Source code for mph.session

"""Manages the local Comsol session."""
__license__ = 'MIT'


########################################
# Components                           #
########################################
from .client import Client             # client class
from .server import Server             # server class
from .config import option             # configuration


########################################
# Dependencies                         #
########################################
import jpype                           # Java bridge
import atexit                          # exit handler
import sys                             # system specifics
import platform                        # platform information
import threading                       # multi-threading
from logging import getLogger          # event logging


########################################
# Globals                              #
########################################
client = None                          # client instance
server = None                          # server instance
logger = getLogger(__package__)        # event logger


########################################
# Start                                #
########################################

[docs]def start(cores=None, version=None, port=None): """ Starts a local Comsol session. This convenience function provides for the typical use case of running a Comsol session on the local machine, i.e. *not* have a client connect to a remote server elsewhere on the network. Example: ```python import mph client = mph.start(cores=1) model = client.load('model.mph') model.solve() model.save() client.remove(model) ``` Depending on the platform, this may either be a stand-alone client (on Windows) or a thin client connected to a server running locally (on Linux and macOS). The reason for this disparity is that, while stand-alone clients are more lightweight and faster to start up, and therefore preferable, support for this mode of operation is somewhat limited on Unix-like operating systems, and thus not the default. Find more details in documentation chapter "Limitations". Due to limitations of the Java bridge, provided by the JPype library, only one client can be instantiated at a time. This is because JPype cannot manage more than one Java virtual machine within the same Python session. Therefore `start()` can only be called once. Subsequent calls will raise `NotImplementedError`. Separate Python processes would have to be started, or spawned, to work around this limitation. Refer to chapter "Demonstrations" in the documentation for an example of how to achieve this. The number of `cores` (threads) the Comsol instance uses can be restricted by specifying a number. Otherwise all available cores will be used. A specific Comsol `version` can be selected if several are installed, for example `version='5.3a'`. Otherwise the latest version is used. The server `port` can be specified if client–server mode is used. If omitted, the port number is chosen automatically by the server. Hand-selecting a port is helpful when starting multiple processes in order to run sessions in parallel. Servers started at roughly the same time will all try to reserve the same network port, a collision that will usually produce run-time errors. In stand-alone mode this problem cannot occur and the argument is just ignored. """ global client, server if client or server: error = 'Only one Comsol session can be started in the same process.' logger.critical(error) raise NotImplementedError(error) session = option('session') if session == 'platform-dependent': if platform.system() == 'Windows': session = 'stand-alone' else: session = 'client-server' logger.info('Starting local Comsol session.') if session == 'stand-alone': client = Client(cores=cores, version=version) elif session == 'client-server': server = Server(cores=cores, version=version, port=port) client = Client(cores=cores, version=version, port=server.port) else: error = f'Invalid session type "{session}".' logger.critical(error) raise ValueError(error) return client
######################################## # Stop # ######################################## def exit_hook(code=None): """Monkey-patches `sys.exit()` to preserve exit code at shutdown.""" global exit_code if isinstance(code, int): exit_code = code exit_function(code) def exception_hook_sys(exc_type, exc_value, exc_traceback): """Sets exit code to 1 if exception raised in main thread.""" global exit_code exit_code = 1 exception_handler_sys(exc_type, exc_value, exc_traceback) def exception_hook_threads(info): """Sets exit code to 1 if exception raised in any other thread.""" global exit_code exit_code = 1 exception_handler_threads(info) exit_code = 0 exit_function = sys.exit sys.exit = exit_hook exception_handler_sys = sys.excepthook sys.excepthook = exception_hook_sys # Only available as of Python 3.8, see bugs.python.org/issue1230540. if hasattr(threading, 'excepthook'): exception_handler_threads = threading.excepthook threading.excepthook = exception_hook_threads @atexit.register def cleanup(): """ Cleans up resources at the end of the Python session. This function is not part of the public API. It runs automatically at the end of the Python session and is not intended to be called directly from application code. Stops the local server instance possibly created by `start()` and shuts down the Java Virtual Machine hosting the client instance. """ if client and client.port: try: client.disconnect() except Exception: error = 'Error while disconnecting client at session clean-up.' logger.error(error, exc_info=True) if server and server.running(): server.stop() if jpype.isJVMStarted(): logger.info('Exiting the Java virtual machine.') sys.stdout.flush() sys.stderr.flush() jpype.java.lang.Runtime.getRuntime().exit(exit_code) # No code is reached from here on due to the hard exit of the JVM. logger.info('Java virtual machine has exited.')