|
| 1 | +[](https://pythonhealthdatascience.github.io/stars-simpy-jupterlite/notebooks/?path=01_urgent_care_model.ipynb) |
| 2 | +[](https://opensource.org/licenses/MIT) |
| 3 | +[](https://doi.org/10.5281/zenodo.10987817) |
| 4 | +[](https://www.python.org/downloads/release/python-3100/) |
| 5 | +[](https://orcid.org/0000-0001-5274-5037) |
| 6 | +[](https://orcid.org/0000-0003-2631-4481) |
| 7 | + |
| 8 | +# Towards Sharing Tools and Artifacts for **Reproducible** Simulation **(v1.5)**: a JuypterLite template for `simpy` models |
| 9 | + |
| 10 | +## 1. Overview |
| 11 | + |
| 12 | +The materials and methods in this repository support work towards developing the S.T.A.R.S healthcare framework version 1.5 (**S**haring **T**ools and **A**rtifacts for **R**eproducible **S**imulations in healthcare). The code and written materials here are a work in progress to demonstrate the application of S.T.A.R.S' version to sharing a `simpy` discrete-event simuilation model and associated research artifacts. |
| 13 | + |
| 14 | +The model will run on a users browser without the need to install any components. This is achieved using Web Assembly technology i.e. [JupterLite](https://github.com/jupyterlite/jupyterlite) and [xeus-python](https://github.com/jupyter-xeus/xeus-python). A model notebook is downloaded to the users local machine and all dependencies are pre-installed via conda-forge. The model then lives in the browsers cache. The user can make changes to the model or create new files and these are persisted (until the browser cache is cleared). |
| 15 | + |
| 16 | +> Try it in your browser now: https://pythonhealthdatascience.github.io/stars-simpy-jupterlite |
| 17 | +
|
| 18 | +### 1.1. Use case |
| 19 | + |
| 20 | +* A researcher wishes to share a runnable version of a simulation model with their publication (e.g. written in `simpy`). The code allows others to replicate the simulation results, tables and charts in a paper and allows others to reuse the model. |
| 21 | +* The researcher wants the model to be immediately usable. Users should not need to install python, `simpy` or any dependencies. |
| 22 | +* The researcher either wants to reduce load on online open science compute infrastructure (e.g. mybinder.org) or does not want to rely on it. |
| 23 | +* Users may want to use a version of their own data due to governance, ethics or other reasons **cannot upload the data to a remote instance of the model.** |
| 24 | +* Loading the model is as simple as clicking a URL. |
| 25 | + |
| 26 | +### 1.2. Credits ✨ |
| 27 | + |
| 28 | +> We would like to thank the [JupterLite](https://github.com/jupyterlite/jupyterlite) and [xeus-python](https://github.com/jupyter-xeus/xeus-python) developers for making this work possible. This discrete-event simulation focussed repository was based on the learning materials and template provided by [Jupyterlite xeus-python demo](https://github.com/jupyterlite/xeus-python-demo) and [tutorial given at PyData 2023](https://www.youtube.com/watch?v=WXRslU9D3bo) by Jeremy Tuloup. |
| 29 | +
|
| 30 | +### 1.3. Citation |
| 31 | + |
| 32 | +If you use the template in your work we would greatly appreciate a citation when you publish your work. |
| 33 | + |
| 34 | +> Monks, T., & Harper, A. (2024). Simpy JupyterLite Template (v0.1.0). Zenodo. https://doi.org/10.5281/zenodo.10987817 |
| 35 | +
|
| 36 | +``` |
| 37 | +@software{monks_harper_jupyterlite_template, |
| 38 | + author = {Monks, Thomas and |
| 39 | + Harper, Alison}, |
| 40 | + title = {Simpy JupyterLite Template}, |
| 41 | + month = apr, |
| 42 | + year = 2024, |
| 43 | + publisher = {Zenodo}, |
| 44 | + version = {v0.1.0}, |
| 45 | + doi = {10.5281/zenodo.10987817}, |
| 46 | + url = {https://doi.org/10.5281/zenodo.10987817} |
| 47 | +} |
| 48 | +``` |
| 49 | + |
| 50 | + |
| 51 | +## 2. The example model included |
| 52 | + |
| 53 | +The `simpy` model is adapted from [Monks and Harper (2023)](https://github.com/pythonhealthdatascience/stars-simpy-example-docs) |
| 54 | + |
| 55 | +> Monks, T., & Harper, A. (2023). Towards Sharing Tools and Artifacts for Reusable Simulation: example enhanced documentation for a simpy model. (v1.1.0). Zenodo. https://doi.org/10.5281/zenodo.10054063 |
| 56 | +
|
| 57 | +Full documentation of this model is available in our [JupyterBook](https://pythonhealthdatascience.github.io/stars-simpy-example-docs) |
| 58 | + |
| 59 | +In summary, we adapt a textbook example from Nelson (2013): a terminating discrete-event simulation model of a U.S based treatment centren summary the model. The example is based on exercise 13 from Nelson (2013) page 170. |
| 60 | + |
