Python vs Java

PureEdgeSim for Python is a faithful port of the Java simulator: same scenario model, same parameter names, same CSV columns. This page is for readers coming from Java. It maps the API, translates the commands, and lists the honest gaps.

Class to module map

Java uses many small classes; Python groups them into modules. The table lists the notable ones.

Java package and class	Python module and name
simulationengine.PureEdgeSim	engine.Engine
simulationengine.Event, FutureQueue, SimEntity	engine.Event, engine.FutureQueue, engine.SimEntity
distribution.Distribution and its subclasses	distribution.Distribution and matching subclasses
distribution.Randomness	randomness.Randomness
scenariomanager.SimulationParameters	parameters.SimulationParameters (a dataclass)
scenariomanager.ParametersParser, FileParserAbstract, XmlFileParser, DistributionReader	parameters.load_parameters, distribution.distribution_from_element
scenariomanager.ApplicationFileParser, ComputingNodesParser, DatacentersParser, EdgeDevicesParser	config.load_settings, config.load_applications, config.load_cloud, config.load_edge_datacenters, config.load_devices
scenariomanager.Scenario	manager.Scenario
taskgenerator.Application, Subtask	config.Application, config.Subtask
taskgenerator.Task, DefaultTask, TaskAbstract	tasks.Task
taskgenerator.DefaultTaskGenerator, TaskGenerator	tasks.DefaultTaskGenerator
datacentersmanager.ComputingNode, DefaultComputingNode, AbstractNode, EnergyAwareNode, LocationAwareNode, NetworkingNode	nodes.ComputingNode
datacentersmanager.ComputingNodeNull	nodes.ComputingNode.NULL
datacentersmanager.DefaultComputingNodesGenerator, ComputingNodesGenerator, DataCentersManager	topology.ComputingNodesGenerator
datacentersmanager.DefaultTopologyCreator, TopologyCreator	topology.DefaultTopologyCreator
datacentersmanager.Router	nodes.Router
energy.EnergyModelComputingNode	nodes.EnergyModelComputingNode
locationmanager.MobilityModel and subclasses, Location	mobility.MobilityModel, its subclasses, mobility.Location
network.NetworkModel, DefaultNetworkModel	network.DefaultNetworkModel
network.NetworkLink and its up/down/type subclasses (NetworkLinkWifiUp, NetworkLinkCellularDown, …)	network.NetworkLink plus the LinkType (WAN, MAN, LAN, IGNORE) and LinkTechnology enums
network.InfrastructureGraph, TransferProgress	network.InfrastructureGraph, network.TransferProgress
taskorchestrator.Orchestrator, DefaultOrchestrator	orchestrator.Orchestrator, orchestrator.DefaultOrchestrator
simulationmanager.Simulation, SimulationAbstract	runner.Simulation
simulationmanager.DefaultSimulationManager, SimulationManager	manager.SimulationManager
simulationmanager.RunMetadata	metadata.write_run_metadata
MainApplication	cli.main
simulationmanager.ChartsGenerator, all of simulationvisualizer.*	not ported (see gaps)
simulationmanager.SimLog	not ported (see gaps)
simulationmanager.SimulationThread	replaced by runner.Simulation._run_parallel with a ProcessPoolExecutor

A few naming differences worth knowing:

• Java getMips() maps to the Python attribute mips_per_core (it is always per core; total_mips is the product with cores).
• An application’s result size is results_size in config.Application (mirroring the XML) but result_size on a tasks.Task (singular).
• The Java TaskType/tier concept is the Python NodeType enum: CLOUD, EDGE_DATACENTER, EDGE_DEVICE, ROUTER.
• Java’s integer event tags are the Python EventType enum in nodes.

Build and run commands

Task	Java	Python
Install/build	mvn install (or mvn package)	pip install -e ".[test]"
Run the bundled scenario	mvn exec:java -Dexec.mainClass="com.mechalikh.pureedgesim.MainApplication"	pureedgesim run --settings examples/default_settings
Run one example	mvn exec:java -Dexec.mainClass="...Example1"	python examples/custom_orchestrator/custom_orchestrator.py
Configure the scenario	edit PureEdgeSim/settings/	edit examples/default_settings/ (or pass --settings <folder>)

The Python package has no runtime dependencies (Maven pulls in libraries; here the standard library is enough). Each example folder carries the settings it runs with, so a scenario is self-contained; Examples lists the runnable scripts and the Java example each one mirrors.

Tests

Task	Java	Python
Framework	JUnit	pytest
Run the suite	mvn test	pytest
Coverage	JaCoCo (jacoco.xml)	pytest --cov=pureedgesim (needs pytest-cov)
Shared fixtures	test utility classes	tests/conftest.py fixtures (parameters, files, manager)

The Python tests build a real SimulationManager from examples/default_settings/ and assert on the resulting Task states and metrics, the same spirit as the JUnit suite. See Contributing.

What the Python port does not have yet

These are real gaps against the Java simulator, verified against src/pureedgesim/.

• No visualizer or charts. There is no simulationvisualizer equivalent and no ChartsGenerator. The properties display_real_time_charts, auto_close_real_time_charts, charts_update_interval, and save_charts are parsed and validated but do nothing. No live map/CPU/WAN/tasks charts and no saved chart images.
• No per-event log. deep_log_enabled is parsed but there is no SimLog; nothing reads the flag.
• No pluggable registry model. Container download is modeled in the core, but only in registry_mode = CLOUD over the default network (the first cloud node is the registry, and a node that has run an app caches it in installed_containers). The CACHE mode is accepted by the validator and implemented in the container_cache example instead of in the core, because there is no separate registry class to subclass the way Java has one.
• No fuzzy logic orchestrator. The two Java examples that need the jFuzzyLogic jar, Example5 and Example8, have no Python counterpart; the package takes no third-party dependency. custom_orchestrator exercises the same injection point with a distance rule.
• Two SimulationParameters fields are not read from properties. enable_location_updates and settings_folder exist on the dataclass (defaults True and "settings"), but load_parameters never assigns them from the file.
• The two device-list columns are written differently. Columns 50 and 51 hold the remaining battery of each device and the death times. Java wraps them in brackets, [99.13- 99.77- ...], and writes [] when empty; Python joins with - and writes an empty field. The numbers are the same, the punctuation is not, so compare those two columns by parsing rather than by diffing the text.
• A heuristic example settles on a nearby, not identical, mix. Both simulators draw placement decisions from a random stream, and the two streams are not the same generator. Run the same settings and seed side by side and the totals, the energy and the traffic agree to within a few percent, while the split between the three tiers and the count of delay failures move with the sample. The deterministic scenario, the trace replay in its recorded mode, draws nothing and gives the same 8819 requests, the same 3180 successes, the same compute total and the same traffic as Java. Two columns are not identical: the total waiting time differs by a hundredth of a percent and the energy of the run by two hundredths.

What the Python port does have, matching Java:

• A parallel executor (--parallel, ProcessPoolExecutor), producing Parallel_simulation.csv and pooled run.json.
• Nine mobility models, sixteen distributions, service graphs with dependencies and optional subtasks, request deadline accounting, battery and static/dynamic energy, and the full WAN/MAN/LAN Wi-Fi/cellular/Ethernet network model.
• Eight set_custom_... injection points, one per pluggable class in Java: mobility model, orchestrator, network model, computing-nodes generator, topology creator, task generator, computing node, and simulation manager. See Architecture.

Cross-links

• The Java manual home: Java docs.
• Replacing the mobility model or orchestrator in Python: Simulation models and Writing orchestrators.