perf(benchmarks): improves benchmarks (#26)

Author: micheleriva

.eslintignore

@@ -9,4 +9,5 @@ jest.config.js
 *.sh
 *.png
 *.html
-*.svg
+*.svg
+get-imdb-dataset.mjs

.eslintrc.cjs

@@ -9,6 +9,7 @@ module.exports = {
     'plugin:@typescript-eslint/recommended',
   ],
   rules: {
-    '@typescript-eslint/no-non-null-assertion': 0
+    '@typescript-eslint/no-non-null-assertion': 0,
+    'no-async-promise-executor': 0,
   }
 };

.gitignore

@@ -4,6 +4,7 @@
 
 dist
 *.csv
+*.tsv
 
 ### Intellij ###
 # Covers JetBrains IDEs: IntelliJ, RubyMine, PhpStorm, AppCode, PyCharm, CLion, Android Studio, WebStorm and Rider

packages/benchmarks/benchmark.js

@@ -1,7 +1,7 @@
-import fs from "fs";
 /*global console*/
 
-import { parse } from "csv-parse";
+import fs from "fs";
+import readline from "readline";
 import { Lyra } from "@nearform/lyra";
 
 const db = new Lyra({
@@ -14,30 +14,32 @@ const db = new Lyra({
 
 function populateDB() {
   console.log("Populating the database...");
-  return new Promise((resolve) => {
-    fs.createReadStream("./dataset/title.csv")
-      .pipe(parse({ delimiter: ";", from_line: 2 }))
-      .on("data", (row) => {
-        const [, type, title, , , , , , category] = row;
-
-        db.insert({
-          type,
-          title,
-          category,
-        });
-      })
-      .on("end", () => {
-        console.log("Database ready");
-        resolve(1);
+  return new Promise(async (resolve) => {
+    const fileStream = fs.createReadStream("./dataset/title.tsv");
+    const rl = readline.createInterface({
+      input: fileStream,
+      crlfDelay: Infinity,
+    });
+
+    for await (const row of rl) {
+      const [, type, title, , , , , , category] = row.split("\t");
+
+      db.insert({
+        type,
+        title,
+        category,
       });
+    }
+
+    resolve(1);
   });
 }
 
 async function main() {
   await populateDB();
 
   console.log("--------------------------------");
-  console.log("Results after 100.000 iterations");
+  console.log("Results after 1000 iterations");
   console.log("--------------------------------");
 
   const searchOnAllIndices = await searchBenchmark(db, {
@@ -48,6 +50,24 @@ async function main() {
     `Searching "believe" through 1M entries in all indices: ${searchOnAllIndices}`
   );
 
+  const exactSearchOnAllIndices = await searchBenchmark(db, {
+    term: "believe",
+    properties: "*",
+    exact: true,
+  });
+  console.log(
+    `Exact search for "believe" through 1M entries in all indices: ${exactSearchOnAllIndices}`
+  );
+
+  const typoTolerantSearch = await searchBenchmark(db, {
+    term: "belve",
+    properties: "*",
+    tolerance: 2,
+  });
+  console.log(
+    `Typo-tolerant search for "belve" through 1M entries in all indices: ${typoTolerantSearch}`
+  );
+
   const searchOnSpecificIndex = await searchBenchmark(db, {
     term: "believe",
     properties: ["title"],
@@ -67,6 +87,7 @@ async function main() {
   const searchOnSpecificIndex3 = await searchBenchmark(db, {
     term: "musical",
     properties: ["category"],
+    exact: true,
   });
   console.log(
     `Searching "musical" through 1M entries in the "category" index: ${searchOnSpecificIndex3}`
@@ -82,7 +103,7 @@ async function main() {
 }
 
 async function searchBenchmark(db, query) {
-  const results = Array.from({ length: 100_000 });
+  const results = Array.from({ length: 1000 });
 
   for (let i = 0; i < results.length; i++) {
     const { elapsed } = await db.search(query);

