references/sources.md
Sources and verification record
Current API/source/runtime review: 2026-10-01, SimPy 4.1.2, Python 3.13.3.
Official PyPI metadata still identifies 4.1.2 as the latest release. The 410,129-byte
official source distribution and the cached installed release were inspected;
core/event/resource/real-time signatures and implementation behavior agree.
Versioned documentation is preferred over the moving latest build.
The release and bibliographic notes below retain their original 2026-07-23 access dates where that historical review was not repeated. They are not fresh live-probe claims. On this refresh, the core/events/resources/rt/util/exceptions API pages, monitoring guide, and Law/L'Ecuyer/Sargent proceedings were opened directly. Several topical pages were inaccessible through the browser extraction tool; the corresponding official 4.1.2 sdist documentation/source supplied the fallback.
Current audit findings
| Surface | Verified behavior / correction |
|---|---|
| Environment(initial_time), run, step, peek, schedule, now, active_process | Queue minimum chosen at each step; callbacks can insert urgent same-time work. Numeric horizon excludes ordinary boundary events. Stop callback reschedules remaining callbacks at priority -1. |
| Event, Timeout, Process, AnyOf, AllOf, ConditionValue, Interrupt | Values, failure propagation, conditions and interrupt lifecycle reproduced. Event.trigger() returns None despite an inconsistent prose docstring; fail() source accepts BaseException despite its Exception annotation. |
| Resource/Priority/Preemptive, Request/Release, Preempted, Put/Get cancellation | Resource request key includes (priority, time, not preempt); context exit cancels pending requests and releases acquired slots. |
| Container/Store/FilterStore/PriorityStore/PriorityItem | Blocked requests wait; filtering can bypass earlier unmatched gets. PriorityStore is a heap and equal-priority items are not FIFO without an explicit tie-breaker. |
| RealtimeEnvironment, factor, strict, sync | Mapping and lag check inspected in released rt.py. Constructor does not enforce positive finite factor. sync resets wall origin, not the initial simulation origin. |
| Monitoring | Native deterministic checks cover urgent trace insertion, final time weighting, warm-up windows, cancellation, and monitor non-interference. |
| Queue/replication helpers | Schema 1.2 counts throughput by window departures and loss by window arrivals; arrival-cohort completed customer means remain censored. Runtime version and metric definitions are recorded. |
There are no service/API endpoints, credentials, request bodies, pagination, or remote mutations in these helpers. The URLs here are package documentation, source, and methodological references. Native discrete-event fixtures verify software mechanics, not a calibrated operational system. Real-time strict/sync unit checks use a controlled clock; a short wall-clock smoke is only a timing sanity check, not a hardware-in-the-loop performance claim.
The isolated suite is under repository-level tests/simpy/. CLI round trips use
tiny local JSON/JSONL/CSV files; no user data or live service was exercised.
The complete minimal model in SKILL.md and ten CLI workflows were executed.
All 46 Python documentation fragments parsed; context-dependent reference snippets
are illustrative patterns, with representative behaviors exercised by the native
suite rather than every fragment executed independently.
Release, packaging, and source
- SimPy on PyPI — 4.1.2, released
2026-05-24;
Requires-Python >=3.8; classifiers list Python 3.8-3.14, CPython, and PyPy; MIT; no runtime dependencies. Verified 2026-07-23. - GitLab source project — canonical repository, MIT license, tagged source, tests, and documentation. Accessed 2026-07-23.
- GitLab tag 4.1.2 —
protected/verified tag, commit
f43816490c6f76f336ad6e457d3cab9f386894af, dated 2026-05-24. - Tagged CHANGES.rst
— 4.1.2 entry dated 2026-05-23: Python 3.13/3.14 support,
modern-interpreter test/doc fixes,
ConditionValueexplicitly unhashable, lint/type cleanup. - Tagged pyproject.toml — Python requirement/classifiers, no dependencies, setuptools build, pytest, mypy, and ruff configuration. Dated 2026-05-23; accessed 2026-07-23.
- Tagged tox.ini and tagged tests — upstream test matrix and behavior tests. Accessed 2026-07-23.
- Tagged
simpy/core.py— authoritativeEnvironment.run(),step(), queue tuple, andStopSimulationimplementation. Accessed 2026-07-23.
PyPI file hashes observed 2026-07-23:
- wheel
simpy-4.1.2-py3-none-any.whlSHA-25643071f84b6512c9b4fcb33ef057f240ccb1d1f3b263f9b4f9229d072e310b372; - sdist
simpy-4.1.2.tar.gzSHA-25676ef36b71e0436ba94e55febc001c78879e493a323f045bbcfbb0b216e9b1fbc.
Use the pinned package command in SKILL.md; hashes are recorded for provenance,
not embedded into uv pip install syntax.
