instance('zmq.eventloop.zmqstream.ZMQStream')
     pongstream=Instance('zmq.eventloop.zmqstream.ZMQStream')
     loop = Instance('zmq.eventloop.ioloop.IOLoop')
     def _loop_default(self):
         return ioloop.IOLoop.instance()
 
     # not settable:
     hearts=Set()
     responses=Set()
     on_probation=Set()
     last_ping=CFloat(0)
     _new_handlers = Set()
     _failure_handlers = Set()
     lifetime = CFloat(0)
     tic = CFloat(0)
 
     def __init__(self, **kwargs):
         super(HeartMonitor, self).__init__(**kwargs)
 
         self.pongstream.on_recv(self.handle_pong)
 
     def start(self):
         self.tic = time.time()
         self.caller = ioloop.PeriodicCallback(self.beat, self.period, self.loop)
         self.caller.start()
 
     def add_new_heart_handler(self, handler):
         """add a new handler for new hearts"""
         self.log.debug("heartbeat::new_heart_handler: %s", handler)
         self._new_handlers.add(handler)
 
     def add_heart_failure_handler(self, handler):
         """add a new handler for heart failure"""
         self.log.debug("heartbeat::new heart failure handler: %s", handler)
         self._failure_handlers.add(handler)
 
     def beat(self):
         self.pongstream.flush()
         self.last_ping = self.lifetime
 
         toc = time.time()
         self.lifetime += toc-self.tic
         self.tic = toc
         self.log.debug("heartbeat::sending %s", self.lifetime)
         goodhearts = self.hearts.intersection(self.responses)
         missed_beats = self.hearts.difference(goodhearts)
         heartfailures = self.on_probation.intersection(missed_beats)
         newhearts = self.responses.difference(goodhearts)
         map(self.handle_new_heart, newhearts)
         map(self.handle_heart_failure, heartfailures)
         self.on_probation = missed_beats.intersection(self.hearts)
         self.responses = set()
         # print self.on_probation, self.hearts
         # self.log.debug("heartbeat::beat %.3f, %i beating hearts", self.lifetime, len(self.hearts))
         self.pingstream.send(str_to_bytes(str(self.lifetime)))
         # flush stream to force immediate socket send
         self.pingstream.flush()
 
     def handle_new_heart(self, heart):
         if self._new_handlers:
             for handler in self._new_handlers:
                 handler(heart)
         else:
             self.log.info("heartbeat::yay, got new heart %s!", heart)
         self.hearts.add(heart)
 
     def handle_heart_failure(self, heart):
         if self._failure_handlers:
             for handler in self._failure_handlers:
                 try:
                     handler(heart)
                 except Exception as e:
                     self.log.error("heartbeat::Bad Handler! %s", handler, exc_info=True)
                     pass
         else:
             self.log.info("heartbeat::Heart %s failed :(", heart)
         self.hearts.remove(heart)
 
 
     @log_errors
     def handle_pong(self, msg):
         "a heart just beat"
         current = str_to_bytes(str(self.lifetime))
         last = str_to_bytes(str(self.last_ping))
         if msg[1] == current:
             delta = time.time()-self.tic
             # self.log.debug("heartbeat::heart %r took %.2f ms to respond"%(msg[0], 1000*delta))
             self.responses.add(msg[0])
         elif msg[1] == last:
             delta = time.time()-self.tic + (self.lifetime-self.last_ping)
             self.log.warn("heartbeat::heart %r missed a beat, and took %.2f ms to respond", msg[0], 1000*delta)
             self.responses.add(msg[0])
         else:
             self.log.warn("heartbeat::got bad heartbeat (possibly old?): %s (current=%.3f)", msg[1], self.lifetime)
 
 
 if __name__ == '__main__':
     loop = ioloop.IOLoop.instance()
     context = zmq.Context()
     pub = context.socket(zmq.PUB)
     pub.bind('tcp://127.0.0.1:5555')
-    xrep = context.socket(zmq.ROUTER)
-    xrep.bind('tcp://127.0.0.1:5556')
+    router = context.socket(zmq.ROUTER)
+    router.bind('tcp://127.0.0.1:5556')
 
     outstream = zmqstream.ZMQStream(pub, loop)
-    instream = zmqstream.ZMQStream(xrep, loop)
+    instream = zmqstream.ZMQStream(router, loop)
 
     hb = HeartMonitor(loop, outstream, instream)
 
     loop.start()

IPython/parallel/controller/hub.py

@@ -1,1314 +1,1314 b''
 """The IPython Controller Hub with 0MQ
 This is the master object that handles connections from engines and clients,
 and monitors traffic through the various queues.
 
 Authors:
 
 * Min RK
 """
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2010-2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Imports
 #-----------------------------------------------------------------------------
 from __future__ import print_function
 
 import sys
 import time
 from datetime import datetime
 
 import zmq
 from zmq.eventloop import ioloop
 from zmq.eventloop.zmqstream import ZMQStream
 
 # internal:
 from IPython.utils.importstring import import_item
 from IPython.utils.py3compat import cast_bytes
 from IPython.utils.traitlets import (
         HasTraits, Instance, Integer, Unicode, Dict, Set, Tuple, CBytes, DottedObjectName
         )
 
 from IPython.parallel import error, util
 from IPython.parallel.factory import RegistrationFactory
 
 from IPython.zmq.session import SessionFactory
 
 from .heartmonitor import HeartMonitor
 
 #-----------------------------------------------------------------------------
 # Code
 #-----------------------------------------------------------------------------
 
 def _passer(*args, **kwargs):
     return
 
 def _printer(*args, **kwargs):
     print (args)
     print (kwargs)
 
 def empty_record():
     """Return an empty dict with all record keys."""
     return {
         'msg_id' : None,
         'header' : None,
         'content': None,
         'buffers': None,
         'submitted': None,
         'client_uuid' : None,
         'engine_uuid' : None,
         'started': None,
         'completed': None,
         'resubmitted': None,
         'received': None,
         'result_header' : None,
         'result_content' : None,
         'result_buffers' : None,
         'queue' : None,
         'pyin' : None,
         'pyout': None,
         'pyerr': None,
         'stdout': '',
         'stderr': '',
     }
 
 def init_record(msg):
     """Initialize a TaskRecord based on a request."""
     header = msg['header']
     return {
         'msg_id' : header['msg_id'],
         'header' : header,
         'content': msg['content'],
         'buffers': msg['buffers'],
         'submitted': header['date'],
         'client_uuid' : None,
         'engine_uuid' : None,
         'started': None,
         'completed': None,
         'resubmitted': None,
         'received': None,
         'result_header' : None,
         'result_content' : None,
         'result_buffers' : None,
         'queue' : None,
         'pyin' : None,
         'pyout': None,
         'pyerr': None,
         'stdout': '',
         'stderr': '',
     }
 
 
 class EngineConnector(HasTraits):
     """A simple object for accessing the various zmq connections of an object.
     Attributes are:
     id (int): engine ID
     uuid (str): uuid (unused?)
-    queue (str): identity of queue's XREQ socket
-    registration (str): identity of registration XREQ socket
-    heartbeat (str): identity of heartbeat XREQ socket
+    queue (str): identity of queue's DEALER socket
+    registration (str): identity of registration DEALER socket
+    heartbeat (str): identity of heartbeat DEALER socket
     """
     id=Integer(0)
     queue=CBytes()
     control=CBytes()
     registration=CBytes()
     heartbeat=CBytes()
     pending=Set()
 
 class HubFactory(RegistrationFactory):
     """The Configurable for setting up a Hub."""
 
     # port-pairs for monitoredqueues:
     hb = Tuple(Integer,Integer,config=True,
-        help="""XREQ/SUB Port pair for Engine heartbeats""")
+        help="""DEALER/SUB Port pair for Engine heartbeats""")
     def _hb_default(self):
         return tuple(util.select_random_ports(2))
 
     mux = Tuple(Integer,Integer,config=True,
         help="""Engine/Client Port pair for MUX queue""")
 
     def _mux_default(self):
         return tuple(util.select_random_ports(2))
 
     task = Tuple(Integer,Integer,config=True,
         help="""Engine/Client Port pair for Task queue""")
     def _task_default(self):
         return tuple(util.select_random_ports(2))
 
     control = Tuple(Integer,Integer,config=True,
         help="""Engine/Client Port pair for Control queue""")
 
     def _control_default(self):
         return tuple(util.select_random_ports(2))
 
     iopub = Tuple(Integer,Integer,config=True,
         help="""Engine/Client Port pair for IOPub relay""")
 
     def _iopub_default(self):
         return tuple(util.select_random_ports(2))
 
     # single ports:
     mon_port = Integer(config=True,
         help="""Monitor (SUB) port for queue traffic""")
 
     def _mon_port_default(self):
         return util.select_random_ports(1)[0]
 
     notifier_port = Integer(config=True,
         help="""PUB port for sending engine status notifications""")
 
     def _notifier_port_default(self):
         return util.select_random_ports(1)[0]
 
     engine_ip = Unicode('127.0.0.1', config=True,
         help="IP on which to listen for engine connections. [default: loopback]")
     engine_transport = Unicode('tcp', config=True,
         help="0MQ transport for engine connections. [default: tcp]")
 
     client_ip = Unicode('127.0.0.1', config=True,
         help="IP on which to listen for client connections. [default: loopback]")
     client_transport = Unicode('tcp', config=True,
         help="0MQ transport for client connections. [default : tcp]")
 
     monitor_ip = Unicode('127.0.0.1', config=True,
         help="IP on which to listen for monitor messages. [default: loopback]")
     monitor_transport = Unicode('tcp', config=True,
         help="0MQ transport for monitor messages. [default : tcp]")
 
     monitor_url = Unicode('')
 
     db_class = DottedObjectName('IPython.parallel.controller.dictdb.DictDB',
         config=True, help="""The class to use for the DB backend""")
 
     # not configurable
     db = Instance('IPython.parallel.controller.dictdb.BaseDB')
     heartmonitor = Instance('IPython.parallel.controller.heartmonitor.HeartMonitor')
 
     def _ip_changed(self, name, old, new):
         self.engine_ip = new
         self.client_ip = new
         self.monitor_ip = new
         self._update_monitor_url()
 
     def _update_monitor_url(self):
         self.monitor_url = "%s://%s:%i" % (self.monitor_transport, self.monitor_ip, self.mon_port)
 
     def _transport_changed(self, name, old, new):
         self.engine_transport = new
         self.client_transport = new
         self.monitor_transport = new
         self._update_monitor_url()
 
     def __init__(self, **kwargs):
         super(HubFactory, self).__init__(**kwargs)
         self._update_monitor_url()
 
 
     def construct(self):
         self.init_hub()
 
     def start(self):
         self.heartmonitor.start()
         self.log.info("Heartmonitor started")
 
     def init_hub(self):
         """construct"""
         client_iface = "%s://%s:" % (self.client_transport, self.client_ip) + "%i"
         engine_iface = "%s://%s:" % (self.engine_transport, self.engine_ip) + "%i"
 
         ctx = self.context
         loop = self.loop
 
         # Registrar socket
         q = ZMQStream(ctx.socket(zmq.ROUTER), loop)
         q.bind(client_iface % self.regport)
         self.log.info("Hub listening on %s for registration.", client_iface % self.regport)
         if self.client_ip != self.engine_ip:
             q.bind(engine_iface % self.regport)
             self.log.info("Hub listening on %s for registration.", engine_iface % self.regport)
 
         ### Engine connections ###
 
         # heartbeat
         hpub = ctx.socket(zmq.PUB)
         hpub.bind(engine_iface % self.hb[0])
         hrep = ctx.socket(zmq.ROUTER)
         hrep.bind(engine_iface % self.hb[1])
         self.heartmonitor = HeartMonitor(loop=loop, config=self.config, log=self.log,
                                 pingstream=ZMQStream(hpub,loop),
                                 pongstream=ZMQStream(hrep,loop)
                             )
 
         ### Client connections ###
         # Notifier socket
         n = ZMQStream(ctx.socket(zmq.PUB), loop)
         n.bind(client_iface%self.notifier_port)
 
         ### build and launch the queues ###
 
         # monitor socket
         sub = ctx.socket(zmq.SUB)
         sub.setsockopt(zmq.SUBSCRIBE, b"")
         sub.bind(self.monitor_url)
         sub.bind('inproc://monitor')
         sub = ZMQStream(sub, loop)
 
         # connect the db
         self.log.info('Hub using DB backend: %r'%(self.db_class.split()[-1]))
         # cdir = self.config.Global.cluster_dir
         self.db = import_item(str(self.db_class))(session=self.session.session,
                                             config=self.config, log=self.log)
         time.sleep(.25)
         try:
             scheme = self.config.TaskScheduler.scheme_name
         except AttributeError:
             from .scheduler import TaskScheduler
             scheme = TaskScheduler.scheme_name.get_default_value()
         # build connection dicts
         self.engine_info = {
             'control' : engine_iface%self.control[1],
             'mux': engine_iface%self.mux[1],
             'heartbeat': (engine_iface%self.hb[0], engine_iface%self.hb[1]),
             'task' : engine_iface%self.task[1],
             'iopub' : engine_iface%self.iopub[1],
             # 'monitor' : engine_iface%self.mon_port,
             }
 
         self.client_info = {
             'control' : client_iface%self.control[0],
             'mux': client_iface%self.mux[0],
             'task' : (scheme, client_iface%self.task[0]),
             'iopub' : client_iface%self.iopub[0],
             'notification': client_iface%self.notifier_port
             }
         self.log.debug("Hub engine addrs: %s", self.engine_info)
         self.log.debug("Hub client addrs: %s", self.client_info)
 
         # resubmit stream
         r = ZMQStream(ctx.socket(zmq.DEALER), loop)
         url = util.disambiguate_url(self.client_info['task'][-1])
         r.setsockopt(zmq.IDENTITY, self.session.bsession)
         r.connect(url)
 
         self.hub = Hub(loop=loop, session=self.session, monitor=sub, heartmonitor=self.heartmonitor,
                 query=q, notifier=n, resubmit=r, db=self.db,
                 engine_info=self.engine_info, client_info=self.client_info,
                 log=self.log)
 
 
 class Hub(SessionFactory):
     """The IPython Controller Hub with 0MQ connections
 
     Parameters
     ==========
     loop: zmq IOLoop instance
     session: Session object
     <removed> context: zmq context for creating new connections (?)
     queue: ZMQStream for monitoring the command queue (SUB)
-    query: ZMQStream for engine registration and client queries requests (XREP)
+    query: ZMQStream for engine registration and client queries requests (ROUTER)
     heartbeat: HeartMonitor object checking the pulse of the engines
     notifier: ZMQStream for broadcasting engine registration changes (PUB)
     db: connection to db for out of memory logging of commands
                 NotImplemented
     engine_info: dict of zmq connection information for engines to connect
                 to the queues.
     client_info: dict of zmq connection information for engines to connect
                 to the queues.
     """
     # internal data structures:
     ids=Set() # engine IDs
     keytable=Dict()
     by_ident=Dict()
     engines=Dict()
     clients=Dict()
     hearts=Dict()
     pending=Set()
     queues=Dict()  # pending msg_ids keyed by engine_id
     tasks=Dict() # pending msg_ids submitted as tasks, keyed by client_id
     completed=Dict() # completed msg_ids keyed by engine_id
     all_completed=Set() # completed msg_ids keyed by engine_id
     dead_engines=Set() # completed msg_ids keyed by engine_id
     unassigned=Set() # set of task msg_ds not yet assigned a destination
     incoming_registrations=Dict()
     registration_timeout=Integer()
     _idcounter=Integer(0)
 
