feat: support parallelize scrapers (daniel-hauser#356)

* [email protected] * patch israeli-bank-scrapers to allow passing a browser context * Use a single browser (and context) for all scrapers * Better logs * Use a separate browser context for each scraper * Move ScraperOptions to own var * Extract `scrapeAccount` * Add option to run parallel scrapers
gczobel · Oct 16, 2024 · c14863a · c14863a
1 parent d19cd02
commit c14863a
Show file tree

Hide file tree

Showing 10 changed files with 503 additions and 151 deletions.
diff --git a/.github/workflows/scrape.yml b/.github/workflows/scrape.yml
@@ -11,10 +11,14 @@ on:
         default: "30"
         required: false
         description: "Days back to scrape"
-      # worksheetName:
-      #   default: "auto_moneyman"
-      #   required: false
-      #   description: "The name of the worksheet to write to"
+      worksheetName:
+        default: "auto_moneyman"
+        required: false
+        description: "The name of the worksheet to write to"
+      parallelScrapes:
+        default: "1"
+        required: false
+        description: "Number of parallel scrapes to run"
   schedule:
     - cron: "0 22 * * *"
 env:
@@ -61,6 +65,7 @@ jobs:
           -e BUXFER_ACCOUNTS
           -e TRANSACTION_HASH_TYPE
           -e WEB_POST_URL
+          -e MAX_PARALLEL_SCRAPERS
           ${{ env.REGISTRY }}/${{ steps.normalize-repository-name.outputs.repository }}:latest
         env:
           DEBUG: ${{ vars.DEBUG }}
@@ -90,3 +95,4 @@ jobs:
           BUXFER_ACCOUNTS: ${{ secrets.BUXFER_ACCOUNTS }}
           TRANSACTION_HASH_TYPE: ${{ secrets.TRANSACTION_HASH_TYPE }}
           WEB_POST_URL: ${{ secrets.WEB_POST_URL }}
+          MAX_PARALLEL_SCRAPERS: ${{ github.event.inputs.parallelScrapes }}
diff --git a/README.md b/README.md
@@ -97,6 +97,7 @@ Example:
 | `TRANSACTION_HASH_TYPE`     | ``                 | The hash type to use for the transaction hash. Can be `moneyman` or empty. The default will be changed to `moneyman` in the upcoming versions |
 | `HIDDEN_DEPRECATIONS`       | ''                 | A comma separated list of deprecations to hide                                                                                                |
 | `PUPPETEER_EXECUTABLE_PATH` | `undefined`        | An ExecutablePath for the scraper. if undefined defaults to system.                                                                           |
+| `MAX_PARALLEL_SCRAPERS`     | `1`                | The maximum number of parallel scrapers to run                                                                                                |
 
 ### Get notified in telegram
 

diff --git a/package-lock.json b/package-lock.json
diff --git a/package.json b/package.json
@@ -28,6 +28,7 @@
   },
   "homepage": "https://github.com/daniel-hauser/moneyman#readme",
   "dependencies": {
+    "async": "^3.2.6",
     "azure-kusto-data": "^6.0.2",
     "azure-kusto-ingest": "^6.0.2",
     "buxfer-ts-client": "^1.1.0",
@@ -38,11 +39,12 @@
     "google-auth-library": "^9.14.1",
     "google-spreadsheet": "^4.1.4",
     "hash-it": "^6.0.0",
-    "israeli-bank-scrapers": "^5.1.4",
+    "israeli-bank-scrapers": "^5.2.0",
     "telegraf": "^4.16.3",
     "ynab": "^2.5.0"
   },
   "devDependencies": {
+    "@types/async": "^3.2.24",
     "@types/debug": "^4.1.12",
     "@types/jest": "^29.5.13",
     "husky": "^9.1.6",

diff --git a/patches/israeli-bank-scrapers+5.2.0.patch b/patches/israeli-bank-scrapers+5.2.0.patch
diff --git a/src/browser.ts b/src/browser.ts
@@ -0,0 +1,21 @@
+import puppeteer, {
+  type Browser,
+  type PuppeteerLaunchOptions,
+} from "puppeteer";
+import { createLogger } from "./utils/logger.js";
+
+export const browserArgs = ["--disable-dev-shm-usage", "--no-sandbox"];
+export const browserExecutablePath =
+  process.env.PUPPETEER_EXECUTABLE_PATH || undefined;
+
+const logger = createLogger("browser");
+
+export async function createBrowser(): Promise<Browser> {
+  const options = {
+    args: browserArgs,
+    executablePath: browserExecutablePath,
+  } satisfies PuppeteerLaunchOptions;
+
+  logger("Creating browser", options);
+  return puppeteer.launch(options);
+}
diff --git a/src/config.ts b/src/config.ts
@@ -25,6 +25,7 @@ const {
   BUXFER_ACCOUNTS = "",
   TRANSACTION_HASH_TYPE = "",
   WEB_POST_URL = "",
+  MAX_PARALLEL_SCRAPERS = "",
 } = process.env;
 
