API timings

uber · Sep 3, 2021 · 90c688c · 90c688c
1 parent 2088cfb
commit 90c688c
Show file tree

Hide file tree

Showing 2 changed files with 334 additions and 2 deletions.
diff --git a/makefile b/makefile
@@ -1,5 +1,6 @@
 init: purge
-	virtualenv -p python3 env
+	python -m venv env
+	env/bin/pip install --upgrade pip wheel setuptools
 	env/bin/pip install -r requirements.txt
 
 purge:
@@ -9,4 +10,4 @@ lab:
 	env/bin/jupyter lab
 
 ipython:
-	env/bin/ipython
+	env/bin/ipython
diff --git a/notebooks/time_h3_apis.ipynb b/notebooks/time_h3_apis.ipynb
@@ -0,0 +1,331 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "95776ece-9628-40ad-a6cb-bf68690167ec",
+   "metadata": {},
+   "source": [
+    "# API Speed Comparison\n",
+    "\n",
+    "We time a simple toy computational task across the various APIs.\n",
+    "We also compare the option of doing most of the computation in one of the faster APIs (using the `int` representation of H3 indexes),\n",
+    "and then converting the results to the more familiar format of Python `str` objects."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "cef5ba28-c8b0-42de-b55f-7a3d92436b32",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import h3\n",
+    "import h3.api.numpy_int\n",
+    "from time import sleep\n",
+    "\n",
+    "sleep_seconds = 20\n",
+    "\n",
+    "def compute(h3_lib, N=100):\n",
+    "    h   = h3_lib.geo_to_h3(0, 0, 9)\n",
+    "    out = h3_lib.k_ring(h, N)\n",
+    "    out = h3_lib.compact(out)\n",
+    "    \n",
+    "    return out\n",
+    "\n",
+    "def compute_and_convert(h3_lib, N=100):\n",
+    "    out = [\n",
+    "        h3.h3_to_string(h)\n",
+    "        for h in compute(h3_lib, N)\n",
+    "    ]\n",
+    "    \n",
+    "    return out"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "9c933e85-5a9a-4d07-b4c4-20a88c97fa88",
+   "metadata": {},
+   "source": [
+    "# Compute with each API\n",
+    "\n",
+    "**Benchmarking note**: We put a `sleep()` before each timing to help stablize the results (at least on my laptop)."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "1889e459-0482-44f8-a4a9-a363f6a64231",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "sleep(sleep_seconds)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "e143ba5e-618b-47a9-913b-66f25877b240",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "54.2 ms ± 643 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)\n"
+     ]
+    }
+   ],
+   "source": [
+    "%%timeit\n",
+    "compute(h3.api.basic_str)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "e4b4f37e-9c23-408a-84e0-ee8a4e92d793",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "sleep(sleep_seconds)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "6d80bca9-c0f5-4aaa-8488-f7bd174c7ee3",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "31.1 ms ± 372 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)\n"
+     ]
+    }
+   ],
+   "source": [
+    "%%timeit\n",
+    "compute(h3.api.basic_int)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "766baaec-a40f-4a5c-8d9f-3c48c2835697",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "sleep(sleep_seconds)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "f2713e84-c4fb-4a60-b913-c1fb7ea7fd4b",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "6.69 ms ± 75.5 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
+     ]
+    }
+   ],
+   "source": [
+    "%%timeit\n",
+    "compute(h3.api.numpy_int)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "fa9186bb-1ffe-48be-99c3-f08f9aae2ce2",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "sleep(sleep_seconds)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "e9931ac6-eeba-4168-92e1-f8f330b1ca42",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "6.67 ms ± 72.8 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
+     ]
+    }
+   ],
+   "source": [
+    "%%timeit\n",
+    "compute(h3.api.memview_int)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "97e8e885-401b-42aa-a28e-534f828bf003",
+   "metadata": {},
+   "source": [
+    "# Compute with `int` APIs and convert to `str`"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "id": "ea7e8528-d84d-4e23-af12-9daab077aea7",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "sleep(sleep_seconds)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "id": "51c02c9f-0a03-4e84-943d-09969949b83d",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "34 ms ± 421 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)\n"
+     ]
+    }
+   ],
+   "source": [
+    "%%timeit\n",
+    "compute_and_convert(h3.api.basic_int)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "d56ba926-9c5e-49d2-9fb6-b897070cc474",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "sleep(sleep_seconds)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "id": "b11e3d6d-58d3-418f-9798-2a80dc16b133",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "7.85 ms ± 99.1 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
+     ]
+    }
+   ],
+   "source": [
+    "%%timeit\n",
+    "compute_and_convert(h3.api.memview_int)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "id": "eab6f2d9-f52a-44e7-9079-19d1754c5a20",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "sleep(sleep_seconds)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "id": "5cf6e8a8-583d-4309-8ed2-5b4c31c8765a",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "7.83 ms ± 101 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
+     ]
+    }
+   ],
+   "source": [
+    "%%timeit\n",
+    "compute_and_convert(h3.api.numpy_int)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "381e7c85-a70a-4563-8504-29e39639821f",
+   "metadata": {},
+   "source": [
+    "# Speedup\n",
+    "\n",
+    "We typically see about a 6--8x speedup, comparing the `h3.api.basic_str` interface against computing with the `h3.api.numpy_int` and then converting the results back to `str`."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "id": "188da0e8-efd8-43f5-9f45-b821e085b8a7",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "6.922094508301405"
+      ]
+     },
+     "execution_count": 16,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# REMEMBER! update the numbers from the `%%timeit`s above!\n",
+    "\n",
+    "c_str = 54.2\n",
+    "cnc_numpy = 7.83\n",
+    "\n",
+    "c_str/cnc_numpy"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "233d4a44-f8e6-42d2-b54d-b6001c776102",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "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": "3.9.1"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}