     # objects from constructor:
     query=Instance(ZMQStream)
     monitor=Instance(ZMQStream)
     notifier=Instance(ZMQStream)
     resubmit=Instance(ZMQStream)
     heartmonitor=Instance(HeartMonitor)
     db=Instance(object)
     client_info=Dict()
     engine_info=Dict()
 
 
     def __init__(self, **kwargs):
         """
         # universal:
         loop: IOLoop for creating future connections
         session: streamsession for sending serialized data
         # engine:
         queue: ZMQStream for monitoring queue messages
         query: ZMQStream for engine+client registration and client requests
         heartbeat: HeartMonitor object for tracking engines
         # extra:
         db: ZMQStream for db connection (NotImplemented)
         engine_info: zmq address/protocol dict for engine connections
         client_info: zmq address/protocol dict for client connections
         """
 
         super(Hub, self).__init__(**kwargs)
         self.registration_timeout = max(5000, 2*self.heartmonitor.period)
 
         # validate connection dicts:
         for k,v in self.client_info.iteritems():
             if k == 'task':
                 util.validate_url_container(v[1])
             else:
                 util.validate_url_container(v)
         # util.validate_url_container(self.client_info)
         util.validate_url_container(self.engine_info)
 
         # register our callbacks
         self.query.on_recv(self.dispatch_query)
         self.monitor.on_recv(self.dispatch_monitor_traffic)
 
         self.heartmonitor.add_heart_failure_handler(self.handle_heart_failure)
         self.heartmonitor.add_new_heart_handler(self.handle_new_heart)
 
         self.monitor_handlers = {b'in' : self.save_queue_request,
                                 b'out': self.save_queue_result,
                                 b'intask': self.save_task_request,
                                 b'outtask': self.save_task_result,
                                 b'tracktask': self.save_task_destination,
                                 b'incontrol': _passer,
                                 b'outcontrol': _passer,
                                 b'iopub': self.save_iopub_message,
         }
 
         self.query_handlers = {'queue_request': self.queue_status,
                                 'result_request': self.get_results,
                                 'history_request': self.get_history,
                                 'db_request': self.db_query,
                                 'purge_request': self.purge_results,
                                 'load_request': self.check_load,
                                 'resubmit_request': self.resubmit_task,
                                 'shutdown_request': self.shutdown_request,
                                 'registration_request' : self.register_engine,
                                 'unregistration_request' : self.unregister_engine,
                                 'connection_request': self.connection_request,
         }
 
         # ignore resubmit replies
         self.resubmit.on_recv(lambda msg: None, copy=False)
 
         self.log.info("hub::created hub")
 
     @property
     def _next_id(self):
         """gemerate a new ID.
 
         No longer reuse old ids, just count from 0."""
         newid = self._idcounter
         self._idcounter += 1
         return newid
         # newid = 0
         # incoming = [id[0] for id in self.incoming_registrations.itervalues()]
         # # print newid, self.ids, self.incoming_registrations
         # while newid in self.ids or newid in incoming:
         #     newid += 1
         # return newid
 
     #-----------------------------------------------------------------------------
     # message validation
     #-----------------------------------------------------------------------------
 
     def _validate_targets(self, targets):
         """turn any valid targets argument into a list of integer ids"""
         if targets is None:
             # default to all
             return self.ids
 
         if isinstance(targets, (int,str,unicode)):
             # only one target specified
             targets = [targets]
         _targets = []
         for t in targets:
             # map raw identities to ids
             if isinstance(t, (str,unicode)):
                 t = self.by_ident.get(cast_bytes(t), t)
             _targets.append(t)
         targets = _targets
         bad_targets = [ t for t in targets if t not in self.ids ]
         if bad_targets:
             raise IndexError("No Such Engine: %r" % bad_targets)
         if not targets:
             raise IndexError("No Engines Registered")
         return targets
 
     #-----------------------------------------------------------------------------
     # dispatch methods (1 per stream)
     #-----------------------------------------------------------------------------
 
 
     @util.log_errors
     def dispatch_monitor_traffic(self, msg):
         """all ME and Task queue messages come through here, as well as
         IOPub traffic."""
         self.log.debug("monitor traffic: %r", msg[0])
         switch = msg[0]
         try:
             idents, msg = self.session.feed_identities(msg[1:])
         except ValueError:
             idents=[]
         if not idents:
             self.log.error("Monitor message without topic: %r", msg)
             return
         handler = self.monitor_handlers.get(switch, None)
         if handler is not None:
             handler(idents, msg)
         else:
             self.log.error("Unrecognized monitor topic: %r", switch)
 
 
     @util.log_errors
     def dispatch_query(self, msg):
         """Route registration requests and queries from clients."""
         try:
             idents, msg = self.session.feed_identities(msg)
         except ValueError:
             idents = []
         if not idents:
             self.log.error("Bad Query Message: %r", msg)
             return
         client_id = idents[0]
         try:
             msg = self.session.unserialize(msg, content=True)
         except Exception:
             content = error.wrap_exception()
             self.log.error("Bad Query Message: %r", msg, exc_info=True)
             self.session.send(self.query, "hub_error", ident=client_id,
                     content=content)
             return
         # print client_id, header, parent, content
         #switch on message type:
         msg_type = msg['header']['msg_type']
         self.log.info("client::client %r requested %r", client_id, msg_type)
         handler = self.query_handlers.get(msg_type, None)
         try:
             assert handler is not None, "Bad Message Type: %r" % msg_type
         except:
             content = error.wrap_exception()
             self.log.error("Bad Message Type: %r", msg_type, exc_info=True)
             self.session.send(self.query, "hub_error", ident=client_id,
                     content=content)
             return
 
         else:
             handler(idents, msg)
 
     def dispatch_db(self, msg):
         """"""
         raise NotImplementedError
 
     #---------------------------------------------------------------------------
     # handler methods (1 per event)
     #---------------------------------------------------------------------------
 
     #----------------------- Heartbeat --------------------------------------
 
     def handle_new_heart(self, heart):
         """handler to attach to heartbeater.
         Called when a new heart starts to beat.
         Triggers completion of registration."""
         self.log.debug("heartbeat::handle_new_heart(%r)", heart)
         if heart not in self.incoming_registrations:
             self.log.info("heartbeat::ignoring new heart: %r", heart)
         else:
             self.finish_registration(heart)
 
 
     def handle_heart_failure(self, heart):
         """handler to attach to heartbeater.
         called when a previously registered heart fails to respond to beat request.
         triggers unregistration"""
         self.log.debug("heartbeat::handle_heart_failure(%r)", heart)
         eid = self.hearts.get(heart, None)
         queue = self.engines[eid].queue
         if eid is None or self.keytable[eid] in self.dead_engines:
             self.log.info("heartbeat::ignoring heart failure %r (not an engine or already dead)", heart)
         else:
             self.unregister_engine(heart, dict(content=dict(id=eid, queue=queue)))
 
     #----------------------- MUX Queue Traffic ------------------------------
 
     def save_queue_request(self, idents, msg):
         if len(idents) < 2:
             self.log.error("invalid identity prefix: %r", idents)
             return
         queue_id, client_id = idents[:2]
         try:
             msg = self.session.unserialize(msg)
         except Exception:
             self.log.error("queue::client %r sent invalid message to %r: %r", client_id, queue_id, msg, exc_info=True)
             return
 
         eid = self.by_ident.get(queue_id, None)
         if eid is None:
             self.log.error("queue::target %r not registered", queue_id)
             self.log.debug("queue::    valid are: %r", self.by_ident.keys())
             return
         record = init_record(msg)
         msg_id = record['msg_id']
         self.log.info("queue::client %r submitted request %r to %s", client_id, msg_id, eid)
         # Unicode in records
         record['engine_uuid'] = queue_id.decode('ascii')
         record['client_uuid'] = client_id.decode('ascii')
         record['queue'] = 'mux'
 
         try:
             # it's posible iopub arrived first:
             existing = self.db.get_record(msg_id)
             for key,evalue in existing.iteritems():
                 rvalue = record.get(key, None)
                 if evalue and rvalue and evalue != rvalue:
                     self.log.warn("conflicting initial state for record: %r:%r <%r> %r", msg_id, rvalue, key, evalue)
                 elif evalue and not rvalue:
                     record[key] = evalue
             try:
                 self.db.update_record(msg_id, record)
             except Exception:
                 self.log.error("DB Error updating record %r", msg_id, exc_info=True)
         except KeyError:
             try:
                 self.db.add_record(msg_id, record)
             except Exception:
                 self.log.error("DB Error adding record %r", msg_id, exc_info=True)
 
 
         self.pending.add(msg_id)
         self.queues[eid].append(msg_id)
 
     def save_queue_result(self, idents, msg):
         if len(idents) < 2:
             self.log.error("invalid identity prefix: %r", idents)
             return
 
         client_id, queue_id = idents[:2]
         try:
             msg = self.session.unserialize(msg)
         except Exception:
             self.log.error("queue::engine %r sent invalid message to %r: %r",
                     queue_id, client_id, msg, exc_info=True)
             return
 
         eid = self.by_ident.get(queue_id, None)
         if eid is None:
             self.log.error("queue::unknown engine %r is sending a reply: ", queue_id)
             return
 
         parent = msg['parent_header']
         if not parent:
             return
         msg_id = parent['msg_id']
         if msg_id in self.pending:
             self.pending.remove(msg_id)
             self.all_completed.add(msg_id)
             self.queues[eid].remove(msg_id)
             self.completed[eid].append(msg_id)
             self.log.info("queue::request %r completed on %s", msg_id, eid)
         elif msg_id not in self.all_completed:
             # it could be a result from a dead engine that died before delivering the
             # result
             self.log.warn("queue:: unknown msg finished %r", msg_id)
             return
         # update record anyway, because the unregistration could have been premature
         rheader = msg['header']
         completed = rheader['date']
         started = rheader.get('started', None)
         result = {
             'result_header' : rheader,
             'result_content': msg['content'],
             'received': datetime.now(),
             'started' : started,
             'completed' : completed
         }
 
         result['result_buffers'] = msg['buffers']
         try:
             self.db.update_record(msg_id, result)
         except Exception:
             self.log.error("DB Error updating record %r", msg_id, exc_info=True)
 
 
     #--------------------- Task Queue Traffic ------------------------------
 
     def save_task_request(self, idents, msg):
         """Save the submission of a task."""
         client_id = idents[0]
 
         try:
             msg = self.session.unserialize(msg)
         except Exception:
             self.log.error("task::client %r sent invalid task message: %r",
                     client_id, msg, exc_info=True)
             return
         record = init_record(msg)
 
         record['client_uuid'] = client_id.decode('ascii')
         record['queue'] = 'task'
         header = msg['header']
         msg_id = header['msg_id']
         self.pending.add(msg_id)
         self.unassigned.add(msg_id)
         try:
             # it's posible iopub arrived first:
             existing = self.db.get_record(msg_id)
             if existing['resubmitted']:
                 for key in ('submitted', 'client_uuid', 'buffers'):
                     # don't clobber these keys on resubmit
                     # submitted and client_uuid should be different
                     # and buffers might be big, and shouldn't have changed
                     record.pop(key)
                     # still check content,header which should not change
                     # but are not expensive to compare as buffers
 
             for key,evalue in existing.iteritems():
                 if key.endswith('buffers'):
                     # don't compare buffers
                     continue
                 rvalue = record.get(key, None)
                 if evalue and rvalue and evalue != rvalue:
                     self.log.warn("conflicting initial state for record: %r:%r <%r> %r", msg_id, rvalue, key, evalue)
                 elif evalue and not rvalue:
                     record[key] = evalue
             try:
                 self.db.update_record(msg_id, record)
             except Exception:
                 self.log.error("DB Error updating record %r", msg_id, exc_info=True)
         except KeyError:
             try:
                 self.db.add_record(msg_id, record)
             except Exception:
                 self.log.error("DB Error adding record %r", msg_id, exc_info=True)
         except Exception:
             self.log.error("DB Error saving task request %r", msg_id, exc_info=True)
 
     def save_task_result(self, idents, msg):
         """save the result of a completed task."""
         client_id = idents[0]
         try:
             msg = self.session.unserialize(msg)
         except Exception:
             self.log.error("task::invalid task result message send to %r: %r",
                     client_id, msg, exc_info=True)
             return
 
         parent = msg['parent_header']
         if not parent:
             # print msg
             self.log.warn("Task %r had no parent!", msg)
             return
         msg_id = parent['msg_id']
         if msg_id in self.unassigned:
             self.unassigned.remove(msg_id)
 
         header = msg['header']
         engine_uuid = header.get('engine', u'')
         eid = self.by_ident.get(cast_bytes(engine_uuid), None)
         
         status = header.get('status', None)
 
         if msg_id in self.pending:
             self.log.info("task::task %r finished on %s", msg_id, eid)
             self.pending.remove(msg_id)
             self.all_completed.add(msg_id)
             if eid is not None:
                 if status != 'aborted':
                     self.completed[eid].append(msg_id)
                 if msg_id in self.tasks[eid]:
                     self.tasks[eid].remove(msg_id)
             completed = header['date']
             started = header.get('started', None)
             result = {
                 'result_header' : header,
                 'result_content': msg['content'],
                 'started' : started,
                 'completed' : completed,
                 'received' : datetime.now(),
                 'engine_uuid': engine_uuid,
             }
 
             result['result_buffers'] = msg['buffers']
             try:
                 self.db.update_record(msg_id, result)
             except Exception:
                 self.log.error("DB Error saving task request %r", msg_id, exc_info=True)
 
         else:
             self.log.debug("task::unknown task %r finished", msg_id)
 
     def save_task_destination(self, idents, msg):
         try:
             msg = self.session.unserialize(msg, content=True)
         except Exception:
             self.log.error("task::invalid task tracking message", exc_info=True)
             return
         content = msg['content']
         # print (content)
         msg_id = content['msg_id']
         engine_uuid = content['engine_id']
         eid = self.by_ident[cast_bytes(engine_uuid)]
 
         self.log.info("task::task %r arrived on %r", msg_id, eid)
         if msg_id in self.unassigned:
             self.unassigned.remove(msg_id)
         # else:
         #     self.log.debug("task::task %r not listed as MIA?!"%(msg_id))
 
         self.tasks[eid].append(msg_id)
         # self.pending[msg_id][1].update(received=datetime.now(),engine=(eid,engine_uuid))
         try:
             self.db.update_record(msg_id, dict(engine_uuid=engine_uuid))
         except Exception:
             self.log.error("DB Error saving task destination %r", msg_id, exc_info=True)
 
 
     def mia_task_request(self, idents, msg):
         raise NotImplementedError
         client_id = idents[0]
         # content = dict(mia=self.mia,status='ok')
         # self.session.send('mia_reply', content=content, idents=client_id)
 