 /**
@@ -34,6 +35,7 @@ export const daysBackToScrape = DAYS_BACK || 10;
 export const worksheetName = WORKSHEET_NAME || "_moneyman";
 export const futureMonthsToScrape = parseInt(FUTURE_MONTHS, 10);
 export const systemTimezone = Intl.DateTimeFormat().resolvedOptions().timeZone;
+export const parallelScrapers = MAX_PARALLEL_SCRAPERS || 1;
 
 const accountsToScrape = ACCOUNTS_TO_SCRAPE.split(",")
   .filter(Boolean)

diff --git a/src/data/index.ts b/src/data/index.ts
@@ -2,8 +2,15 @@ import { performance } from "perf_hooks";
 import { getAccountTransactions } from "./scrape.js";
 import { AccountConfig, AccountScrapeResult } from "../types.js";
 import { createLogger } from "../utils/logger.js";
+import { createBrowser } from "../browser.js";
+import { send, sendError } from "../notifier.js";
+import { getFailureScreenShotPath } from "../utils/failureScreenshot.js";
+import { ScraperOptions } from "israeli-bank-scrapers";
+import { parallelScrapers } from "../config.js";
+import { parallelLimit } from "async";
+import os from "node:os";
 
-const logger = createLogger("data");
+const logger = createLogger("scraper");
 
 export async function scrapeAccounts(
   accounts: Array<AccountConfig>,
@@ -18,67 +25,74 @@ export async function scrapeAccounts(
 
   logger(`scraping %d accounts`, accounts.length);
   logger(`start date %s`, startDate.toISOString());
+
+  let futureMonths: number | undefined = undefined;
   if (!Number.isNaN(futureMonthsToScrape)) {
     logger(`months to scrap: %d`, futureMonthsToScrape);
+    futureMonths = futureMonthsToScrape;
   }
 
   const status: Array<string> = [];
-  const results: Array<AccountScrapeResult> = [];
-
-  for (let i = 0; i < accounts.length; i++) {
-    const account = accounts[i];
-
-    logger(`scraping account #${i} (type=${account.companyId})`);
-    const result = await scrapeAccount(
-      account,
-      startDate,
-      futureMonthsToScrape,
-      async (message) => {
-        status[i] = message;
-        await scrapeStatusChanged?.(status);
-      },
-    );
 
-    results.push({
-      companyId: account.companyId,
-      result,
-    });
-    logger(`scraping account #${i} ended`);
+  logger("Creating a browser");
+  const browser = await createBrowser();
+  logger(`Browser created, starting to scrape ${accounts.length} accounts`);
+
+  if (Number(parallelScrapers) > 1) {
+    logger(`Running with ${parallelScrapers} parallel scrapers`);
+    send(
+      `System info: ${JSON.stringify(
+        {
+          parallelScrapers: Number(parallelScrapers),
+          availableParallelism: os.availableParallelism(),
+          totalMemoryGB: (os.totalmem() / 1000000000).toFixed(2),
+          freeMemoryGB: (os.freemem() / 1000000000).toFixed(2),
+          cpus: os.cpus().length,
+        },
+        null,
+        2,
+      )}`,
+    );
   }
 
-  logger(`scraping ended`);
-  const stats = getStats(results);
-  logger(
-    `Got ${stats.transactions} transactions from ${stats.accounts} accounts`,
+  const results = await parallelLimit<AccountConfig, AccountScrapeResult[]>(
+    accounts.map(
+      (account, i) => async () =>
+        scrapeAccount(
+          logger.extend(`#${i} (${account.companyId})`),
+          account,
+          {
+            browserContext: await browser.createBrowserContext(),
+            startDate,
+            companyId: account.companyId,
+            futureMonthsToScrape: futureMonths,
+            storeFailureScreenShotPath: getFailureScreenShotPath(
+              account.companyId,
+            ),
+          },
+          async (message, append = false) => {
+            status[i] = append ? `${status[i]} ${message}` : message;
+            return scrapeStatusChanged?.(status);
+          },
+        ),
+    ),
+    Number(parallelScrapers),
   );
 
   const duration = (performance.now() - start) / 1000;
-  logger(`total duration: ${duration}s`);
-
+  logger(`scraping ended, total duration: ${duration.toFixed(1)}s`);
   await scrapeStatusChanged?.(status, duration);
 
