{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "6ca39b80-dd96-449e-91d2-42031c3a5e3f",
   "metadata": {},
   "source": [
    "# Tutorial about mutiprocessing using ray"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5779d015-a270-4944-8159-6377635195b9",
   "metadata": {},
   "source": [
    "We will describe how to set up an analysis pipeline to process multiple datasets in parallel using the framework [ray](https://ray.io/)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "930146a8-4017-4a6a-8b7d-ccf0037e9b37",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import sys\n",
    "import logging\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import ray\n",
    "\n",
    "import locan as lc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1b0095f4-3817-4084-9993-9b3e642bfab1",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "lc.show_versions(dependencies=False, verbose=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "38ac0b33-738a-44ad-8282-2544153acdf1",
   "metadata": {},
   "source": [
    "## Activate logging"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "29bf9112-cb7b-4b9c-bd19-ec0224e4c8a1",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "logging.basicConfig(stream=sys.stdout, level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')\n",
    "logger = logging.getLogger()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "94931283-67d0-4005-b153-8d1541225f37",
   "metadata": {},
   "source": [
    "For changing the configuration logging has to be reloaded or the kernel be restarted."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c3d02767-bf67-47b3-8732-5f266210ffc3",
   "metadata": {},
   "source": [
    "## Synthetic data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c0a57554-5bab-4034-a697-90ead8a32bd5",
   "metadata": {},
   "source": [
    "Simulate 3 datasets of localization data that is homogeneously Poisson distributed and treat them as files."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "718aeae8-a4d3-48ca-bada-d0d4183237e7",
   "metadata": {},
   "outputs": [],
   "source": [
    "rng = np.random.default_rng(seed=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c2703702-24e3-446e-bd50-3a4cf9c3161a",
   "metadata": {},
   "outputs": [],
   "source": [
    "locdatas = [lc.simulate_Poisson(intensity=1e-3, region=((0,1000), (0,1000)), seed=rng) for _ in range(3)]\n",
    "files = locdatas\n",
    "\n",
    "print(\"Element_counts:\", [locdata.meta.element_count for locdata in locdatas])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5494729d-661d-4f44-b41d-919e12d57c6d",
   "metadata": {},
   "source": [
    "## Analysis pipeline"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ef41d726-1d46-461b-90b8-8ff66c101ffd",
   "metadata": {},
   "source": [
    "Define an analysis pipeline. Typically a pipeline processes a single file, which in this example will be a an element of locdatas.\n",
    "\n",
    "Within the analysis procedure there will be more random number generation involved. Therefore a correctly generated seed has to be passed."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b9bd707b-f167-4c54-981d-6cd0ee98d0f6",
   "metadata": {},
   "outputs": [],
   "source": [
    "def computation(self, file, seed):\n",
    "    logging.basicConfig(level=logging.INFO)\n",
    "    logger.info(f'computation started for file: {file}')\n",
    "    \n",
    "    rng = np.random.default_rng(seed=seed)\n",
    "    \n",
    "    other_locdata = lc.simulate_Poisson(intensity=1e-3, region=((0,1000), (0,1000)), seed=rng)\n",
    "    self.nn = lc.NearestNeighborDistances().compute(locdata=file, other_locdata=other_locdata)\n",
    "        \n",
    "    return self"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c33ba275-1139-45d9-a767-4ea4f4b67eb4",
   "metadata": {},
   "source": [
    "## Run analysis in parallel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "51949da6-12bf-42b6-8d47-1c071cdea091",
   "metadata": {},
   "outputs": [],
   "source": [
    "ray.init()\n",
    "# ray.init(num_cpus = 4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b6c0362e-2629-4dc4-921b-d6d0c5bf3d25",
   "metadata": {},
   "outputs": [],
   "source": [
    "%%time\n",
    "@ray.remote\n",
    "def worker(file, seed):\n",
    "    pipe = lc.Pipeline(computation=computation, file=file, seed=seed).compute()\n",
    "    return pipe\n",
    "\n",
    "n_processes = len(files)\n",
    "ss = np.random.SeedSequence()\n",
    "child_seeds = ss.spawn(n_processes)\n",
    "\n",
    "futures = [worker.remote(file=file, seed=seed) for file, seed in zip(locdatas, child_seeds)]\n",
    "pipes = ray.get(futures)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "23b5d32d-f762-48a2-a467-aa2df6b5cdf0",
   "metadata": {},
   "source": [
    "## Visualize the combined results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "53a13f7a-94cc-4e83-bfac-8971eaa9aeba",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "[pipe.meta for pipe in pipes]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a9111063-d43d-4d18-9f55-08fc05f9f8b5",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "fig, ax = plt.subplots(nrows=1, ncols=1)\n",
    "for pipe in pipes:\n",
    "    pipe.nn.hist(ax=ax)\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.8"
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "state": {},
    "version_major": 2,
    "version_minor": 0
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}