 
     #--------------------- IOPub Traffic ------------------------------
 
     def save_iopub_message(self, topics, msg):
         """save an iopub message into the db"""
         # print (topics)
         try:
             msg = self.session.unserialize(msg, content=True)
         except Exception:
             self.log.error("iopub::invalid IOPub message", exc_info=True)
             return
 
         parent = msg['parent_header']
         if not parent:
             self.log.warn("iopub::IOPub message lacks parent: %r", msg)
             return
         msg_id = parent['msg_id']
         msg_type = msg['header']['msg_type']
         content = msg['content']
 
         # ensure msg_id is in db
         try:
             rec = self.db.get_record(msg_id)
         except KeyError:
             rec = empty_record()
             rec['msg_id'] = msg_id
             self.db.add_record(msg_id, rec)
         # stream
         d = {}
         if msg_type == 'stream':
             name = content['name']
             s = rec[name] or ''
             d[name] = s + content['data']
 
         elif msg_type == 'pyerr':
             d['pyerr'] = content
         elif msg_type == 'pyin':
             d['pyin'] = content['code']
         else:
             d[msg_type] = content.get('data', '')
 
         try:
             self.db.update_record(msg_id, d)
         except Exception:
             self.log.error("DB Error saving iopub message %r", msg_id, exc_info=True)
 
 
 
     #-------------------------------------------------------------------------
     # Registration requests
     #-------------------------------------------------------------------------
 
     def connection_request(self, client_id, msg):
         """Reply with connection addresses for clients."""
         self.log.info("client::client %r connected", client_id)
         content = dict(status='ok')
         content.update(self.client_info)
         jsonable = {}
         for k,v in self.keytable.iteritems():
             if v not in self.dead_engines:
                 jsonable[str(k)] = v.decode('ascii')
         content['engines'] = jsonable
         self.session.send(self.query, 'connection_reply', content, parent=msg, ident=client_id)
 
     def register_engine(self, reg, msg):
         """Register a new engine."""
         content = msg['content']
         try:
             queue = cast_bytes(content['queue'])
         except KeyError:
             self.log.error("registration::queue not specified", exc_info=True)
             return
         heart = content.get('heartbeat', None)
         if heart:
             heart = cast_bytes(heart)
         """register a new engine, and create the socket(s) necessary"""
         eid = self._next_id
         # print (eid, queue, reg, heart)
 
         self.log.debug("registration::register_engine(%i, %r, %r, %r)", eid, queue, reg, heart)
 
         content = dict(id=eid,status='ok')
         content.update(self.engine_info)
         # check if requesting available IDs:
         if queue in self.by_ident:
             try:
                 raise KeyError("queue_id %r in use" % queue)
             except:
                 content = error.wrap_exception()
                 self.log.error("queue_id %r in use", queue, exc_info=True)
         elif heart in self.hearts: # need to check unique hearts?
             try:
                 raise KeyError("heart_id %r in use" % heart)
             except:
                 self.log.error("heart_id %r in use", heart, exc_info=True)
                 content = error.wrap_exception()
         else:
             for h, pack in self.incoming_registrations.iteritems():
                 if heart == h:
                     try:
                         raise KeyError("heart_id %r in use" % heart)
                     except:
                         self.log.error("heart_id %r in use", heart, exc_info=True)
                         content = error.wrap_exception()
                     break
                 elif queue == pack[1]:
                     try:
                         raise KeyError("queue_id %r in use" % queue)
                     except:
                         self.log.error("queue_id %r in use", queue, exc_info=True)
                         content = error.wrap_exception()
                     break
 
         msg = self.session.send(self.query, "registration_reply",
                 content=content,
                 ident=reg)
 
         if content['status'] == 'ok':
             if heart in self.heartmonitor.hearts:
                 # already beating
                 self.incoming_registrations[heart] = (eid,queue,reg[0],None)
                 self.finish_registration(heart)
             else:
                 purge = lambda : self._purge_stalled_registration(heart)
                 dc = ioloop.DelayedCallback(purge, self.registration_timeout, self.loop)
                 dc.start()
                 self.incoming_registrations[heart] = (eid,queue,reg[0],dc)
         else:
             self.log.error("registration::registration %i failed: %r", eid, content['evalue'])
         return eid
 
     def unregister_engine(self, ident, msg):
         """Unregister an engine that explicitly requested to leave."""
         try:
             eid = msg['content']['id']
         except:
             self.log.error("registration::bad engine id for unregistration: %r", ident, exc_info=True)
             return
         self.log.info("registration::unregister_engine(%r)", eid)
         # print (eid)
         uuid = self.keytable[eid]
         content=dict(id=eid, queue=uuid.decode('ascii'))
         self.dead_engines.add(uuid)
         # self.ids.remove(eid)
         # uuid = self.keytable.pop(eid)
         #
         # ec = self.engines.pop(eid)
         # self.hearts.pop(ec.heartbeat)
         # self.by_ident.pop(ec.queue)
         # self.completed.pop(eid)
         handleit = lambda : self._handle_stranded_msgs(eid, uuid)
         dc = ioloop.DelayedCallback(handleit, self.registration_timeout, self.loop)
         dc.start()
         ############## TODO: HANDLE IT ################
 
         if self.notifier:
             self.session.send(self.notifier, "unregistration_notification", content=content)
 
     def _handle_stranded_msgs(self, eid, uuid):
         """Handle messages known to be on an engine when the engine unregisters.
 
         It is possible that this will fire prematurely - that is, an engine will
         go down after completing a result, and the client will be notified
         that the result failed and later receive the actual result.
         """
 
         outstanding = self.queues[eid]
 
         for msg_id in outstanding:
             self.pending.remove(msg_id)
             self.all_completed.add(msg_id)
             try:
                 raise error.EngineError("Engine %r died while running task %r" % (eid, msg_id))
             except:
                 content = error.wrap_exception()
             # build a fake header:
             header = {}
             header['engine'] = uuid
             header['date'] = datetime.now()
             rec = dict(result_content=content, result_header=header, result_buffers=[])
             rec['completed'] = header['date']
             rec['engine_uuid'] = uuid
             try:
                 self.db.update_record(msg_id, rec)
             except Exception:
                 self.log.error("DB Error handling stranded msg %r", msg_id, exc_info=True)
 
 
     def finish_registration(self, heart):
         """Second half of engine registration, called after our HeartMonitor
         has received a beat from the Engine's Heart."""
         try:
             (eid,queue,reg,purge) = self.incoming_registrations.pop(heart)
         except KeyError:
             self.log.error("registration::tried to finish nonexistant registration", exc_info=True)
             return
         self.log.info("registration::finished registering engine %i:%r", eid, queue)
         if purge is not None:
             purge.stop()
         control = queue
         self.ids.add(eid)
         self.keytable[eid] = queue
         self.engines[eid] = EngineConnector(id=eid, queue=queue, registration=reg,
                                     control=control, heartbeat=heart)
         self.by_ident[queue] = eid
         self.queues[eid] = list()
         self.tasks[eid] = list()
         self.completed[eid] = list()
         self.hearts[heart] = eid
         content = dict(id=eid, queue=self.engines[eid].queue.decode('ascii'))
         if self.notifier:
             self.session.send(self.notifier, "registration_notification", content=content)
         self.log.info("engine::Engine Connected: %i", eid)
 
     def _purge_stalled_registration(self, heart):
         if heart in self.incoming_registrations:
             eid = self.incoming_registrations.pop(heart)[0]
             self.log.info("registration::purging stalled registration: %i", eid)
         else:
             pass
 
     #-------------------------------------------------------------------------
     # Client Requests
     #-------------------------------------------------------------------------
 
     def shutdown_request(self, client_id, msg):
         """handle shutdown request."""
         self.session.send(self.query, 'shutdown_reply', content={'status': 'ok'}, ident=client_id)
         # also notify other clients of shutdown
         self.session.send(self.notifier, 'shutdown_notice', content={'status': 'ok'})
         dc = ioloop.DelayedCallback(lambda : self._shutdown(), 1000, self.loop)
         dc.start()
 
     def _shutdown(self):
         self.log.info("hub::hub shutting down.")
         time.sleep(0.1)
         sys.exit(0)
 
 
     def check_load(self, client_id, msg):
         content = msg['content']
         try:
             targets = content['targets']
             targets = self._validate_targets(targets)
         except:
             content = error.wrap_exception()
             self.session.send(self.query, "hub_error",
                     content=content, ident=client_id)
             return
 
         content = dict(status='ok')
         # loads = {}
         for t in targets:
             content[bytes(t)] = len(self.queues[t])+len(self.tasks[t])
         self.session.send(self.query, "load_reply", content=content, ident=client_id)
 
 
     def queue_status(self, client_id, msg):
         """Return the Queue status of one or more targets.
         if verbose: return the msg_ids
         else: return len of each type.
         keys: queue (pending MUX jobs)
             tasks (pending Task jobs)
             completed (finished jobs from both queues)"""
         content = msg['content']
         targets = content['targets']
         try:
             targets = self._validate_targets(targets)
         except:
             content = error.wrap_exception()
             self.session.send(self.query, "hub_error",
                     content=content, ident=client_id)
             return
         verbose = content.get('verbose', False)
         content = dict(status='ok')
         for t in targets:
             queue = self.queues[t]
             completed = self.completed[t]
             tasks = self.tasks[t]
             if not verbose:
                 queue = len(queue)
                 completed = len(completed)
                 tasks = len(tasks)
             content[str(t)] = {'queue': queue, 'completed': completed , 'tasks': tasks}
         content['unassigned'] = list(self.unassigned) if verbose else len(self.unassigned)
         # print (content)
         self.session.send(self.query, "queue_reply", content=content, ident=client_id)
 
     def purge_results(self, client_id, msg):
         """Purge results from memory. This method is more valuable before we move
         to a DB based message storage mechanism."""
         content = msg['content']
         self.log.info("Dropping records with %s", content)
         msg_ids = content.get('msg_ids', [])
         reply = dict(status='ok')
         if msg_ids == 'all':
             try:
                 self.db.drop_matching_records(dict(completed={'$ne':None}))
             except Exception:
                 reply = error.wrap_exception()
         else:
             pending = filter(lambda m: m in self.pending, msg_ids)
             if pending:
                 try:
                     raise IndexError("msg pending: %r" % pending[0])
                 except:
                     reply = error.wrap_exception()
             else:
                 try:
                     self.db.drop_matching_records(dict(msg_id={'$in':msg_ids}))
                 except Exception:
                     reply = error.wrap_exception()
 
             if reply['status'] == 'ok':
                 eids = content.get('engine_ids', [])
                 for eid in eids:
                     if eid not in self.engines:
                         try:
                             raise IndexError("No such engine: %i" % eid)
                         except:
                             reply = error.wrap_exception()
                         break
                     uid = self.engines[eid].queue
                     try:
                         self.db.drop_matching_records(dict(engine_uuid=uid, completed={'$ne':None}))
                     except Exception:
                         reply = error.wrap_exception()
                         break
 
         self.session.send(self.query, 'purge_reply', content=reply, ident=client_id)
 
     def resubmit_task(self, client_id, msg):
         """Resubmit one or more tasks."""
         def finish(reply):
             self.session.send(self.query, 'resubmit_reply', content=reply, ident=client_id)
 
         content = msg['content']
         msg_ids = content['msg_ids']
         reply = dict(status='ok')
         try:
             records = self.db.find_records({'msg_id' : {'$in' : msg_ids}}, keys=[
                 'header', 'content', 'buffers'])
         except Exception:
             self.log.error('db::db error finding tasks to resubmit', exc_info=True)
             return finish(error.wrap_exception())
 
         # validate msg_ids
         found_ids = [ rec['msg_id'] for rec in records ]
         pending_ids = [ msg_id for msg_id in found_ids if msg_id in self.pending ]
         if len(records) > len(msg_ids):
             try:
                 raise RuntimeError("DB appears to be in an inconsistent state."
                     "More matching records were found than should exist")
             except Exception:
                 return finish(error.wrap_exception())
         elif len(records) < len(msg_ids):
             missing = [ m for m in msg_ids if m not in found_ids ]
             try:
                 raise KeyError("No such msg(s): %r" % missing)
             except KeyError:
                 return finish(error.wrap_exception())
         elif pending_ids:
             pass
             # no need to raise on resubmit of pending task, now that we
             # resubmit under new ID, but do we want to raise anyway?
             # msg_id = invalid_ids[0]
             # try:
             #     raise ValueError("Task(s) %r appears to be inflight" % )
             # except Exception:
             #     return finish(error.wrap_exception())
 
         # mapping of original IDs to resubmitted IDs
         resubmitted = {}
 
         # send the messages
         for rec in records:
             header = rec['header']
             msg = self.session.msg(header['msg_type'], parent=header)
             msg_id = msg['msg_id']
             msg['content'] = rec['content']
             
             # use the old header, but update msg_id and timestamp
             fresh = msg['header']
             header['msg_id'] = fresh['msg_id']
             header['date'] = fresh['date']
             msg['header'] = header
 
             self.session.send(self.resubmit, msg, buffers=rec['buffers'])
 
             resubmitted[rec['msg_id']] = msg_id
             self.pending.add(msg_id)
             msg['buffers'] = rec['buffers']
             try:
                 self.db.add_record(msg_id, init_record(msg))
             except Exception:
                 self.log.error("db::DB Error updating record: %s", msg_id, exc_info=True)
 
         finish(dict(status='ok', resubmitted=resubmitted))
         
         # store the new IDs in the Task DB
         for msg_id, resubmit_id in resubmitted.iteritems():
             try:
                 self.db.update_record(msg_id, {'resubmitted' : resubmit_id})
             except Exception:
                 self.log.error("db::DB Error updating record: %s", msg_id, exc_info=True)
 
 
     def _extract_record(self, rec):
         """decompose a TaskRecord dict into subsection of reply for get_result"""
         io_dict = {}
         for key in ('pyin', 'pyout', 'pyerr', 'stdout', 'stderr'):
                 io_dict[key] = rec[key]
         content = { 'result_content': rec['result_content'],
                             'header': rec['header'],
                             'result_header' : rec['result_header'],
                             'received' : rec['received'],
                             'io' : io_dict,
                           }
         if rec['result_buffers']:
             buffers = map(bytes, rec['result_buffers'])
         else:
             buffers = []
 
         return content, buffers
 
     def get_results(self, client_id, msg):
         """Get the result of 1 or more messages."""
         content = msg['content']
         msg_ids = sorted(set(content['msg_ids']))
         statusonly = content.get('status_only', False)
         pending = []
         completed = []
         content = dict(status='ok')
         content['pending'] = pending
         content['completed'] = completed
         buffers = []
         if not statusonly:
             try:
                 matches = self.db.find_records(dict(msg_id={'$in':msg_ids}))
                 # turn match list into dict, for faster lookup
                 records = {}
                 for rec in matches:
                     records[rec['msg_id']] = rec
             except Exception:
                 content = error.wrap_exception()
                 self.session.send(self.query, "result_reply", content=content,
                                                     parent=msg, ident=client_id)
                 return
         else:
             records = {}
         for msg_id in msg_ids:
             if msg_id in self.pending:
                 pending.append(msg_id)
             elif msg_id in self.all_completed:
                 completed.append(msg_id)
                 if not statusonly:
                     c,bufs = self._extract_record(records[msg_id])
                     content[msg_id] = c
                     buffers.extend(bufs)
             elif msg_id in records:
                 if rec['completed']:
                     completed.append(msg_id)
                     c,bufs = self._extract_record(records[msg_id])
                     content[msg_id] = c
                     buffers.extend(bufs)
                 else:
                     pending.append(msg_id)
             else:
                 try:
                     raise KeyError('No such message: '+msg_id)
                 except:
                     content = error.wrap_exception()
                 break
         self.session.send(self.query, "result_reply", content=content,
                                             parent=msg, ident=client_id,
                                             buffers=buffers)
 
     def get_history(self, client_id, msg):
         """Get a list of all msg_ids in our DB records"""
         try:
             msg_ids = self.db.get_history()
         except Exception as e:
             content = error.wrap_exception()
         else:
             content = dict(status='ok', history=msg_ids)
 
         self.session.send(self.query, "history_reply", content=content,
                                             parent=msg, ident=client_id)
 
     def db_query(self, client_id, msg):
         """Perform a raw query on the task record database."""
         content = msg['content']
         query = content.get('query', {})
         keys = content.get('keys', None)
         buffers = []
         empty = list()
         try:
             records = self.db.find_records(query, keys)
         except Exception as e:
             content = error.wrap_exception()
         else:
             # extract buffers from reply content:
             if keys is not None:
                 buffer_lens = [] if 'buffers' in keys else None
                 result_buffer_lens = [] if 'result_buffers' in keys else None
             else:
                 buffer_lens = None
                 result_buffer_lens = None
 
             for rec in records:
                 # buffers may be None, so double check
                 b = rec.pop('buffers', empty) or empty
                 if buffer_lens is not None:
                     buffer_lens.append(len(b))
                     buffers.extend(b)
                 rb = rec.pop('result_buffers', empty) or empty
                 if result_buffer_lens is not None:
                     result_buffer_lens.append(len(rb))
                     buffers.extend(rb)
             content = dict(status='ok', records=records, buffer_lens=buffer_lens,
                                     result_buffer_lens=result_buffer_lens)
         # self.log.debug (content)
         self.session.send(self.query, "db_reply", content=content,
                                             parent=msg, ident=client_id,
                                             buffers=buffers)
 

IPython/parallel/controller/scheduler.py

@@ -1,768 +1,768 b''
 """The Python scheduler for rich scheduling.
 
