scrapear urls a BD

coto-link-scraper/index.ts

@@ -1,23 +1,24 @@
 import { getHtml } from "../scraper/fetch.js";
 import { parseHTML } from "linkedom";
 import PQueue from "p-queue";
+import { saveUrls } from "db-datos/urlHelpers.js";
 
-// let fetched = new Set<string>();
+export async function scrapCotoProducts() {
-{
   const initial =
     "https://www.cotodigital3.com.ar/sitios/cdigi/browse?Nf=product.endDate%7CGTEQ+1.7032032E12%7C%7Cproduct.startDate%7CLTEQ+1.7032032E12&No=2200&Nr=AND%28product.sDisp_200%3A1004%2Cproduct.language%3Aespa%C3%B1ol%2COR%28product.siteId%3ACotoDigital%29%29&Nrpp=200";
 
-  const queue = new PQueue({ concurrency: 2 });
+  const queue = new PQueue({ concurrency: 4 });
 
   const pageSize = 300; // hasta 1000
-  const links = Array.from({ length: Math.ceil(29000 / 300) }, (x, i) => i).map(
+  const links = Array.from(
-    (i) => {
+    { length: Math.ceil(29000 / pageSize) },
+    (x, i) => i
+  ).map((i) => {
     const url = new URL(initial);
     url.searchParams.set("No", `${i * pageSize}`);
     url.searchParams.set("Nrpp", `${pageSize}`);
     return url.toString();
-    }
+  });
-  );
 
   const promises = links.map((l) => queue.add(getPage(l)));
   await Promise.all(promises);
@@ -38,22 +39,6 @@ function getPage(url: string) {
       document.querySelectorAll<HTMLAnchorElement>(".product_info_container a"),
       (a) => new URL(a.href, url).toString()
     );
-    hrefs.forEach((h) => process.stdout.write(h + "\n"));
+    saveUrls(hrefs);
-
-    // const nextLinks = Array.from(
-    //   document.querySelectorAll<HTMLAnchorElement>(
-    //     "#atg_store_pagination a[href]"
-    //   ),
-    //   (a) => new URL(a.href, url).toString()
-    // );
-
-    // await Promise.all(
-    //   nextLinks
-    //     .filter((l) => !fetched.has(l))
-    //     .map((l) => {
-    //       fetched.add(l);
-    //       return queue.add(getPage(l));
-    //     })
-    // );
   };
 }

db-datos/db.ts

@@ -0,0 +1,10 @@
+import { Database } from "bun:sqlite";
+import { drizzle } from "drizzle-orm/bun-sqlite";
+import { DB_PATH } from "./drizzle.config.js";
+import { migrateDb } from "./migrate.js";
+import * as schema from "./schema.js";
+
+migrateDb();
+
+export const sqlite = new Database(DB_PATH);
+export const db = drizzle(sqlite, { schema });

db-datos/drizzle/0009_breezy_forge.sql

@@ -0,0 +1,8 @@
+CREATE TABLE `producto_urls` (
+	`id` integer PRIMARY KEY AUTOINCREMENT NOT NULL,
+	`url` text NOT NULL,
+	`first_seen` integer NOT NULL,
+	`last_seen` integer NOT NULL
+);
+--> statement-breakpoint
+CREATE UNIQUE INDEX `producto_urls_url_unique` ON `producto_urls` (`url`);