| 61 | +> *Nelson. B.L. (2013). [Foundations and methods of stochastic simulation](https://www.amazon.co.uk/Foundations-Methods-Stochastic-Simulation-International/dp/1461461596/ref=sr_1_1?dchild=1&keywords=foundations+and+methods+of+stochastic+simulation&qid=1617050801&sr=8-1). Springer.* |
| 62 | +
|
| 63 | +## 3. Try the example DES in your browser |
| 64 | + |
| 65 | +* Jupyterlab: https://pythonhealthdatascience.github.io/stars-simpy-jupterlite |
| 66 | +* Classic notebook: https://pythonhealthdatascience.github.io/stars-simpy-jupterlite/notebooks/?path=01_urgent_care_model.ipynb |
| 67 | + |
| 68 | +## 4. Using the template to create a new repo. |
| 69 | + |
| 70 | +> There are three steps: i.) create a new repo form the template; ii.) modify you repo settings so that GitHub pages are built from Actions. iii.) Commit changes and trigger the GitHub Action and deployment. We recommend reading all instructions first. |
| 71 | +
|
| 72 | +Let's assume you wanted to create a new discrete-event simulation model of cancer services. |
| 73 | + |
| 74 | +1. The first step is to copy the template. Click on the green "use this template" button in the top right and select "create a new repository" |
| 75 | +2. You will be prompted to enter a name of the repository - e.g. `cancer_model` - and a short description |
| 76 | +3. Click on "Create Repository" |
| 77 | + |
| 78 | + |
| 79 | + |
| 80 | +The JuypterLite interactive website is built from GitHub actions. **In the newly created repo for your model** do the following |
| 81 | + |
| 82 | +4. Navigate to "Settings->Pages" |
| 83 | +5. Under "Build and Deployment" set the "Source" to "GitHub actions". |
| 84 | + |
| 85 | + |
| 86 | + |
| 87 | +By default the build is trigged on any commit to the `main` branch. Push a small commit and it will trigger the build. This will take a few minutes. Your site will be published under https://{USERNAME}.github.io/{DEMO_REPO_NAME} |
| 88 | + |
| 89 | +## 5. How to install extra packages supporting your DES model. 📦 |
| 90 | + |
| 91 | +The repo contains two environment files. To install more dependencies for your DES model and analysis you need to edit the ``environment.yml`` file. |
| 92 | + |
| 93 | +The template ``environment.yml`` is as follows: |
| 94 | + |
| 95 | + |
| 96 | +```yml |
| 97 | +name: xeus-python-kernel |
| 98 | +channels: |
| 99 | + - https://repo.mamba.pm/emscripten-forge |
| 100 | + - conda-forge |
| 101 | +dependencies: |
| 102 | + - xeus-python |
| 103 | + - ipycanvas |
| 104 | + - simpy=4.1.1 |
| 105 | + - numpy |
| 106 | + - pandas |
| 107 | + - matplotlib |
| 108 | +``` |
| 109 | +
|
| 110 | +**Key points:** |
| 111 | +
|
| 112 | +* There are two channels in use. |
| 113 | + * `encription-forge` contains specific versions of the packages for web assembly These include `numpy` `pandas`, and `matplotlib`. Other popular packages include `scipy`, `scikit-learn` and `pytest`. |
| 114 | + * `conda-forge` for other installs you can use conda-forge. Only ``no-arch`` packages from ``conda-forge`` can be installed (simpy qualifies) |
| 115 | +* Note that `numpy`, `pandas` and `matplotlib` have specific versions available on `enscription-forge`. For this reason we recommend not including the package version number. |
| 116 | +* `simpy` is installed from `conda-forge` we were therefore able to freeze the version to 4.1.1 to aid reproducibility. |
| 117 | +* At the time of writing the xeus-python kernal will use python 3.11.3 |
| 118 | + |
| 119 | +As an example modification assume that you wanted to add two new packages: `plotly` and `scipy`. The first `plotly` is available ``no-arch`` from conda-forge so it is safe to include and if you wanted to you could try to include a version number. There is a specific version of `scipy` is available on `encription-forge` |
| 120 | + |
| 121 | +Our modified environment looks like: |
| 122 | + |
| 123 | +```yml |
| 124 | +name: xeus-python-kernel |
| 125 | +channels: |
| 126 | + - https://repo.mamba.pm/emscripten-forge |
| 127 | + - conda-forge |
| 128 | +dependencies: |
| 129 | + - xeus-python |
| 130 | + - ipycanvas |
| 131 | + - simpy=4.1.1 |
| 132 | + - numpy |
| 133 | + - pandas |
| 134 | + - matplotlib |
| 135 | + - plotly |
| 136 | + - scipy |
| 137 | +``` |
| 138 | + |
| 139 | +If you wanted to use an alternative simulation package to `simpy` this would need to be available on `conda-forge` and be ``no-arch``. An example package is `salabim`. A modification of the enviroment is: |
| 140 | + |
| 141 | + |
| 142 | +```yml |
| 143 | +name: xeus-python-kernel |
| 144 | +channels: |
| 145 | + - https://repo.mamba.pm/emscripten-forge |
| 146 | + - conda-forge |
| 147 | +dependencies: |
| 148 | + - xeus-python |
| 149 | + - ipycanvas |
| 150 | + - salabim |
| 151 | + - numpy |
| 152 | + - pandas |
| 153 | + - matplotlib |
| 154 | + - plotly |
| 155 | + - scipy |
| 156 | +``` |
0 commit comments