 The Pure ZMQ scheduler does not allow routing schemes other than LRU,
 nor does it check msg_id DAG dependencies. For those, a slightly slower
 Python Scheduler exists.
 
 Authors:
 
 * Min RK
 """
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2010-2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 #----------------------------------------------------------------------
 # Imports
 #----------------------------------------------------------------------
 
 from __future__ import print_function
 
 import logging
 import sys
 import time
 
 from datetime import datetime, timedelta
 from random import randint, random
 from types import FunctionType
 
 try:
     import numpy
 except ImportError:
     numpy = None
 
 import zmq
 from zmq.eventloop import ioloop, zmqstream
 
 # local imports
 from IPython.external.decorator import decorator
 from IPython.config.application import Application
 from IPython.config.loader import Config
 from IPython.utils.traitlets import Instance, Dict, List, Set, Integer, Enum, CBytes
 from IPython.utils.py3compat import cast_bytes
 
 from IPython.parallel import error, util
 from IPython.parallel.factory import SessionFactory
 from IPython.parallel.util import connect_logger, local_logger
 
 from .dependency import Dependency
 
 @decorator
 def logged(f,self,*args,**kwargs):
     # print ("#--------------------")
     self.log.debug("scheduler::%s(*%s,**%s)", f.func_name, args, kwargs)
     # print ("#--")
     return f(self,*args, **kwargs)
 
 #----------------------------------------------------------------------
 # Chooser functions
 #----------------------------------------------------------------------
 
 def plainrandom(loads):
     """Plain random pick."""
     n = len(loads)
     return randint(0,n-1)
 
 def lru(loads):
     """Always pick the front of the line.
 
     The content of `loads` is ignored.
 
     Assumes LRU ordering of loads, with oldest first.
     """
     return 0
 
 def twobin(loads):
     """Pick two at random, use the LRU of the two.
 
     The content of loads is ignored.
 
     Assumes LRU ordering of loads, with oldest first.
     """
     n = len(loads)
     a = randint(0,n-1)
     b = randint(0,n-1)
     return min(a,b)
 
 def weighted(loads):
     """Pick two at random using inverse load as weight.
 
     Return the less loaded of the two.
     """
     # weight 0 a million times more than 1:
     weights = 1./(1e-6+numpy.array(loads))
     sums = weights.cumsum()
     t = sums[-1]
     x = random()*t
     y = random()*t
     idx = 0
     idy = 0
     while sums[idx] < x:
         idx += 1
     while sums[idy] < y:
         idy += 1
     if weights[idy] > weights[idx]:
         return idy
     else:
         return idx
 
 def leastload(loads):
     """Always choose the lowest load.
 
     If the lowest load occurs more than once, the first
     occurance will be used.  If loads has LRU ordering, this means
     the LRU of those with the lowest load is chosen.
     """
     return loads.index(min(loads))
 
 #---------------------------------------------------------------------
 # Classes
 #---------------------------------------------------------------------
 
 
 # store empty default dependency:
 MET = Dependency([])
 
 
 class Job(object):
     """Simple container for a job"""
     def __init__(self, msg_id, raw_msg, idents, msg, header, targets, after, follow, timeout):
         self.msg_id = msg_id
         self.raw_msg = raw_msg
         self.idents = idents
         self.msg = msg
         self.header = header
         self.targets = targets
         self.after = after
         self.follow = follow
         self.timeout = timeout
         
         
         self.timestamp = time.time()
         self.blacklist = set()
 
     @property
     def dependents(self):
         return self.follow.union(self.after)
 
 class TaskScheduler(SessionFactory):
     """Python TaskScheduler object.
 
     This is the simplest object that supports msg_id based
     DAG dependencies. *Only* task msg_ids are checked, not
     msg_ids of jobs submitted via the MUX queue.
 
     """
 
     hwm = Integer(1, config=True,
         help="""specify the High Water Mark (HWM) for the downstream
         socket in the Task scheduler. This is the maximum number
         of allowed outstanding tasks on each engine.
         
         The default (1) means that only one task can be outstanding on each
         engine.  Setting TaskScheduler.hwm=0 means there is no limit, and the
         engines continue to be assigned tasks while they are working,
         effectively hiding network latency behind computation, but can result
         in an imbalance of work when submitting many heterogenous tasks all at
         once.  Any positive value greater than one is a compromise between the
         two.
 
         """
     )
     scheme_name = Enum(('leastload', 'pure', 'lru', 'plainrandom', 'weighted', 'twobin'),
         'leastload', config=True, allow_none=False,
         help="""select the task scheduler scheme  [default: Python LRU]
         Options are: 'pure', 'lru', 'plainrandom', 'weighted', 'twobin','leastload'"""
     )
     def _scheme_name_changed(self, old, new):
         self.log.debug("Using scheme %r"%new)
         self.scheme = globals()[new]
 
     # input arguments:
     scheme = Instance(FunctionType) # function for determining the destination
     def _scheme_default(self):
         return leastload
     client_stream = Instance(zmqstream.ZMQStream) # client-facing stream
     engine_stream = Instance(zmqstream.ZMQStream) # engine-facing stream
     notifier_stream = Instance(zmqstream.ZMQStream) # hub-facing sub stream
     mon_stream = Instance(zmqstream.ZMQStream) # hub-facing pub stream
 
     # internals:
     graph = Dict() # dict by msg_id of [ msg_ids that depend on key ]
     retries = Dict() # dict by msg_id of retries remaining (non-neg ints)
     # waiting = List() # list of msg_ids ready to run, but haven't due to HWM
     depending = Dict() # dict by msg_id of Jobs
     pending = Dict() # dict by engine_uuid of submitted tasks
     completed = Dict() # dict by engine_uuid of completed tasks
     failed = Dict() # dict by engine_uuid of failed tasks
     destinations = Dict() # dict by msg_id of engine_uuids where jobs ran (reverse of completed+failed)
     clients = Dict() # dict by msg_id for who submitted the task
     targets = List() # list of target IDENTs
     loads = List() # list of engine loads
     # full = Set() # set of IDENTs that have HWM outstanding tasks
     all_completed = Set() # set of all completed tasks
     all_failed = Set() # set of all failed tasks
     all_done = Set() # set of all finished tasks=union(completed,failed)
     all_ids = Set() # set of all submitted task IDs
 
     auditor = Instance('zmq.eventloop.ioloop.PeriodicCallback')
 
     ident = CBytes() # ZMQ identity. This should just be self.session.session
                      # but ensure Bytes
     def _ident_default(self):
         return self.session.bsession
 
     def start(self):
         self.engine_stream.on_recv(self.dispatch_result, copy=False)
         self.client_stream.on_recv(self.dispatch_submission, copy=False)
 
         self._notification_handlers = dict(
             registration_notification = self._register_engine,
             unregistration_notification = self._unregister_engine
         )
         self.notifier_stream.on_recv(self.dispatch_notification)
         self.auditor = ioloop.PeriodicCallback(self.audit_timeouts, 2e3, self.loop) # 1 Hz
         self.auditor.start()
         self.log.info("Scheduler started [%s]"%self.scheme_name)
 
     def resume_receiving(self):
         """Resume accepting jobs."""
         self.client_stream.on_recv(self.dispatch_submission, copy=False)
 
     def stop_receiving(self):
         """Stop accepting jobs while there are no engines.
         Leave them in the ZMQ queue."""
         self.client_stream.on_recv(None)
 
     #-----------------------------------------------------------------------
     # [Un]Registration Handling
     #-----------------------------------------------------------------------
 
     
     @util.log_errors
     def dispatch_notification(self, msg):
         """dispatch register/unregister events."""
         try:
             idents,msg = self.session.feed_identities(msg)
         except ValueError:
             self.log.warn("task::Invalid Message: %r",msg)
             return
         try:
             msg = self.session.unserialize(msg)
         except ValueError:
             self.log.warn("task::Unauthorized message from: %r"%idents)
             return
 
         msg_type = msg['header']['msg_type']
 
         handler = self._notification_handlers.get(msg_type, None)
         if handler is None:
             self.log.error("Unhandled message type: %r"%msg_type)
         else:
             try:
                 handler(cast_bytes(msg['content']['queue']))
             except Exception:
                 self.log.error("task::Invalid notification msg: %r", msg, exc_info=True)
 
     def _register_engine(self, uid):
         """New engine with ident `uid` became available."""
         # head of the line:
         self.targets.insert(0,uid)
         self.loads.insert(0,0)
 
         # initialize sets
         self.completed[uid] = set()
         self.failed[uid] = set()
         self.pending[uid] = {}
 
         # rescan the graph:
         self.update_graph(None)
 
     def _unregister_engine(self, uid):
         """Existing engine with ident `uid` became unavailable."""
         if len(self.targets) == 1:
             # this was our only engine
             pass
 
         # handle any potentially finished tasks:
         self.engine_stream.flush()
 
         # don't pop destinations, because they might be used later
         # map(self.destinations.pop, self.completed.pop(uid))
         # map(self.destinations.pop, self.failed.pop(uid))
 
         # prevent this engine from receiving work
         idx = self.targets.index(uid)
         self.targets.pop(idx)
         self.loads.pop(idx)
 
         # wait 5 seconds before cleaning up pending jobs, since the results might
         # still be incoming
         if self.pending[uid]:
             dc = ioloop.DelayedCallback(lambda : self.handle_stranded_tasks(uid), 5000, self.loop)
             dc.start()
         else:
             self.completed.pop(uid)
             self.failed.pop(uid)
 
 
     def handle_stranded_tasks(self, engine):
         """Deal with jobs resident in an engine that died."""
         lost = self.pending[engine]
         for msg_id in lost.keys():
             if msg_id not in self.pending[engine]:
                 # prevent double-handling of messages
                 continue
 
             raw_msg = lost[msg_id].raw_msg
             idents,msg = self.session.feed_identities(raw_msg, copy=False)
             parent = self.session.unpack(msg[1].bytes)
             idents = [engine, idents[0]]
 
             # build fake error reply
             try:
                 raise error.EngineError("Engine %r died while running task %r"%(engine, msg_id))
             except:
                 content = error.wrap_exception()
             # build fake header
             header = dict(
                 status='error',
                 engine=engine,
                 date=datetime.now(),
             )
             msg = self.session.msg('apply_reply', content, parent=parent, subheader=header)
             raw_reply = map(zmq.Message, self.session.serialize(msg, ident=idents))
             # and dispatch it
             self.dispatch_result(raw_reply)
 
         # finally scrub completed/failed lists
         self.completed.pop(engine)
         self.failed.pop(engine)
 
 
     #-----------------------------------------------------------------------
     # Job Submission
     #-----------------------------------------------------------------------
     
     
     @util.log_errors
     def dispatch_submission(self, raw_msg):
         """Dispatch job submission to appropriate handlers."""
         # ensure targets up to date:
         self.notifier_stream.flush()
         try:
             idents, msg = self.session.feed_identities(raw_msg, copy=False)
             msg = self.session.unserialize(msg, content=False, copy=False)
         except Exception:
             self.log.error("task::Invaid task msg: %r"%raw_msg, exc_info=True)
             return
 
 
         # send to monitor
         self.mon_stream.send_multipart([b'intask']+raw_msg, copy=False)
 
         header = msg['header']
         msg_id = header['msg_id']
         self.all_ids.add(msg_id)
 
         # get targets as a set of bytes objects
         # from a list of unicode objects
         targets = header.get('targets', [])
         targets = map(cast_bytes, targets)
         targets = set(targets)
 
         retries = header.get('retries', 0)
         self.retries[msg_id] = retries
 
         # time dependencies
         after = header.get('after', None)
         if after:
             after = Dependency(after)
             if after.all:
                 if after.success:
                     after = Dependency(after.difference(self.all_completed),
                                 success=after.success,
                                 failure=after.failure,
                                 all=after.all,
                     )
                 if after.failure:
                     after = Dependency(after.difference(self.all_failed),
                                 success=after.success,
                                 failure=after.failure,
                                 all=after.all,
                     )
             if after.check(self.all_completed, self.all_failed):
                 # recast as empty set, if `after` already met,
                 # to prevent unnecessary set comparisons
                 after = MET
         else:
             after = MET
 
         # location dependencies
         follow = Dependency(header.get('follow', []))
 
         # turn timeouts into datetime objects:
         timeout = header.get('timeout', None)
         if timeout:
             # cast to float, because jsonlib returns floats as decimal.Decimal,
             # which timedelta does not accept
             timeout = datetime.now() + timedelta(0,float(timeout),0)
 
         job = Job(msg_id=msg_id, raw_msg=raw_msg, idents=idents, msg=msg,
                  header=header, targets=targets, after=after, follow=follow,
                  timeout=timeout,
         )
 