db-datos/drizzle/meta/0009_snapshot.json

@@ -0,0 +1,146 @@
+{
+  "version": "5",
+  "dialect": "sqlite",
+  "id": "2e398920-ffaf-4d55-ae13-d906cb9e0efa",
+  "prevId": "082630a9-3744-4e33-bde5-06045ca57d36",
+  "tables": {
+    "precios": {
+      "name": "precios",
+      "columns": {
+        "id": {
+          "name": "id",
+          "type": "integer",
+          "primaryKey": true,
+          "notNull": true,
+          "autoincrement": true
+        },
+        "ean": {
+          "name": "ean",
+          "type": "text",
+          "primaryKey": false,
+          "notNull": true,
+          "autoincrement": false
+        },
+        "fetched_at": {
+          "name": "fetched_at",
+          "type": "integer",
+          "primaryKey": false,
+          "notNull": true,
+          "autoincrement": false
+        },
+        "precio_centavos": {
+          "name": "precio_centavos",
+          "type": "integer",
+          "primaryKey": false,
+          "notNull": false,
+          "autoincrement": false
+        },
+        "in_stock": {
+          "name": "in_stock",
+          "type": "integer",
+          "primaryKey": false,
+          "notNull": false,
+          "autoincrement": false
+        },
+        "url": {
+          "name": "url",
+          "type": "text",
+          "primaryKey": false,
+          "notNull": true,
+          "autoincrement": false
+        },
+        "warc_record_id": {
+          "name": "warc_record_id",
+          "type": "text",
+          "primaryKey": false,
+          "notNull": false,
+          "autoincrement": false
+        },
+        "parser_version": {
+          "name": "parser_version",
+          "type": "integer",
+          "primaryKey": false,
+          "notNull": false,
+          "autoincrement": false
+        },
+        "name": {
+          "name": "name",
+          "type": "text",
+          "primaryKey": false,
+          "notNull": false,
+          "autoincrement": false
+        },
+        "image_url": {
+          "name": "image_url",
+          "type": "text",
+          "primaryKey": false,
+          "notNull": false,
+          "autoincrement": false
+        }
+      },
+      "indexes": {
+        "precios_ean_idx": {
+          "name": "precios_ean_idx",
+          "columns": [
+            "ean"
+          ],
+          "isUnique": false
+        }
+      },
+      "foreignKeys": {},
+      "compositePrimaryKeys": {},
+      "uniqueConstraints": {}
+    },
+    "producto_urls": {
+      "name": "producto_urls",
+      "columns": {
+        "id": {
+          "name": "id",
+          "type": "integer",
+          "primaryKey": true,
+          "notNull": true,
+          "autoincrement": true
+        },
+        "url": {
+          "name": "url",
+          "type": "text",
+          "primaryKey": false,
+          "notNull": true,
+          "autoincrement": false
+        },
+        "first_seen": {
+          "name": "first_seen",
+          "type": "integer",
+          "primaryKey": false,
+          "notNull": true,
+          "autoincrement": false
+        },
+        "last_seen": {
+          "name": "last_seen",
+          "type": "integer",
+          "primaryKey": false,
+          "notNull": true,
+          "autoincrement": false
+        }
+      },
+      "indexes": {
+        "producto_urls_url_unique": {
+          "name": "producto_urls_url_unique",
+          "columns": [
+            "url"
+          ],
+          "isUnique": true
+        }
+      },
+      "foreignKeys": {},
+      "compositePrimaryKeys": {},
+      "uniqueConstraints": {}
+    }
+  },
+  "enums": {},
+  "_meta": {
+    "schemas": {},
+    "tables": {},
+    "columns": {}
+  }
+}

db-datos/drizzle/meta/_journal.json

@@ -64,6 +64,13 @@
       "when": 1703807460152,
       "tag": "0008_funny_nighthawk",
       "breakpoints": true
+    },
+    {
+      "idx": 9,
+      "version": "5",
+      "when": 1703895109501,
+      "tag": "0009_breezy_forge",
+      "breakpoints": true
     }
   ]
 }

db-datos/schema.ts

@@ -22,3 +22,12 @@ export const precios = sqliteTable(
 );
 
 export type Precio = typeof precios.$inferSelect;
+
+export const productoUrls = sqliteTable("producto_urls", {
+  id: integer("id", { mode: "number" }).primaryKey({ autoIncrement: true }),
+  url: text("url").unique().notNull(),
+  firstSeen: integer("first_seen", { mode: "timestamp" }).notNull(),
+  lastSeen: integer("last_seen", { mode: "timestamp" }).notNull(),
+});
+
+export type ProductUrl = typeof productoUrls.$inferSelect;

db-datos/urlHelpers.ts

@@ -0,0 +1,25 @@
+import { sql } from "drizzle-orm";
+import { db } from "./db.js";
+import { productoUrls } from "./schema.js";
+
+export function saveUrls(urls: string[]) {
+  db.transaction((tx) => {
+    const now = new Date();
+    const insertUrlTra = tx
+      .insert(productoUrls)
+      .values({
+        url: sql.placeholder("url"),
+        firstSeen: now,
+        lastSeen: now,
+      })
+      .onConflictDoUpdate({
+        target: productoUrls.url,
+        set: { lastSeen: now },
+      })
+      .prepare();
+
+    for (const href of urls) {
+      insertUrlTra.run({ url: href });
+    }
+  });
+}