packages/benchmarks/engines/exact-search.js

@@ -0,0 +1,31 @@
+import cronometro from "cronometro";
+import { Lyra } from "@nearform/lyra";
+import lines from "../dataset/divinaCommedia.json" assert { type: "json" };
+
+const db = new Lyra({
+  schema: {
+    id: "string",
+    txt: "string",
+  },
+});
+
+for (const line of lines) {
+  await db.insert(line);
+}
+
+const testCases = {
+  ['Lyra exact search. Searching "comandamento" in Divina Commedia, "txt" index']() {
+    return db.search({
+      term: "comandamento",
+      properties: ["txt"],
+      exact: true,
+    });
+  },
+  ['Lyra exact search. Searching "incominciai" in Divina Commedia all indexes']() {
+    return db.search({ term: "incominciai", exact: true });
+  },
+};
+
+cronometro(testCases);
+
+export default testCases;

packages/benchmarks/engines/prefix-search-movies.js

@@ -2,7 +2,7 @@
 
 import fs from "fs";
 import cronometro from "cronometro";
-import { parse } from "csv-parse";
+import readline from "readline";
 import { Lyra } from "@nearform/lyra";
 
 const db = new Lyra({
@@ -15,22 +15,24 @@ const db = new Lyra({
 
 function populateDB() {
   console.log("Populating the database...");
-  return new Promise((resolve) => {
-    fs.createReadStream("./dataset/title.csv")
-      .pipe(parse({ delimiter: ";", from_line: 2 }))
-      .on("data", (row) => {
-        const [, type, title, , , , , , category] = row;
+  return new Promise(async (resolve) => {
+    const fileStream = fs.createReadStream("./dataset/title.tsv");
+    const rl = readline.createInterface({
+      input: fileStream,
+      crlfDelay: Infinity,
+    });
+
+    for await (const row of rl) {
+      const [, type, title, , , , , , category] = row.split("\t");
 
-        db.insert({
-          type,
-          title,
-          category,
-        });
-      })
-      .on("end", () => {
-        console.log("Database ready");
-        resolve(1);
+      db.insert({
+        type,
+        title,
+        category,
       });
+    }
+
+    resolve(1);
   });
 }
 

packages/benchmarks/engines/typo-tolerance.js

@@ -0,0 +1,32 @@
+import cronometro from "cronometro";
+import { Lyra } from "@nearform/lyra";
+import lines from "../dataset/divinaCommedia.json" assert { type: "json" };
+
+const db = new Lyra({
+  schema: {
+    id: "string",
+    txt: "string",
+  },
+});
+
+for (const line of lines) {
+  await db.insert(line);
+}
+
+const testCases = {
+  ['Lyra typo-tolerant search. Searching "confonderi" in Divina Commedia, "txt" index']() {
+    return db.search({
+      term: "confonderi",
+      properties: ["txt"],
+      exact: true,
+      tolerance: 1,
+    });
+  },
+  ['Lyra typo-tolerant search. Searching "confondersi" in Divina Commedia all indexes']() {
+    return db.search({ term: "confondersi", exact: true, tolerance: 2 });
+  },
+};
+
+cronometro(testCases);
+
+export default testCases;

packages/benchmarks/get-imdb-dataset.mjs

@@ -0,0 +1,11 @@
+#!/usr/bin/env zx
+
+const datasetName = `title.basics.tsv`;
+const dataset = `https://datasets.imdbws.com/${datasetName}`;
+const datasetOutputDir = `./dataset`;
+const datasetNewName = `title.tsv`
+
+await $`wget ${dataset} -O ${datasetOutputDir}/${datasetName}.gz`;
+await $`gunzip ${datasetOutputDir}/${datasetName}`;
+await $`head -1000000 ${datasetOutputDir}/${datasetName} >> ${datasetOutputDir}/${datasetNewName}`
+await $`rm ${datasetOutputDir}/${datasetName}.gz`

packages/benchmarks/index.js

@@ -1,8 +1,12 @@
 import cronometro from "cronometro";
 import prefixSearch from "./engines/prefix-search.js";
-import prefixSearchMovies from "./engines/prefix-search-movies.js";
+import exactSearch from "./engines/exact-search.js";
 import indexing from "./engines/indexing.js";
+import typoTolerance from "./engines/typo-tolerance.js";
+import prefixSearchMovies from "./engines/prefix-search-movies.js";
 
-cronometro(prefixSearch);
 cronometro(indexing);
+cronometro(prefixSearch);
+cronometro(exactSearch);
+cronometro(typoTolerance);
 cronometro(prefixSearchMovies);