         # validate and reduce dependencies:
         for dep in after,follow:
             if not dep: # empty dependency
                 continue
             # check valid:
             if msg_id in dep or dep.difference(self.all_ids):
                 self.depending[msg_id] = job
                 return self.fail_unreachable(msg_id, error.InvalidDependency)
             # check if unreachable:
             if dep.unreachable(self.all_completed, self.all_failed):
                 self.depending[msg_id] = job
                 return self.fail_unreachable(msg_id)
 
         if after.check(self.all_completed, self.all_failed):
             # time deps already met, try to run
             if not self.maybe_run(job):
                 # can't run yet
                 if msg_id not in self.all_failed:
                     # could have failed as unreachable
                     self.save_unmet(job)
         else:
             self.save_unmet(job)
 
     def audit_timeouts(self):
         """Audit all waiting tasks for expired timeouts."""
         now = datetime.now()
         for msg_id in self.depending.keys():
             # must recheck, in case one failure cascaded to another:
             if msg_id in self.depending:
                 job = self.depending[msg_id]
                 if job.timeout and job.timeout < now:
                     self.fail_unreachable(msg_id, error.TaskTimeout)
 
     def fail_unreachable(self, msg_id, why=error.ImpossibleDependency):
         """a task has become unreachable, send a reply with an ImpossibleDependency
         error."""
         if msg_id not in self.depending:
             self.log.error("msg %r already failed!", msg_id)
             return
         job = self.depending.pop(msg_id)
         for mid in job.dependents:
             if mid in self.graph:
                 self.graph[mid].remove(msg_id)
 
         try:
             raise why()
         except:
             content = error.wrap_exception()
 
         self.all_done.add(msg_id)
         self.all_failed.add(msg_id)
 
         msg = self.session.send(self.client_stream, 'apply_reply', content,
                                                 parent=job.header, ident=job.idents)
         self.session.send(self.mon_stream, msg, ident=[b'outtask']+job.idents)
 
         self.update_graph(msg_id, success=False)
 
     def maybe_run(self, job):
         """check location dependencies, and run if they are met."""
         msg_id = job.msg_id
         self.log.debug("Attempting to assign task %s", msg_id)
         if not self.targets:
             # no engines, definitely can't run
             return False
         
         if job.follow or job.targets or job.blacklist or self.hwm:
             # we need a can_run filter
             def can_run(idx):
                 # check hwm
                 if self.hwm and self.loads[idx] == self.hwm:
                     return False
                 target = self.targets[idx]
                 # check blacklist
                 if target in job.blacklist:
                     return False
                 # check targets
                 if job.targets and target not in job.targets:
                     return False
                 # check follow
                 return job.follow.check(self.completed[target], self.failed[target])
 
             indices = filter(can_run, range(len(self.targets)))
 
             if not indices:
                 # couldn't run
                 if job.follow.all:
                     # check follow for impossibility
                     dests = set()
                     relevant = set()
                     if job.follow.success:
                         relevant = self.all_completed
                     if job.follow.failure:
                         relevant = relevant.union(self.all_failed)
                     for m in job.follow.intersection(relevant):
                         dests.add(self.destinations[m])
                     if len(dests) > 1:
                         self.depending[msg_id] = job
                         self.fail_unreachable(msg_id)
                         return False
                 if job.targets:
                     # check blacklist+targets for impossibility
                     job.targets.difference_update(job.blacklist)
                     if not job.targets or not job.targets.intersection(self.targets):
                         self.depending[msg_id] = job
                         self.fail_unreachable(msg_id)
                         return False
                 return False
         else:
             indices = None
 
         self.submit_task(job, indices)
         return True
 
     def save_unmet(self, job):
         """Save a message for later submission when its dependencies are met."""
         msg_id = job.msg_id
         self.depending[msg_id] = job
         # track the ids in follow or after, but not those already finished
         for dep_id in job.after.union(job.follow).difference(self.all_done):
             if dep_id not in self.graph:
                 self.graph[dep_id] = set()
             self.graph[dep_id].add(msg_id)
 
     def submit_task(self, job, indices=None):
         """Submit a task to any of a subset of our targets."""
         if indices:
             loads = [self.loads[i] for i in indices]
         else:
             loads = self.loads
         idx = self.scheme(loads)
         if indices:
             idx = indices[idx]
         target = self.targets[idx]
         # print (target, map(str, msg[:3]))
         # send job to the engine
         self.engine_stream.send(target, flags=zmq.SNDMORE, copy=False)
         self.engine_stream.send_multipart(job.raw_msg, copy=False)
         # update load
         self.add_job(idx)
         self.pending[target][job.msg_id] = job
         # notify Hub
         content = dict(msg_id=job.msg_id, engine_id=target.decode('ascii'))
         self.session.send(self.mon_stream, 'task_destination', content=content,
                         ident=[b'tracktask',self.ident])
 
 
     #-----------------------------------------------------------------------
     # Result Handling
     #-----------------------------------------------------------------------
     
     
     @util.log_errors
     def dispatch_result(self, raw_msg):
         """dispatch method for result replies"""
         try:
             idents,msg = self.session.feed_identities(raw_msg, copy=False)
             msg = self.session.unserialize(msg, content=False, copy=False)
             engine = idents[0]
             try:
                 idx = self.targets.index(engine)
             except ValueError:
                 pass # skip load-update for dead engines
             else:
                 self.finish_job(idx)
         except Exception:
             self.log.error("task::Invaid result: %r", raw_msg, exc_info=True)
             return
 
         header = msg['header']
         parent = msg['parent_header']
         if header.get('dependencies_met', True):
             success = (header['status'] == 'ok')
             msg_id = parent['msg_id']
             retries = self.retries[msg_id]
             if not success and retries > 0:
                 # failed
                 self.retries[msg_id] = retries - 1
                 self.handle_unmet_dependency(idents, parent)
             else:
                 del self.retries[msg_id]
                 # relay to client and update graph
                 self.handle_result(idents, parent, raw_msg, success)
                 # send to Hub monitor
                 self.mon_stream.send_multipart([b'outtask']+raw_msg, copy=False)
         else:
             self.handle_unmet_dependency(idents, parent)
 
     def handle_result(self, idents, parent, raw_msg, success=True):
         """handle a real task result, either success or failure"""
         # first, relay result to client
         engine = idents[0]
         client = idents[1]
-        # swap_ids for XREP-XREP mirror
+        # swap_ids for ROUTER-ROUTER mirror
         raw_msg[:2] = [client,engine]
         # print (map(str, raw_msg[:4]))
         self.client_stream.send_multipart(raw_msg, copy=False)
         # now, update our data structures
         msg_id = parent['msg_id']
         self.pending[engine].pop(msg_id)
         if success:
             self.completed[engine].add(msg_id)
             self.all_completed.add(msg_id)
         else:
             self.failed[engine].add(msg_id)
             self.all_failed.add(msg_id)
         self.all_done.add(msg_id)
         self.destinations[msg_id] = engine
 
         self.update_graph(msg_id, success)
 
     def handle_unmet_dependency(self, idents, parent):
         """handle an unmet dependency"""
         engine = idents[0]
         msg_id = parent['msg_id']
 
         job = self.pending[engine].pop(msg_id)
         job.blacklist.add(engine)
 
         if job.blacklist == job.targets:
             self.depending[msg_id] = job
             self.fail_unreachable(msg_id)
         elif not self.maybe_run(job):
             # resubmit failed
             if msg_id not in self.all_failed:
                 # put it back in our dependency tree
                 self.save_unmet(job)
 
         if self.hwm:
             try:
                 idx = self.targets.index(engine)
             except ValueError:
                 pass # skip load-update for dead engines
             else:
                 if self.loads[idx] == self.hwm-1:
                     self.update_graph(None)
 
 
 
     def update_graph(self, dep_id=None, success=True):
         """dep_id just finished. Update our dependency
         graph and submit any jobs that just became runable.
 
         Called with dep_id=None to update entire graph for hwm, but without finishing
         a task.
         """
         # print ("\n\n***********")
         # pprint (dep_id)
         # pprint (self.graph)
         # pprint (self.depending)
         # pprint (self.all_completed)
         # pprint (self.all_failed)
         # print ("\n\n***********\n\n")
         # update any jobs that depended on the dependency
         jobs = self.graph.pop(dep_id, [])
 
         # recheck *all* jobs if
         # a) we have HWM and an engine just become no longer full
         # or b) dep_id was given as None
         
         if dep_id is None or self.hwm and any( [ load==self.hwm-1 for load in self.loads ]):
             jobs = self.depending.keys()
         
         for msg_id in sorted(jobs, key=lambda msg_id: self.depending[msg_id].timestamp):
             job = self.depending[msg_id]
 
             if job.after.unreachable(self.all_completed, self.all_failed)\
                     or job.follow.unreachable(self.all_completed, self.all_failed):
                 self.fail_unreachable(msg_id)
 
             elif job.after.check(self.all_completed, self.all_failed): # time deps met, maybe run
                 if self.maybe_run(job):
 
                     self.depending.pop(msg_id)
                     for mid in job.dependents:
                         if mid in self.graph:
                             self.graph[mid].remove(msg_id)
 
     #----------------------------------------------------------------------
     # methods to be overridden by subclasses
     #----------------------------------------------------------------------
 
     def add_job(self, idx):
         """Called after self.targets[idx] just got the job with header.
         Override with subclasses.  The default ordering is simple LRU.
         The default loads are the number of outstanding jobs."""
         self.loads[idx] += 1
         for lis in (self.targets, self.loads):
             lis.append(lis.pop(idx))
 
 
     def finish_job(self, idx):
         """Called after self.targets[idx] just finished a job.
         Override with subclasses."""
         self.loads[idx] -= 1
 
 
 
 def launch_scheduler(in_addr, out_addr, mon_addr, not_addr, config=None,
                         logname='root', log_url=None, loglevel=logging.DEBUG,
                         identity=b'task', in_thread=False):
 
     ZMQStream = zmqstream.ZMQStream
 
     if config:
         # unwrap dict back into Config
         config = Config(config)
 
     if in_thread:
         # use instance() to get the same Context/Loop as our parent
         ctx = zmq.Context.instance()
         loop = ioloop.IOLoop.instance()
     else:
         # in a process, don't use instance()
         # for safety with multiprocessing
         ctx = zmq.Context()
         loop = ioloop.IOLoop()
     ins = ZMQStream(ctx.socket(zmq.ROUTER),loop)
     ins.setsockopt(zmq.IDENTITY, identity)
     ins.bind(in_addr)
 
     outs = ZMQStream(ctx.socket(zmq.ROUTER),loop)
     outs.setsockopt(zmq.IDENTITY, identity)
     outs.bind(out_addr)
     mons = zmqstream.ZMQStream(ctx.socket(zmq.PUB),loop)
     mons.connect(mon_addr)
     nots = zmqstream.ZMQStream(ctx.socket(zmq.SUB),loop)
     nots.setsockopt(zmq.SUBSCRIBE, b'')
     nots.connect(not_addr)
 
     # setup logging.
     if in_thread:
         log = Application.instance().log
     else:
         if log_url:
             log = connect_logger(logname, ctx, log_url, root="scheduler", loglevel=loglevel)
         else:
             log = local_logger(logname, loglevel)
 
     scheduler = TaskScheduler(client_stream=ins, engine_stream=outs,
                             mon_stream=mons, notifier_stream=nots,
                             loop=loop, log=log,
                             config=config)
     scheduler.start()
     if not in_thread:
         try:
             loop.start()
         except KeyboardInterrupt:
             scheduler.log.critical("Interrupted, exiting...")
 

IPython/zmq/__init__.py

@@ -1,68 +1,68 b''
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2010-2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING.txt, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Verify zmq version dependency >= 2.1.4
 #-----------------------------------------------------------------------------
 
 import warnings
 from distutils.version import LooseVersion as V
 
 
 def patch_pyzmq():
     """backport a few patches from newer pyzmq
     
     These can be removed as we bump our minimum pyzmq version
     """
     
     import zmq
     
     # ioloop.install, introduced in pyzmq 2.1.7
     from zmq.eventloop import ioloop
     
     def install():
         import tornado.ioloop
         tornado.ioloop.IOLoop = ioloop.IOLoop
     
     if not hasattr(ioloop, 'install'):
         ioloop.install = install
     
     # fix missing DEALER/ROUTER aliases in pyzmq < 2.1.9
     if not hasattr(zmq, 'DEALER'):
-        zmq.DEALER = zmq.XREQ
+        zmq.DEALER = zmq.DEALER
     if not hasattr(zmq, 'ROUTER'):
-        zmq.ROUTER = zmq.XREP
+        zmq.ROUTER = zmq.ROUTER
     
     # fallback on stdlib json if jsonlib is selected, because jsonlib breaks things.
     # jsonlib support is removed from pyzmq >= 2.2.0
 
     from zmq.utils import jsonapi
     if jsonapi.jsonmod.__name__ == 'jsonlib':
         import json
         jsonapi.jsonmod = json
 
 
 def check_for_zmq(minimum_version, module='IPython.zmq'):
     try:
         import zmq
     except ImportError:
         raise ImportError("%s requires pyzmq >= %s"%(module, minimum_version))
 
     pyzmq_version = zmq.__version__
     
     if 'dev' not in pyzmq_version and V(pyzmq_version) < V(minimum_version):
         raise ImportError("%s requires pyzmq >= %s, but you have %s"%(
                         module, minimum_version, pyzmq_version))
 
     if V(zmq.zmq_version()) >= V('4.0.0'):
         warnings.warn("""libzmq 4 detected.
         It is unlikely that IPython's zmq code will work properly.
         Please install libzmq stable, which is 2.1.x or 2.2.x""",
         RuntimeWarning)
 
 check_for_zmq('2.1.4')
 patch_pyzmq()

IPython/zmq/entry_point.py

@@ -1,208 +1,208 b''
 """ Defines helper functions for creating kernel entry points and process
 launchers.
 """
 
 # Standard library imports.
 import atexit
 import json
 import os
 import socket
 from subprocess import Popen, PIPE
 import sys
 import tempfile
 
 # System library imports
 
 # IPython imports
 from IPython.utils.localinterfaces import LOCALHOST
 from IPython.utils.py3compat import bytes_to_str
 
 # Local imports.
 from parentpoller import ParentPollerWindows
 
 def write_connection_file(fname=None, shell_port=0, iopub_port=0, stdin_port=0, hb_port=0,
                          ip=LOCALHOST, key=b''):
     """Generates a JSON config file, including the selection of random ports.
     
     Parameters
     ----------
 
     fname : unicode
         The path to the file to write
 
     shell_port : int, optional
-        The port to use for XREP channel.
+        The port to use for ROUTER channel.
 
     iopub_port : int, optional
         The port to use for the SUB channel.
 
     stdin_port : int, optional
         The port to use for the REQ (raw input) channel.
 
     hb_port : int, optional
         The port to use for the hearbeat REP channel.
 
     ip  : str, optional
         The ip address the kernel will bind to.
 
     key : str, optional
         The Session key used for HMAC authentication.
 
     """
     # default to temporary connector file
     if not fname:
         fname = tempfile.mktemp('.json')
     
     # Find open ports as necessary.
     ports = []
     ports_needed = int(shell_port <= 0) + int(iopub_port <= 0) + \
                    int(stdin_port <= 0) + int(hb_port <= 0)
     for i in xrange(ports_needed):
         sock = socket.socket()
         sock.bind(('', 0))
         ports.append(sock)
     for i, sock in enumerate(ports):
         port = sock.getsockname()[1]
         sock.close()
         ports[i] = port
     if shell_port <= 0:
         shell_port = ports.pop(0)
     if iopub_port <= 0:
         iopub_port = ports.pop(0)
     if stdin_port <= 0:
         stdin_port = ports.pop(0)
     if hb_port <= 0:
         hb_port = ports.pop(0)
     
     cfg = dict( shell_port=shell_port,
                 iopub_port=iopub_port,
                 stdin_port=stdin_port,
                 hb_port=hb_port,
               )
     cfg['ip'] = ip
     cfg['key'] = bytes_to_str(key)
     
     with open(fname, 'w') as f:
         f.write(json.dumps(cfg, indent=2))
     
     return fname, cfg
     
 
 def base_launch_kernel(code, fname, stdin=None, stdout=None, stderr=None,
                         executable=None, independent=False, extra_arguments=[]):
     """ Launches a localhost kernel, binding to the specified ports.
 