dia-link-scraper/index.ts

@@ -1,7 +1,8 @@
 import pMap from "p-map";
 import { parseHTML } from "linkedom";
 import { getHtml } from "../scraper/fetch.js";
-(async () => {
+import { saveUrls } from "db-datos/urlHelpers.js";
+
 const categorias = [
   "https://diaonline.supermercadosdia.com.ar/almacen",
   "https://diaonline.supermercadosdia.com.ar/almacen/conservas",
@@ -66,29 +67,44 @@ import { getHtml } from "../scraper/fetch.js";
   "https://diaonline.supermercadosdia.com.ar/2089?map=productClusterIds&order=OrderByBestDiscountDESC",
 ];
 
-  const links = categorias.flatMap(
+export async function scrapDiaProducts() {
-    (link) =>
+  await Promise.all([scrapBySite(), scrapBySitemap()]);
+}
+
+async function scrapBySitemap() {
+  // de https://diaonline.supermercadosdia.com.ar/sitemap.xml
+  const sitemaps = [
+    "https://diaonline.supermercadosdia.com.ar/sitemap/product-1.xml",
+    "https://diaonline.supermercadosdia.com.ar/sitemap/product-2.xml",
+    "https://diaonline.supermercadosdia.com.ar/sitemap/product-3.xml",
+    "https://diaonline.supermercadosdia.com.ar/sitemap/product-4.xml",
+    "https://diaonline.supermercadosdia.com.ar/sitemap/product-5.xml",
+  ];
+
+  await pMap(sitemaps, async (sitemapUrl) => {
+    const res = await fetch(sitemapUrl);
+    const xml = await res.text();
+    let urls = new Set<string>();
+    new HTMLRewriter()
+      .on("loc", {
+        text(element) {
+          const txt = element.text.trim();
+          if (!txt) return;
+          urls.add(txt);
+        },
+      })
+      .transform(new Response(xml));
+    saveUrls(Array.from(urls));
+  });
+}
+
+async function scrapBySite() {
+  const links = categorias.flatMap((link) =>
     Array.from({ length: 51 }, (x, i) => i).map((i) => {
       const url = new URL(link);
       url.searchParams.set("page", `${i}`);
       return url.toString();
     })
-
-    // el order solo carga con el frontend :(
-    // .flatMap((link) =>
-    //   [
-    //     "OrderByNameASC",
-    //     "OrderByNameDESC",
-    //     "OrderByTopSaleDESC",
-    //     "OrderByPriceDESC",
-    //     "OrderByPriceASC",
-    //     "",
-    //   ].map((order) => {
-    //     const url = new URL(link);
-    //     url.searchParams.set("order", order);
-    //     return url.toString();
-    //   })
-    // )
   );
 
   await pMap(
@@ -103,8 +119,8 @@ import { getHtml } from "../scraper/fetch.js";
         ),
         (a) => new URL(a.href, url).toString()
       );
-      hrefs.forEach((h) => process.stdout.write(h + "\n"));
+      saveUrls(hrefs);
     },
     { concurrency: 32 }
   );
-})();
+}

scraper/auto.ts

@@ -1,14 +1,19 @@
-import { mkdtemp, access } from "node:fs/promises";
+import { mkdtemp, access, writeFile } from "node:fs/promises";
 import { tmpdir } from "node:os";
 import { join, resolve } from "node:path";
 import { spawn } from "node:child_process";
-import { Supermercado } from "db-datos/supermercado.js";
+import { Supermercado, hosts } from "db-datos/supermercado.js";
 import PQueue from "p-queue";
 import { format, formatDuration, intervalToDuration } from "date-fns";
 import { parseWarc } from "./scrap.js";
 import { S3Client } from "@aws-sdk/client-s3";
 import { Upload } from "@aws-sdk/lib-storage";
 import { BunFile } from "bun";
+import { db } from "db-datos/db.js";
+import { like } from "drizzle-orm";
+import { productoUrls } from "db-datos/schema.js";
+import { scrapDiaProducts } from "../dia-link-scraper/index.js";
+import { scrapCotoProducts } from "../coto-link-scraper/index.js";
 
 const supermercados: Supermercado[] = [
   Supermercado.Carrefour,
@@ -71,11 +76,41 @@ class Auto {
   }
 