Official SimPy 4.1.2 documentation
Entry points
- 4.1.2 documentation contents
- 4.1.2 tutorial
- 4.1.2 topical guides
- 4.1.2 examples
- 4.1.2 API reference
- history/change log
- license
All accessed 2026-07-23. The examples index covers condition events, interrupts, Resource, PreemptiveResource, Container, Store, shared events, and process waiting.
Core topical guides
- SimPy basics
- Environments
- Events
- Process Interaction
- Shared Resources
- Monitoring
- Time and Scheduling
- Real-time simulations
All accessed 2026-07-23.
API pages
simpy.core—Environment,run,schedule,peek,step,EmptySchedule.simpy.events—Event,Timeout,Process,Condition,AnyOf,AllOf, priorities.simpy.exceptions—Interrupt.simpy.resources— Resource/Priority/Preemptive, Container, Store/Filter/Priority, base events.simpy.rt—RealtimeEnvironment.simpy.util— helper processes.
All accessed 2026-07-23.
Source-level semantic finding
The 4.1.2 Environments topical guide informally says Event-based run() returns
when the Event has been processed. The 4.1.2 API docstring says "triggered."
Tagged core.py is decisive:
run(until=event)appendsStopSimulation.callback;step()sets callbacks toNone, executes them, catchesStopSimulation, restores callbacks remaining after the stopping callback, and reschedules that Event at priority-1.
Therefore the target value can be returned while target.processed is still false,
until one more step processes the rescheduled empty-callback Event. This was
reproduced under the exact PyPI 4.1.2 wheel on Python 3.13 and is covered by
the repository-level tests/simpy/test_scripts.py.
Numeric run(until=time) creates an urgent internal stop Event, so normal Events at
that exact time are excluded. This was verified from both tagged source and runtime
tests.
Primary simulation-method sources
These sources describe simulation-study design and output analysis. They are not SimPy API documentation.
- Averill M. Law, "Statistical Analysis of Simulation Output Data", Proceedings of the 2020 Winter Simulation Conference, 2020. Defines terminating/nonterminating analyses; independent replications; replication/deletion; warm-up/run length; Student-t intervals; and warns against naive single-run IID intervals. Accessed 2026-07-23.
- Pierre L'Ecuyer, "Random Number Generation with Multiple Streams for Sequential and Parallel Computing",
Proceedings of the 2015 Winter Simulation Conference, 2015,
DOI
10.1109/WSC.2015.7408151. Covers streams/substreams, one stream per random source, replication separation, common random numbers, testing, and exact reproducibility. Accessed 2026-07-23. - Robert G. Sargent, "Verification and Validation of Simulation Models", Proceedings of the 2010 Winter Simulation Conference, 2010. Defines model verification, conceptual/data/operational validity, intended-purpose/domain framing, sensitivity analysis, and validation evidence. Accessed 2026-07-23.
- Stewart Robinson, "Conceptual Modelling for Simulation Part I: Definition and Requirements", Journal of the Operational Research Society 59, 278-290, 2008. Identifies validity, credibility, utility, feasibility, and simplest-adequate-model requirements. Publisher page accessed 2026-07-23.
- Lee W. Schruben, "Detecting Initialization Bias in Simulation Output", Operations Research 30(3), 569-590, June 1982. Original initialization-bias testing method. Metadata/abstract verified via IDEAS/RePEc on 2026-07-23.
- Stewart Robinson, "A Statistical Process Control Approach to Selecting a Warm-up Period for a Discrete-event Simulation", European Journal of Operational Research 176(1), 332-346, January 2007. Accessed 2026-07-23.
- Kathryn Hoad, Stewart Robinson, Ruth Davies, "Automated Selection of the Number of Replications for a Discrete-event Simulation", Journal of the Operational Research Society 61(11), 1632-1644, November 2010. Replication-count/precision methodology; metadata verified 2026-07-23.
- Thomas Monks et al., "Strengthening the Reporting of Empirical Simulation Studies: Introducing the STRESS Guidelines", Journal of Simulation, 2018/2019 publication record. Reporting checklists for objectives, logic, data, experimentation, implementation, and code access. EQUATOR record last updated 2021-11-19; both accessed 2026-07-23.
- Averill M. Law and W. David Kelton, "Confidence Intervals for Steady-State Simulations: I. A Survey of Fixed Sample Size Procedures", Operations Research 32(6), 1221-1239, December 1984. Surveys replication, batch means, autoregressive, spectral, and regenerative fixed-run methods and cautions that adequate run length is model-dependent. Accessed 2026-07-23.
Research queries used
Parallel searches/extractions covered:
- latest SimPy stable version, Python support, PyPI files, GitLab tags/changelog;
- Environment/Event/Process/Timeout/Condition and scheduling;
- all Resource, Container, and Store variants;
- interrupts, monitoring, stepping, real-time, examples, API, and testing config;
- independent replications, confidence intervals, warm-up/transient bias, random streams, conceptual modeling, V&V, sensitivity, and reproducibility.
No Parallel JSON artifacts were written into the repository.