     Parameters
     ----------
     code : str,
         A string of Python code that imports and executes a kernel entry point.
 
     stdin, stdout, stderr : optional (default None)
         Standards streams, as defined in subprocess.Popen.
 
     fname : unicode, optional
         The JSON connector file, containing ip/port/hmac key information.
 
     key : str, optional
         The Session key used for HMAC authentication.
 
     executable : str, optional (default sys.executable)
         The Python executable to use for the kernel process.
 
     independent : bool, optional (default False)
         If set, the kernel process is guaranteed to survive if this process
         dies. If not set, an effort is made to ensure that the kernel is killed
         when this process dies. Note that in this case it is still good practice
         to kill kernels manually before exiting.
 
     extra_arguments = list, optional
         A list of extra arguments to pass when executing the launch code.
 
     Returns
     -------
     A tuple of form:
         (kernel_process, shell_port, iopub_port, stdin_port, hb_port)
     where kernel_process is a Popen object and the ports are integers.
     """
     
     # Build the kernel launch command.
     if executable is None:
         executable = sys.executable
     arguments = [ executable, '-c', code, '-f', fname ]
     arguments.extend(extra_arguments)
 
     # Popen will fail (sometimes with a deadlock) if stdin, stdout, and stderr
     # are invalid. Unfortunately, there is in general no way to detect whether
     # they are valid.  The following two blocks redirect them to (temporary)
     # pipes in certain important cases.
 
     # If this process has been backgrounded, our stdin is invalid. Since there
     # is no compelling reason for the kernel to inherit our stdin anyway, we'll
     # place this one safe and always redirect.
     redirect_in = True
     _stdin = PIPE if stdin is None else stdin
 
     # If this process in running on pythonw, we know that stdin, stdout, and
     # stderr are all invalid.
     redirect_out = sys.executable.endswith('pythonw.exe')
     if redirect_out:
         _stdout = PIPE if stdout is None else stdout
         _stderr = PIPE if stderr is None else stderr
     else:
         _stdout, _stderr = stdout, stderr
 
     # Spawn a kernel.
     if sys.platform == 'win32':
         # Create a Win32 event for interrupting the kernel.
         interrupt_event = ParentPollerWindows.create_interrupt_event()
         arguments += [ '--interrupt=%i'%interrupt_event ]
 
         # If the kernel is running on pythonw and stdout/stderr are not been
         # re-directed, it will crash when more than 4KB of data is written to
         # stdout or stderr. This is a bug that has been with Python for a very
         # long time; see http://bugs.python.org/issue706263.
         # A cleaner solution to this problem would be to pass os.devnull to
         # Popen directly. Unfortunately, that does not work.
         if executable.endswith('pythonw.exe'):
             if stdout is None:
                 arguments.append('--no-stdout')
             if stderr is None:
                 arguments.append('--no-stderr')
 
         # Launch the kernel process.
         if independent:
             proc = Popen(arguments,
                          creationflags=512, # CREATE_NEW_PROCESS_GROUP
                          stdin=_stdin, stdout=_stdout, stderr=_stderr)
         else:
             from _subprocess import DuplicateHandle, GetCurrentProcess, \
                 DUPLICATE_SAME_ACCESS
             pid = GetCurrentProcess()
             handle = DuplicateHandle(pid, pid, pid, 0,
                                      True, # Inheritable by new processes.
                                      DUPLICATE_SAME_ACCESS)
             proc = Popen(arguments + ['--parent=%i'%int(handle)],
                          stdin=_stdin, stdout=_stdout, stderr=_stderr)
 
         # Attach the interrupt event to the Popen objet so it can be used later.
         proc.win32_interrupt_event = interrupt_event
 
     else:
         if independent:
             proc = Popen(arguments, preexec_fn=lambda: os.setsid(),
                          stdin=_stdin, stdout=_stdout, stderr=_stderr)
         else:
             proc = Popen(arguments + ['--parent=1'],
                          stdin=_stdin, stdout=_stdout, stderr=_stderr)
 
     # Clean up pipes created to work around Popen bug.
     if redirect_in:
         if stdin is None:
             proc.stdin.close()
     if redirect_out:
         if stdout is None:
             proc.stdout.close()
         if stderr is None:
             proc.stderr.close()
 
     return proc

IPython/zmq/kernelapp.py

@@ -1,333 +1,333 b''
 """An Application for launching a kernel
 
 Authors
 -------
 * MinRK
 """
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING.txt, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Imports
 #-----------------------------------------------------------------------------
 
 # Standard library imports
 import atexit
 import json
 import os
 import sys
 import signal
 
 # System library imports
 import zmq
 from zmq.eventloop import ioloop
 
 # IPython imports
 from IPython.core.ultratb import FormattedTB
 from IPython.core.application import (
     BaseIPythonApplication, base_flags, base_aliases, catch_config_error
 )
 from IPython.utils import io
 from IPython.utils.localinterfaces import LOCALHOST
 from IPython.utils.path import filefind
 from IPython.utils.py3compat import str_to_bytes
 from IPython.utils.traitlets import (Any, Instance, Dict, Unicode, Integer, Bool,
                                         DottedObjectName)
 from IPython.utils.importstring import import_item
 # local imports
 from IPython.zmq.entry_point import write_connection_file
 from IPython.zmq.heartbeat import Heartbeat
 from IPython.zmq.parentpoller import ParentPollerUnix, ParentPollerWindows
 from IPython.zmq.session import (
     Session, session_flags, session_aliases, default_secure,
 )
 
 
 #-----------------------------------------------------------------------------
 # Flags and Aliases
 #-----------------------------------------------------------------------------
 
 kernel_aliases = dict(base_aliases)
 kernel_aliases.update({
     'ip' : 'KernelApp.ip',
     'hb' : 'KernelApp.hb_port',
     'shell' : 'KernelApp.shell_port',
     'iopub' : 'KernelApp.iopub_port',
     'stdin' : 'KernelApp.stdin_port',
     'f' : 'KernelApp.connection_file',
     'parent': 'KernelApp.parent',
 })
 if sys.platform.startswith('win'):
     kernel_aliases['interrupt'] = 'KernelApp.interrupt'
 
 kernel_flags = dict(base_flags)
 kernel_flags.update({
     'no-stdout' : (
             {'KernelApp' : {'no_stdout' : True}},
             "redirect stdout to the null device"),
     'no-stderr' : (
             {'KernelApp' : {'no_stderr' : True}},
             "redirect stderr to the null device"),
 })
 
 # inherit flags&aliases for Sessions
 kernel_aliases.update(session_aliases)
 kernel_flags.update(session_flags)
 
 
 
 #-----------------------------------------------------------------------------
 # Application class for starting a Kernel
 #-----------------------------------------------------------------------------
 
 class KernelApp(BaseIPythonApplication):
     name='ipkernel'
     aliases = Dict(kernel_aliases)
     flags = Dict(kernel_flags)
     classes = [Session]
     # the kernel class, as an importstring
     kernel_class = DottedObjectName('IPython.zmq.ipkernel.Kernel')
     kernel = Any()
     poller = Any() # don't restrict this even though current pollers are all Threads
     heartbeat = Instance(Heartbeat)
     session = Instance('IPython.zmq.session.Session')
     ports = Dict()
     _full_connection_file = Unicode()
     
     # inherit config file name from parent:
     parent_appname = Unicode(config=True)
     def _parent_appname_changed(self, name, old, new):
         if self.config_file_specified:
             # it was manually specified, ignore
             return
         self.config_file_name = new.replace('-','_') + u'_config.py'
         # don't let this count as specifying the config file
         self.config_file_specified = False
         
     # connection info:
     ip = Unicode(LOCALHOST, config=True,
         help="Set the IP or interface on which the kernel will listen.")
     hb_port = Integer(0, config=True, help="set the heartbeat port [default: random]")
-    shell_port = Integer(0, config=True, help="set the shell (XREP) port [default: random]")
+    shell_port = Integer(0, config=True, help="set the shell (ROUTER) port [default: random]")
     iopub_port = Integer(0, config=True, help="set the iopub (PUB) port [default: random]")
-    stdin_port = Integer(0, config=True, help="set the stdin (XREQ) port [default: random]")
+    stdin_port = Integer(0, config=True, help="set the stdin (DEALER) port [default: random]")
     connection_file = Unicode('', config=True, 
     help="""JSON file in which to store connection info [default: kernel-<pid>.json]
     
     This file will contain the IP, ports, and authentication key needed to connect
     clients to this kernel. By default, this file will be created in the security-dir
     of the current profile, but can be specified by absolute path.
     """)
 
     # streams, etc.
     no_stdout = Bool(False, config=True, help="redirect stdout to the null device")
     no_stderr = Bool(False, config=True, help="redirect stderr to the null device")
     outstream_class = DottedObjectName('IPython.zmq.iostream.OutStream',
         config=True, help="The importstring for the OutStream factory")
     displayhook_class = DottedObjectName('IPython.zmq.displayhook.ZMQDisplayHook',
         config=True, help="The importstring for the DisplayHook factory")
 
     # polling
     parent = Integer(0, config=True,
         help="""kill this process if its parent dies.  On Windows, the argument
         specifies the HANDLE of the parent process, otherwise it is simply boolean.
         """)
     interrupt = Integer(0, config=True,
         help="""ONLY USED ON WINDOWS
         Interrupt this process when the parent is signalled.
         """)
 
     def init_crash_handler(self):
         # Install minimal exception handling
         sys.excepthook = FormattedTB(mode='Verbose', color_scheme='NoColor',
                                      ostream=sys.__stdout__)
 
     def init_poller(self):
         if sys.platform == 'win32':
             if self.interrupt or self.parent:
                 self.poller = ParentPollerWindows(self.interrupt, self.parent)
         elif self.parent:
             self.poller = ParentPollerUnix()
 
     def _bind_socket(self, s, port):
         iface = 'tcp://%s' % self.ip
         if port <= 0:
             port = s.bind_to_random_port(iface)
         else:
             s.bind(iface + ':%i'%port)
         return port
 
     def load_connection_file(self):
         """load ip/port/hmac config from JSON connection file"""
         try:
             fname = filefind(self.connection_file, ['.', self.profile_dir.security_dir])
         except IOError:
             self.log.debug("Connection file not found: %s", self.connection_file)
             # This means I own it, so I will clean it up:
             atexit.register(self.cleanup_connection_file)
             return
         self.log.debug(u"Loading connection file %s", fname)
         with open(fname) as f:
             s = f.read()
         cfg = json.loads(s)
         if self.ip == LOCALHOST and 'ip' in cfg:
             # not overridden by config or cl_args
             self.ip = cfg['ip']
         for channel in ('hb', 'shell', 'iopub', 'stdin'):
             name = channel + '_port'
             if getattr(self, name) == 0 and name in cfg:
                 # not overridden by config or cl_args
                 setattr(self, name, cfg[name])
         if 'key' in cfg:
             self.config.Session.key = str_to_bytes(cfg['key'])
     
     def write_connection_file(self):
         """write connection info to JSON file"""
         if os.path.basename(self.connection_file) == self.connection_file:
             cf = os.path.join(self.profile_dir.security_dir, self.connection_file)
         else:
             cf = self.connection_file
         write_connection_file(cf, ip=self.ip, key=self.session.key,
         shell_port=self.shell_port, stdin_port=self.stdin_port, hb_port=self.hb_port,
         iopub_port=self.iopub_port)
         
         self._full_connection_file = cf
     
     def cleanup_connection_file(self):
         cf = self._full_connection_file
         self.log.debug("cleaning up connection file: %r", cf)
         try:
             os.remove(cf)
         except (IOError, OSError):
             pass
     
     def init_connection_file(self):
         if not self.connection_file:
             self.connection_file = "kernel-%s.json"%os.getpid()
         try:
             self.load_connection_file()
         except Exception:
             self.log.error("Failed to load connection file: %r", self.connection_file, exc_info=True)
             self.exit(1)
     
     def init_sockets(self):
         # Create a context, a session, and the kernel sockets.
         self.log.info("Starting the kernel at pid: %i", os.getpid())
         context = zmq.Context.instance()
         # Uncomment this to try closing the context.
         # atexit.register(context.term)
 
         self.shell_socket = context.socket(zmq.ROUTER)
         self.shell_port = self._bind_socket(self.shell_socket, self.shell_port)
         self.log.debug("shell ROUTER Channel on port: %i"%self.shell_port)
 
         self.iopub_socket = context.socket(zmq.PUB)
         self.iopub_port = self._bind_socket(self.iopub_socket, self.iopub_port)
         self.log.debug("iopub PUB Channel on port: %i"%self.iopub_port)
 
         self.stdin_socket = context.socket(zmq.ROUTER)
         self.stdin_port = self._bind_socket(self.stdin_socket, self.stdin_port)
         self.log.debug("stdin ROUTER Channel on port: %i"%self.stdin_port)
     
     def init_heartbeat(self):
         """start the heart beating"""
         # heartbeat doesn't share context, because it mustn't be blocked
         # by the GIL, which is accessed by libzmq when freeing zero-copy messages
         hb_ctx = zmq.Context()
         self.heartbeat = Heartbeat(hb_ctx, (self.ip, self.hb_port))
         self.hb_port = self.heartbeat.port
         self.log.debug("Heartbeat REP Channel on port: %i"%self.hb_port)
         self.heartbeat.start()
 
         # Helper to make it easier to connect to an existing kernel.
         # set log-level to critical, to make sure it is output
         self.log.critical("To connect another client to this kernel, use:")
     
     def log_connection_info(self):
         """display connection info, and store ports"""
         basename = os.path.basename(self.connection_file)
         if basename == self.connection_file or \
             os.path.dirname(self.connection_file) == self.profile_dir.security_dir:
             # use shortname
             tail = basename
             if self.profile != 'default':
                 tail += " --profile %s" % self.profile
         else:
             tail = self.connection_file
         self.log.critical("--existing %s", tail)
 
 
         self.ports = dict(shell=self.shell_port, iopub=self.iopub_port,
                                 stdin=self.stdin_port, hb=self.hb_port)
 
     def init_session(self):
         """create our session object"""
         default_secure(self.config)
         self.session = Session(config=self.config, username=u'kernel')
 
     def init_blackhole(self):
         """redirects stdout/stderr to devnull if necessary"""
         if self.no_stdout or self.no_stderr:
             blackhole = open(os.devnull, 'w')
             if self.no_stdout:
                 sys.stdout = sys.__stdout__ = blackhole
             if self.no_stderr:
                 sys.stderr = sys.__stderr__ = blackhole
     
     def init_io(self):
         """Redirect input streams and set a display hook."""
         if self.outstream_class:
             outstream_factory = import_item(str(self.outstream_class))
             sys.stdout = outstream_factory(self.session, self.iopub_socket, u'stdout')
             sys.stderr = outstream_factory(self.session, self.iopub_socket, u'stderr')
         if self.displayhook_class:
             displayhook_factory = import_item(str(self.displayhook_class))
             sys.displayhook = displayhook_factory(self.session, self.iopub_socket)
 
     def init_signal(self):
         signal.signal(signal.SIGINT, signal.SIG_IGN)
 
     def init_kernel(self):
         """Create the Kernel object itself"""
         kernel_factory = import_item(str(self.kernel_class))
         self.kernel = kernel_factory(config=self.config, session=self.session,
                                 shell_socket=self.shell_socket,
                                 iopub_socket=self.iopub_socket,
                                 stdin_socket=self.stdin_socket,
                                 log=self.log
         )
         self.kernel.record_ports(self.ports)
 
     @catch_config_error
     def initialize(self, argv=None):
         super(KernelApp, self).initialize(argv)
         self.init_blackhole()
         self.init_connection_file()
         self.init_session()
         self.init_poller()
         self.init_sockets()
         self.init_heartbeat()
         # writing/displaying connection info must be *after* init_sockets/heartbeat
         self.log_connection_info()
         self.write_connection_file()
         self.init_io()
         self.init_signal()
         self.init_kernel()
         # flush stdout/stderr, so that anything written to these streams during
         # initialization do not get associated with the first execution request
         sys.stdout.flush()
         sys.stderr.flush()
 
     def start(self):
         if self.poller is not None:
             self.poller.start()
         self.kernel.start()
         try:
             ioloop.IOLoop.instance().start()
         except KeyboardInterrupt:
             pass
 

IPython/zmq/kernelmanager.py

@@ -1,994 +1,994 b''
 """Base classes to manage the interaction with a running kernel.
 