   async downloadList(supermercado: Supermercado) {
-    const listPath = resolve(
+    const ctxPath = await mkdtemp(join(tmpdir(), "preciazo-scraper-wget-"));
+
+    let listPath: string;
+    if (supermercado === "Carrefour") {
+      // TODO: carrefour todavía no tiene un scraper que guarde a la BD
+      listPath = resolve(
         join(process.env.LISTS_DIR ?? "../data", `${supermercado}.txt`)
       );
+    } else {
+      const t0 = performance.now();
+      switch (supermercado) {
+        case "Dia":
+          await scrapDiaProducts();
+          break;
+        case "Coto":
+          await scrapCotoProducts();
+          break;
+      }
+      this.inform(
+        `[scrapUrls[${supermercado}]] Tardó ${formatMs(performance.now() - t0)}`
+      );
+
+      listPath = join(ctxPath, `lista-${supermercado}.txt`);
+      const host = Object.entries(hosts).find(
+        ([host, supe]) => supe === supermercado
+      )![0];
+      const results = await db.query.productoUrls
+        .findMany({
+          where: like(productoUrls.url, `%${host}%`),
+        })
+        .execute();
+      const urls = results.map((r) => r.url);
+      await writeFile(listPath, urls.join("\n") + "\n");
+    }
     const date = new Date();
-    const ctxPath = await mkdtemp(join(tmpdir(), "preciazo-scraper-wget-"));
     const zstdWarcName = `${supermercado}-${format(
       date,
       "yyyy-MM-dd-HH:mm"
@@ -98,7 +133,7 @@ class Auto {
     const t0 = performance.now();
     await subproc.exited;
     this.inform(
-      `wget para ${zstdWarcName} tardó ${formatMs(performance.now() - t0)}`
+      `[wget] ${zstdWarcName} tardó ${formatMs(performance.now() - t0)}`
     );
 
     const gzippedWarcPath = join(ctxPath, "temp.warc.gz");
@@ -187,7 +222,6 @@ class Auto {
           stdio: ["pipe", null, null],
         }
       );
-      // @ts-expect-error a los types de bun no le gusta????
       decompressor.stdout.pipe(compressor.stdin);
       compressor.on("close", (code) => {
         if (code !== 0) {

scraper/fetch.ts

@@ -1,32 +1,6 @@
-import { request } from "undici";
-import { createBrotliDecompress, createUnzip } from "node:zlib";
-import { pipeline } from "node:stream/promises";
-
 export async function getHtml(url: string) {
-  const res = await request(url, {
+  const res = await fetch(url);
-    headers: {
+  return readableToBuffer(res.body!);
-      "Accept-Encoding": "gzip, deflate, br",
-    },
-    throwOnError: true,
-    bodyTimeout: 10 * 60 * 1000,
-  });
-  let output: Buffer;
-  switch (res.headers["content-encoding"]) {
-    case "gzip":
-    case "deflate":
-      output = await pipeline(res.body, createUnzip(), readableToBuffer);
-      break;
-    case "br":
-      output = await pipeline(
-        res.body,
-        createBrotliDecompress(),
-        readableToBuffer
-      );
-      break;
-    default:
-      output = await readableToBuffer(res.body);
-  }
-  return output;
 }
 
 async function readableToBuffer(source: AsyncIterable<any>) {

scraper/scrap.ts

@@ -1,5 +1,3 @@
-import { Database } from "bun:sqlite";
-import { drizzle } from "drizzle-orm/bun-sqlite";
 import * as schema from "db-datos/schema.js";
 import { WARCParser } from "warcio";
 import { writeFile } from "fs/promises";
@@ -9,17 +7,11 @@ import { getDiaProduct } from "./parsers/dia.js";
 import { getCotoProduct } from "./parsers/coto.js";
 import { join } from "path";
 import { and, eq, sql } from "drizzle-orm";
-import { DB_PATH } from "db-datos/drizzle.config.js";
+import { db } from "db-datos/db.js";
-import { migrateDb } from "db-datos/migrate.js";
 
 const DEBUG = false;
 const PARSER_VERSION = 3;
 
-migrateDb();
-
-const sqlite = new Database(DB_PATH);
-const db = drizzle(sqlite, { schema });
-
 const getPrevPrecio = db
   .select({ id: schema.precios.id })
   .from(schema.precios)