-  return results;
-}
-
-export async function scrapeAccount(
-  account: AccountConfig,
-  startDate: Date,
-  futureMonthsToScrape: number,
-  setStatusMessage: (message: string) => Promise<void>,
-) {
-  let message = "";
-  const start = performance.now();
-  const result = await getAccountTransactions(
-    account,
-    startDate,
-    futureMonthsToScrape,
-    (cid, step) => setStatusMessage((message = `[${cid}] ${step}`)),
-  );
+  try {
+    logger(`closing browser`);
+    await browser?.close();
+  } catch (e) {
+    sendError(e, "browser.close");
+    logger(`failed to close browser`, e);
+  }
 
-  const duration = (performance.now() - start) / 1000;
-  logger(`scraping took ${duration.toFixed(1)}s`);
-  await setStatusMessage(`${message}, took ${duration.toFixed(1)}s`);
-  return result;
+  logger(getStats(results));
+  return results;
 }
 
 function getStats(results: Array<AccountScrapeResult>) {
@@ -99,3 +113,28 @@ function getStats(results: Array<AccountScrapeResult>) {
     transactions,
   };
 }
+
+async function scrapeAccount(
+  logger: debug.IDebugger,
+  account: AccountConfig,
+  scraperOptions: ScraperOptions,
+  setStatusMessage: (message: string, append?: boolean) => Promise<void>,
+) {
+  logger(`scraping`);
+
+  const scraperStart = performance.now();
+  const result = await getAccountTransactions(
+    account,
+    scraperOptions,
+    (cid, step) => setStatusMessage(`[${cid}] ${step}`),
+  );
+
+  const duration = (performance.now() - scraperStart) / 1000;
+  logger(`scraping ended, took ${duration.toFixed(1)}s`);
+  await setStatusMessage(`, took ${duration.toFixed(1)}s`, true);
+
+  return {
+    companyId: account.companyId,
+    result,
+  };
+}
diff --git a/src/data/scrape.ts b/src/data/scrape.ts
@@ -1,29 +1,22 @@
-import { createScraper, ScraperScrapingResult } from "israeli-bank-scrapers";
+import {
+  createScraper,
+  ScraperOptions,
+  ScraperScrapingResult,
+} from "israeli-bank-scrapers";
 import { AccountConfig } from "../types.js";
 import { ScraperErrorTypes } from "israeli-bank-scrapers/lib/scrapers/errors.js";
 import { createLogger } from "../utils/logger.js";
-import { getFailureScreenShotPath } from "../utils/failureScreenshot.js";
 
 const logger = createLogger("scrape");
 
 export async function getAccountTransactions(
   account: AccountConfig,
-  startDate: Date,
-  futureMonthsToScrape: number,
+  options: ScraperOptions,
   onProgress: (companyId: string, status: string) => void,
 ): Promise<ScraperScrapingResult> {
   logger(`started`);
   try {
-    const scraper = createScraper({
-      executablePath: process.env.PUPPETEER_EXECUTABLE_PATH || undefined,
-      startDate,
-      companyId: account.companyId,
-      args: ["--disable-dev-shm-usage", "--no-sandbox"],
-      futureMonthsToScrape: Number.isNaN(futureMonthsToScrape)
-        ? undefined
-        : futureMonthsToScrape,
-      storeFailureScreenShotPath: getFailureScreenShotPath(account.companyId),
-    });
+    const scraper = createScraper(options);
 
     scraper.onProgress((companyId, { type }) => {
       logger(`[${companyId}] ${type}`);

diff --git a/src/utils/failureScreenshot.ts b/src/utils/failureScreenshot.ts
@@ -15,7 +15,7 @@ export function getFailureScreenShotPath(companyId: string) {
   }
 
   const filePath = path.join(companyDir, `${companyId}-${Date.now()}.png`);
-  logger("getFailureScreenShotPath", { filePath });
+  logger("getFailureScreenShotPath %o", filePath);
 
   return filePath;
 }