 TODO
 * Create logger to handle debugging and console messages.
 """
 
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2008-2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Imports
 #-----------------------------------------------------------------------------
 
 # Standard library imports.
 import atexit
 import errno
 import json
 from subprocess import Popen
 import os
 import signal
 import sys
 from threading import Thread
 import time
 
 # System library imports.
 import zmq
 # import ZMQError in top-level namespace, to avoid ugly attribute-error messages
 # during garbage collection of threads at exit:
 from zmq import ZMQError
 from zmq.eventloop import ioloop, zmqstream
 
 # Local imports.
 from IPython.config.loader import Config
 from IPython.utils.localinterfaces import LOCALHOST, LOCAL_IPS
 from IPython.utils.traitlets import (
     HasTraits, Any, Instance, Type, Unicode, Integer, Bool
 )
 from IPython.utils.py3compat import str_to_bytes
 from IPython.zmq.entry_point import write_connection_file
 from session import Session
 
 #-----------------------------------------------------------------------------
 # Constants and exceptions
 #-----------------------------------------------------------------------------
 
 class InvalidPortNumber(Exception):
     pass
 
 #-----------------------------------------------------------------------------
 # Utility functions
 #-----------------------------------------------------------------------------
 
 # some utilities to validate message structure, these might get moved elsewhere
 # if they prove to have more generic utility
 
 def validate_string_list(lst):
     """Validate that the input is a list of strings.
 
     Raises ValueError if not."""
     if not isinstance(lst, list):
         raise ValueError('input %r must be a list' % lst)
     for x in lst:
         if not isinstance(x, basestring):
             raise ValueError('element %r in list must be a string' % x)
 
 
 def validate_string_dict(dct):
     """Validate that the input is a dict with string keys and values.
 
     Raises ValueError if not."""
     for k,v in dct.iteritems():
         if not isinstance(k, basestring):
             raise ValueError('key %r in dict must be a string' % k)
         if not isinstance(v, basestring):
             raise ValueError('value %r in dict must be a string' % v)
 
 
 #-----------------------------------------------------------------------------
 # ZMQ Socket Channel classes
 #-----------------------------------------------------------------------------
 
 class ZMQSocketChannel(Thread):
     """The base class for the channels that use ZMQ sockets.
     """
     context = None
     session = None
     socket = None
     ioloop = None
     stream = None
     _address = None
     _exiting = False
 
     def __init__(self, context, session, address):
         """Create a channel
 
         Parameters
         ----------
         context : :class:`zmq.Context`
             The ZMQ context to use.
         session : :class:`session.Session`
             The session to use.
         address : tuple
             Standard (ip, port) tuple that the kernel is listening on.
         """
         super(ZMQSocketChannel, self).__init__()
         self.daemon = True
 
         self.context = context
         self.session = session
         if address[1] == 0:
             message = 'The port number for a channel cannot be 0.'
             raise InvalidPortNumber(message)
         self._address = address
         atexit.register(self._notice_exit)
     
     def _notice_exit(self):
         self._exiting = True
 
     def _run_loop(self):
         """Run my loop, ignoring EINTR events in the poller"""
         while True:
             try:
                 self.ioloop.start()
             except ZMQError as e:
                 if e.errno == errno.EINTR:
                     continue
                 else:
                     raise
             except Exception:
                 if self._exiting:
                     break
                 else:
                     raise
             else:
                 break
 
     def stop(self):
         """Stop the channel's activity.
 
         This calls :method:`Thread.join` and returns when the thread
         terminates. :class:`RuntimeError` will be raised if
         :method:`self.start` is called again.
         """
         self.join()
 
     @property
     def address(self):
         """Get the channel's address as an (ip, port) tuple.
 
         By the default, the address is (localhost, 0), where 0 means a random
         port.
         """
         return self._address
 
     def _queue_send(self, msg):
         """Queue a message to be sent from the IOLoop's thread.
         
         Parameters
         ----------
         msg : message to send
         
         This is threadsafe, as it uses IOLoop.add_callback to give the loop's
         thread control of the action.
         """
         def thread_send():
             self.session.send(self.stream, msg)
         self.ioloop.add_callback(thread_send)
 
     def _handle_recv(self, msg):
         """callback for stream.on_recv
         
         unpacks message, and calls handlers with it.
         """
         ident,smsg = self.session.feed_identities(msg)
         self.call_handlers(self.session.unserialize(smsg))
     
 
 
 class ShellSocketChannel(ZMQSocketChannel):
-    """The XREQ channel for issues request/replies to the kernel.
+    """The DEALER channel for issues request/replies to the kernel.
     """
 
     command_queue = None
     # flag for whether execute requests should be allowed to call raw_input:
     allow_stdin = True
 
     def __init__(self, context, session, address):
         super(ShellSocketChannel, self).__init__(context, session, address)
         self.ioloop = ioloop.IOLoop()
 
     def run(self):
         """The thread's main activity.  Call start() instead."""
         self.socket = self.context.socket(zmq.DEALER)
         self.socket.setsockopt(zmq.IDENTITY, self.session.bsession)
         self.socket.connect('tcp://%s:%i' % self.address)
         self.stream = zmqstream.ZMQStream(self.socket, self.ioloop)
         self.stream.on_recv(self._handle_recv)
         self._run_loop()
         try:
             self.socket.close()
         except:
             pass
 
     def stop(self):
         self.ioloop.stop()
         super(ShellSocketChannel, self).stop()
 
     def call_handlers(self, msg):
         """This method is called in the ioloop thread when a message arrives.
 
         Subclasses should override this method to handle incoming messages.
         It is important to remember that this method is called in the thread
         so that some logic must be done to ensure that the application leve
         handlers are called in the application thread.
         """
         raise NotImplementedError('call_handlers must be defined in a subclass.')
 
     def execute(self, code, silent=False,
                 user_variables=None, user_expressions=None, allow_stdin=None):
         """Execute code in the kernel.
 
         Parameters
         ----------
         code : str
             A string of Python code.
 
         silent : bool, optional (default False)
             If set, the kernel will execute the code as quietly possible.
 
         user_variables : list, optional
             A list of variable names to pull from the user's namespace.  They
             will come back as a dict with these names as keys and their
             :func:`repr` as values.
 
         user_expressions : dict, optional
             A dict with string keys and  to pull from the user's
             namespace.  They will come back as a dict with these names as keys
             and their :func:`repr` as values.
 
         allow_stdin : bool, optional
             Flag for 
             A dict with string keys and  to pull from the user's
             namespace.  They will come back as a dict with these names as keys
             and their :func:`repr` as values.
 
         Returns
         -------
         The msg_id of the message sent.
         """
         if user_variables is None:
             user_variables = []
         if user_expressions is None:
             user_expressions = {}
         if allow_stdin is None:
             allow_stdin = self.allow_stdin
         
         
         # Don't waste network traffic if inputs are invalid
         if not isinstance(code, basestring):
             raise ValueError('code %r must be a string' % code)
         validate_string_list(user_variables)
         validate_string_dict(user_expressions)
 
         # Create class for content/msg creation. Related to, but possibly
         # not in Session.
         content = dict(code=code, silent=silent,
                        user_variables=user_variables,
                        user_expressions=user_expressions,
                        allow_stdin=allow_stdin,
                        )
         msg = self.session.msg('execute_request', content)
         self._queue_send(msg)
         return msg['header']['msg_id']
 
     def complete(self, text, line, cursor_pos, block=None):
         """Tab complete text in the kernel's namespace.
 
         Parameters
         ----------
         text : str
             The text to complete.
         line : str
             The full line of text that is the surrounding context for the
             text to complete.
         cursor_pos : int
             The position of the cursor in the line where the completion was
             requested.
         block : str, optional
             The full block of code in which the completion is being requested.
 
         Returns
         -------
         The msg_id of the message sent.
         """
         content = dict(text=text, line=line, block=block, cursor_pos=cursor_pos)
         msg = self.session.msg('complete_request', content)
         self._queue_send(msg)
         return msg['header']['msg_id']
 
     def object_info(self, oname, detail_level=0):
         """Get metadata information about an object.
 
         Parameters
         ----------
         oname : str
             A string specifying the object name.
         detail_level : int, optional
             The level of detail for the introspection (0-2)
 
         Returns
         -------
         The msg_id of the message sent.
         """
         content = dict(oname=oname, detail_level=detail_level)
         msg = self.session.msg('object_info_request', content)
         self._queue_send(msg)
         return msg['header']['msg_id']
 
     def history(self, raw=True, output=False, hist_access_type='range', **kwargs):
         """Get entries from the history list.
 
         Parameters
         ----------
         raw : bool
             If True, return the raw input.
         output : bool
             If True, then return the output as well.
         hist_access_type : str
             'range' (fill in session, start and stop params), 'tail' (fill in n)
              or 'search' (fill in pattern param).
 
         session : int
             For a range request, the session from which to get lines. Session
             numbers are positive integers; negative ones count back from the
             current session.
         start : int
             The first line number of a history range.
         stop : int
             The final (excluded) line number of a history range.
 
         n : int
             The number of lines of history to get for a tail request.
 
         pattern : str
             The glob-syntax pattern for a search request.
 
         Returns
         -------
         The msg_id of the message sent.
         """
         content = dict(raw=raw, output=output, hist_access_type=hist_access_type,
                                                                     **kwargs)
         msg = self.session.msg('history_request', content)
         self._queue_send(msg)
         return msg['header']['msg_id']
 
     def shutdown(self, restart=False):
         """Request an immediate kernel shutdown.
 
         Upon receipt of the (empty) reply, client code can safely assume that
         the kernel has shut down and it's safe to forcefully terminate it if
         it's still alive.
 
         The kernel will send the reply via a function registered with Python's
         atexit module, ensuring it's truly done as the kernel is done with all
         normal operation.
         """
         # Send quit message to kernel. Once we implement kernel-side setattr,
         # this should probably be done that way, but for now this will do.
         msg = self.session.msg('shutdown_request', {'restart':restart})
         self._queue_send(msg)
         return msg['header']['msg_id']
 
 
 
 class SubSocketChannel(ZMQSocketChannel):
     """The SUB channel which listens for messages that the kernel publishes.
     """
 
     def __init__(self, context, session, address):
         super(SubSocketChannel, self).__init__(context, session, address)
         self.ioloop = ioloop.IOLoop()
 
     def run(self):
         """The thread's main activity.  Call start() instead."""
         self.socket = self.context.socket(zmq.SUB)
         self.socket.setsockopt(zmq.SUBSCRIBE,b'')
         self.socket.setsockopt(zmq.IDENTITY, self.session.bsession)
         self.socket.connect('tcp://%s:%i' % self.address)
         self.stream = zmqstream.ZMQStream(self.socket, self.ioloop)
         self.stream.on_recv(self._handle_recv)
         self._run_loop()
         try:
             self.socket.close()
         except:
             pass
 
     def stop(self):
         self.ioloop.stop()
         super(SubSocketChannel, self).stop()
 
     def call_handlers(self, msg):
         """This method is called in the ioloop thread when a message arrives.
 
         Subclasses should override this method to handle incoming messages.
         It is important to remember that this method is called in the thread
         so that some logic must be done to ensure that the application leve
         handlers are called in the application thread.
         """
         raise NotImplementedError('call_handlers must be defined in a subclass.')
 
     def flush(self, timeout=1.0):
         """Immediately processes all pending messages on the SUB channel.
 
         Callers should use this method to ensure that :method:`call_handlers`
         has been called for all messages that have been received on the
         0MQ SUB socket of this channel.
 
         This method is thread safe.
 
         Parameters
         ----------
         timeout : float, optional
             The maximum amount of time to spend flushing, in seconds. The
             default is one second.
         """
         # We do the IOLoop callback process twice to ensure that the IOLoop
         # gets to perform at least one full poll.
         stop_time = time.time() + timeout
         for i in xrange(2):
             self._flushed = False
             self.ioloop.add_callback(self._flush)
             while not self._flushed and time.time() < stop_time:
                 time.sleep(0.01)
 
     def _flush(self):
         """Callback for :method:`self.flush`."""
         self.stream.flush()
         self._flushed = True
 
 
 class StdInSocketChannel(ZMQSocketChannel):
     """A reply channel to handle raw_input requests that the kernel makes."""
 
     msg_queue = None
 
     def __init__(self, context, session, address):
         super(StdInSocketChannel, self).__init__(context, session, address)
         self.ioloop = ioloop.IOLoop()
 
     def run(self):
         """The thread's main activity.  Call start() instead."""
         self.socket = self.context.socket(zmq.DEALER)
         self.socket.setsockopt(zmq.IDENTITY, self.session.bsession)
         self.socket.connect('tcp://%s:%i' % self.address)
         self.stream = zmqstream.ZMQStream(self.socket, self.ioloop)
         self.stream.on_recv(self._handle_recv)
         self._run_loop()
         try:
             self.socket.close()
         except:
             pass
         
 
     def stop(self):
         self.ioloop.stop()
         super(StdInSocketChannel, self).stop()
 
     def call_handlers(self, msg):
         """This method is called in the ioloop thread when a message arrives.
 
         Subclasses should override this method to handle incoming messages.
         It is important to remember that this method is called in the thread
         so that some logic must be done to ensure that the application leve
         handlers are called in the application thread.
         """
         raise NotImplementedError('call_handlers must be defined in a subclass.')
 
     def input(self, string):
         """Send a string of raw input to the kernel."""
         content = dict(value=string)
         msg = self.session.msg('input_reply', content)
         self._queue_send(msg)
 
 
 class HBSocketChannel(ZMQSocketChannel):
     """The heartbeat channel which monitors the kernel heartbeat.
 
     Note that the heartbeat channel is paused by default. As long as you start
     this channel, the kernel manager will ensure that it is paused and un-paused
     as appropriate.
     """
 
     time_to_dead = 3.0
     socket = None
     poller = None
     _running = None
     _pause = None
     _beating = None
 
     def __init__(self, context, session, address):
         super(HBSocketChannel, self).__init__(context, session, address)
         self._running = False
         self._pause =True
         self.poller = zmq.Poller()
 
     def _create_socket(self):
         if self.socket is not None:
             # close previous socket, before opening a new one
             self.poller.unregister(self.socket)
             self.socket.close()
         self.socket = self.context.socket(zmq.REQ)
         self.socket.setsockopt(zmq.LINGER, 0)
         self.socket.connect('tcp://%s:%i' % self.address)
         
         self.poller.register(self.socket, zmq.POLLIN)
     
     def _poll(self, start_time):
         """poll for heartbeat replies until we reach self.time_to_dead
         
         Ignores interrupts, and returns the result of poll(), which
         will be an empty list if no messages arrived before the timeout,
         or the event tuple if there is a message to receive.
         """
         
         until_dead = self.time_to_dead - (time.time() - start_time)
         # ensure poll at least once
         until_dead = max(until_dead, 1e-3)
         events = []
         while True:
             try:
                 events = self.poller.poll(1000 * until_dead)
             except ZMQError as e:
                 if e.errno == errno.EINTR:
                     # ignore interrupts during heartbeat
                     # this may never actually happen
                     until_dead = self.time_to_dead - (time.time() - start_time)
                     until_dead = max(until_dead, 1e-3)
                     pass
                 else:
                     raise
             except Exception:
                 if self._exiting:
                     break
                 else:
                     raise
             else:
                 break
         return events
 
     def run(self):
         """The thread's main activity.  Call start() instead."""
         self._create_socket()
         self._running = True
         self._beating = True
         
         while self._running:
             if self._pause:
                 # just sleep, and skip the rest of the loop
                 time.sleep(self.time_to_dead)
                 continue
             
             since_last_heartbeat = 0.0
             # io.rprint('Ping from HB channel') # dbg
             # no need to catch EFSM here, because the previous event was
             # either a recv or connect, which cannot be followed by EFSM
             self.socket.send(b'ping')
             request_time = time.time()
             ready = self._poll(request_time)
             if ready:
                 self._beating = True
                 # the poll above guarantees we have something to recv
                 self.socket.recv()
                 # sleep the remainder of the cycle
                 remainder = self.time_to_dead - (time.time() - request_time)
                 if remainder > 0:
                     time.sleep(remainder)
                 continue
             else:
                 # nothing was received within the time limit, signal heart failure
                 self._beating = False
                 since_last_heartbeat = time.time() - request_time
                 self.call_handlers(since_last_heartbeat)
                 # and close/reopen the socket, because the REQ/REP cycle has been broken
                 self._create_socket()
                 continue
         try:
             self.socket.close()
         except:
             pass
 
     def pause(self):
         """Pause the heartbeat."""
         self._pause = True
 
     def unpause(self):
         """Unpause the heartbeat."""
         self._pause = False
 
     def is_beating(self):
         """Is the heartbeat running and responsive (and not paused)."""
         if self.is_alive() and not self._pause and self._beating:
             return True
         else:
             return False
 
     def stop(self):
         self._running = False
         super(HBSocketChannel, self).stop()
 
     def call_handlers(self, since_last_heartbeat):
         """This method is called in the ioloop thread when a message arrives.
 
         Subclasses should override this method to handle incoming messages.
         It is important to remember that this method is called in the thread
         so that some logic must be done to ensure that the application level
         handlers are called in the application thread.
         """
         raise NotImplementedError('call_handlers must be defined in a subclass.')
 
 
 #-----------------------------------------------------------------------------
 # Main kernel manager class
 #-----------------------------------------------------------------------------
 
 class KernelManager(HasTraits):
     """ Manages a kernel for a frontend.
 
     The SUB channel is for the frontend to receive messages published by the
     kernel.
 
     The REQ channel is for the frontend to make requests of the kernel.
 
     The REP channel is for the kernel to request stdin (raw_input) from the
     frontend.
     """
     # config object for passing to child configurables
     config = Instance(Config)
 
     # The PyZMQ Context to use for communication with the kernel.
     context = Instance(zmq.Context)
     def _context_default(self):
         return zmq.Context.instance()
 
     # The Session to use for communication with the kernel.
     session = Instance(Session)
 
     # The kernel process with which the KernelManager is communicating.
     kernel = Instance(Popen)
 
     # The addresses for the communication channels.
     connection_file = Unicode('')
     ip = Unicode(LOCALHOST)
     def _ip_changed(self, name, old, new):
         if new == '*':
             self.ip = '0.0.0.0'
     shell_port = Integer(0)
     iopub_port = Integer(0)
     stdin_port = Integer(0)
     hb_port = Integer(0)
 
     # The classes to use for the various channels.
     shell_channel_class = Type(ShellSocketChannel)
     sub_channel_class = Type(SubSocketChannel)
     stdin_channel_class = Type(StdInSocketChannel)
     hb_channel_class = Type(HBSocketChannel)
 
     # Protected traits.
     _launch_args = Any
     _shell_channel = Any
     _sub_channel = Any
     _stdin_channel = Any
     _hb_channel = Any
     _connection_file_written=Bool(False)
 
     def __init__(self, **kwargs):
         super(KernelManager, self).__init__(**kwargs)
         if self.session is None:
             self.session = Session(config=self.config)
     
     def __del__(self):
         self.cleanup_connection_file()
     
 
     #--------------------------------------------------------------------------
     # Channel management methods:
     #--------------------------------------------------------------------------
 
     def start_channels(self, shell=True, sub=True, stdin=True, hb=True):
         """Starts the channels for this kernel.
 
         This will create the channels if they do not exist and then start
         them. If port numbers of 0 are being used (random ports) then you
         must first call :method:`start_kernel`. If the channels have been
         stopped and you call this, :class:`RuntimeError` will be raised.
         """
         if shell:
             self.shell_channel.start()
         if sub:
             self.sub_channel.start()
         if stdin:
             self.stdin_channel.start()
             self.shell_channel.allow_stdin = True
         else:
             self.shell_channel.allow_stdin = False
         if hb:
             self.hb_channel.start()
 
     def stop_channels(self):
         """Stops all the running channels for this kernel.
         """
         if self.shell_channel.is_alive():
             self.shell_channel.stop()
         if self.sub_channel.is_alive():
             self.sub_channel.stop()
         if self.stdin_channel.is_alive():
             self.stdin_channel.stop()
         if self.hb_channel.is_alive():
             self.hb_channel.stop()
 
     @property
     def channels_running(self):
         """Are any of the channels created and running?"""
         return (self.shell_channel.is_alive() or self.sub_channel.is_alive() or
                 self.stdin_channel.is_alive() or self.hb_channel.is_alive())
 
     #--------------------------------------------------------------------------
     # Kernel process management methods:
     #--------------------------------------------------------------------------
     
     def cleanup_connection_file(self):
         """cleanup connection file *if we wrote it*
         
         Will not raise if the connection file was already removed somehow.
         """
         if self._connection_file_written:
             # cleanup connection files on full shutdown of kernel we started
             self._connection_file_written = False
             try:
                 os.remove(self.connection_file)
             except OSError:
                 pass
     
     def load_connection_file(self):
         """load connection info from JSON dict in self.connection_file"""
         with open(self.connection_file) as f:
             cfg = json.loads(f.read())
         
         self.ip = cfg['ip']
         self.shell_port = cfg['shell_port']
         self.stdin_port = cfg['stdin_port']
         self.iopub_port = cfg['iopub_port']
         self.hb_port = cfg['hb_port']
         self.session.key = str_to_bytes(cfg['key'])
     
     def write_connection_file(self):
         """write connection info to JSON dict in self.connection_file"""
         if self._connection_file_written:
             return
         self.connection_file,cfg = write_connection_file(self.connection_file,
             ip=self.ip, key=self.session.key,
             stdin_port=self.stdin_port, iopub_port=self.iopub_port,
             shell_port=self.shell_port, hb_port=self.hb_port)
         # write_connection_file also sets default ports:
         self.shell_port = cfg['shell_port']
         self.stdin_port = cfg['stdin_port']
         self.iopub_port = cfg['iopub_port']
         self.hb_port = cfg['hb_port']
         
         self._connection_file_written = True
     
     def start_kernel(self, **kw):
         """Starts a kernel process and configures the manager to use it.
 
         If random ports (port=0) are being used, this method must be called
         before the channels are created.
 
         Parameters:
         -----------
         launcher : callable, optional (default None)
              A custom function for launching the kernel process (generally a
              wrapper around ``entry_point.base_launch_kernel``). In most cases,
              it should not be necessary to use this parameter.
 
         **kw : optional
              See respective options for IPython and Python kernels.
         """
         if self.ip not in LOCAL_IPS:
             raise RuntimeError("Can only launch a kernel on a local interface. "
                                "Make sure that the '*_address' attributes are "
                                "configured properly. "
                                "Currently valid addresses are: %s"%LOCAL_IPS
                                )
         
         # write connection file / get default ports
         self.write_connection_file()
 
         self._launch_args = kw.copy()
         launch_kernel = kw.pop('launcher', None)
         if launch_kernel is None:
             from ipkernel import launch_kernel
         self.kernel = launch_kernel(fname=self.connection_file, **kw)
 
     def shutdown_kernel(self, restart=False):
         """ Attempts to the stop the kernel process cleanly. If the kernel
         cannot be stopped, it is killed, if possible.
         """
         # FIXME: Shutdown does not work on Windows due to ZMQ errors!
         if sys.platform == 'win32':
             self.kill_kernel()
             return
 
         # Pause the heart beat channel if it exists.
         if self._hb_channel is not None:
             self._hb_channel.pause()
 
         # Don't send any additional kernel kill messages immediately, to give
         # the kernel a chance to properly execute shutdown actions. Wait for at
         # most 1s, checking every 0.1s.
         self.shell_channel.shutdown(restart=restart)
         for i in range(10):
             if self.is_alive:
                 time.sleep(0.1)
             else:
                 break
         else:
             # OK, we've waited long enough.
             if self.has_kernel:
                 self.kill_kernel()
 
         if not restart and self._connection_file_written:
             # cleanup connection files on full shutdown of kernel we started
             self._connection_file_written = False
             try:
                 os.remove(self.connection_file)
             except IOError:
                 pass
 
     def restart_kernel(self, now=False, **kw):
         """Restarts a kernel with the arguments that were used to launch it.
 
         If the old kernel was launched with random ports, the same ports will be
         used for the new kernel.
 
         Parameters
         ----------
         now : bool, optional
             If True, the kernel is forcefully restarted *immediately*, without
             having a chance to do any cleanup action.  Otherwise the kernel is
             given 1s to clean up before a forceful restart is issued.
 
             In all cases the kernel is restarted, the only difference is whether
             it is given a chance to perform a clean shutdown or not.
 
         **kw : optional
             Any options specified here will replace those used to launch the
             kernel.
         """
         if self._launch_args is None:
             raise RuntimeError("Cannot restart the kernel. "
                                "No previous call to 'start_kernel'.")
         else:
             # Stop currently running kernel.
             if self.has_kernel:
                 if now:
                     self.kill_kernel()
                 else:
                     self.shutdown_kernel(restart=True)
 
             # Start new kernel.
             self._launch_args.update(kw)
             self.start_kernel(**self._launch_args)
 
             # FIXME: Messages get dropped in Windows due to probable ZMQ bug
             # unless there is some delay here.
             if sys.platform == 'win32':
                 time.sleep(0.2)
 
     @property
     def has_kernel(self):
         """Returns whether a kernel process has been specified for the kernel
         manager.
         """
         return self.kernel is not None
 
     def kill_kernel(self):
         """ Kill the running kernel. """
         if self.has_kernel:
             # Pause the heart beat channel if it exists.
             if self._hb_channel is not None:
                 self._hb_channel.pause()
 
             # Attempt to kill the kernel.
             try:
                 self.kernel.kill()
             except OSError, e:
                 # In Windows, we will get an Access Denied error if the process
                 # has already terminated. Ignore it.
                 if sys.platform == 'win32':
                     if e.winerror != 5:
                         raise
                 # On Unix, we may get an ESRCH error if the process has already
                 # terminated. Ignore it.
                 else:
                     from errno import ESRCH
                     if e.errno != ESRCH:
                         raise
             self.kernel = None
         else:
             raise RuntimeError("Cannot kill kernel. No kernel is running!")
 
     def interrupt_kernel(self):
         """ Interrupts the kernel. Unlike ``signal_kernel``, this operation is
         well supported on all platforms.
         """
         if self.has_kernel:
             if sys.platform == 'win32':
                 from parentpoller import ParentPollerWindows as Poller
                 Poller.send_interrupt(self.kernel.win32_interrupt_event)
             else:
                 self.kernel.send_signal(signal.SIGINT)
         else:
             raise RuntimeError("Cannot interrupt kernel. No kernel is running!")
 
     def signal_kernel(self, signum):
         """ Sends a signal to the kernel. Note that since only SIGTERM is
         supported on Windows, this function is only useful on Unix systems.
         """
         if self.has_kernel:
             self.kernel.send_signal(signum)
         else:
             raise RuntimeError("Cannot signal kernel. No kernel is running!")
 
     @property
     def is_alive(self):
         """Is the kernel process still running?"""
         if self.has_kernel:
             if self.kernel.poll() is None:
                 return True
             else:
                 return False
         elif self._hb_channel is not None:
             # We didn't start the kernel with this KernelManager so we
             # use the heartbeat.
             return self._hb_channel.is_beating()
         else:
             # no heartbeat and not local, we can't tell if it's running,
             # so naively return True
             return True
 
     #--------------------------------------------------------------------------
     # Channels used for communication with the kernel:
     #--------------------------------------------------------------------------
 
     @property
     def shell_channel(self):
         """Get the REQ socket channel object to make requests of the kernel."""
         if self._shell_channel is None:
             self._shell_channel = self.shell_channel_class(self.context,
                                                          self.session,
                                                          (self.ip, self.shell_port))
         return self._shell_channel
 
     @property
     def sub_channel(self):
         """Get the SUB socket channel object."""
         if self._sub_channel is None:
             self._sub_channel = self.sub_channel_class(self.context,
                                                        self.session,
                                                        (self.ip, self.iopub_port))
         return self._sub_channel
 
     @property
     def stdin_channel(self):
         """Get the REP socket channel object to handle stdin (raw_input)."""
         if self._stdin_channel is None:
             self._stdin_channel = self.stdin_channel_class(self.context,
                                                        self.session,
                                                        (self.ip, self.stdin_port))
         return self._stdin_channel
 
     @property
     def hb_channel(self):
         """Get the heartbeat socket channel object to check that the
         kernel is alive."""
         if self._hb_channel is None:
             self._hb_channel = self.hb_channel_class(self.context,
                                                        self.session,
                                                        (self.ip, self.hb_port))
         return self._hb_channel