ignorar paginas status!=200

scraper/cli.ts

@@ -7,16 +7,19 @@ import { parseWarc } from "./scrap.js";
 if (process.argv[2] === "auto") {
   await auto();
 } else if (process.argv[2] === "scrap-carrefour-links") {
-  await scrapCarrefourProducts()
+  await scrapCarrefourProducts();
 } else if (process.argv[2] === "scrap-dia-links") {
-  await scrapDiaProducts()
+  await scrapDiaProducts();
 } else if (process.argv[2] === "scrap-coto-links") {
-  await scrapCotoProducts()
+  await scrapCotoProducts();
 } else if (process.argv[2] === "scrap") {
   const warcPaths = process.argv.slice(3);
   if (warcPaths.length > 0) {
     for (const path of warcPaths) {
-      await parseWarc(path);
+      const res = await parseWarc(path);
+      console.info("=======================================");
+      console.info(path, res);
+      console.info("=======================================");
     }
   } else {
     console.error("Especificá WARCs para scrapear.");

scraper/common.ts

@@ -12,6 +12,10 @@ export function priceFromMeta(dom: Window) {
   const precioCentavos = parseFloat(precioMeta) * 100;
   return precioCentavos;
 }
+export function stockFromMeta(dom: Window) {
+  const stockMeta = getMetaProp(dom, "product:availability");
+  return stockMeta === "instock";
+}
 
 function parseJsonLds(dom: Window): object[] {
   const scripts = dom.window.document.querySelectorAll(

scraper/parsers/carrefour.ts

@@ -1,20 +1,7 @@
 import { parseHTML } from "linkedom";
-import { Precioish, type Precio } from "../scrap.js";
-import { getProductJsonLd, priceFromMeta } from "../common.js";
+import { Precioish } from "../scrap.js";
+import { getProductJsonLd, priceFromMeta, stockFromMeta } from "../common.js";
 
-function getEanByTable(dom: Window): string {
-  const eanLabelEl = dom.window.document.querySelector(
-    'td[data-specification="EAN"]'
-  );
-  const eanValueEl = eanLabelEl?.parentElement?.children[1];
-  if (
-    !eanValueEl ||
-    !(eanValueEl instanceof dom.window.HTMLElement) ||
-    !eanValueEl.dataset.specification
-  )
-    throw new Error("No encontré el EAN");
-  return eanValueEl.dataset.specification;
-}
 function parseScriptJson<T>(dom: Window, varname: string): T {
   const script = dom.window.document.querySelector<HTMLTemplateElement>(
     `template[data-type="json"][data-varname="${varname}"]`
@@ -37,31 +24,27 @@ function eanFromSeedState(dom: Window): string {
   if (!productSkuJson) throw new Error("no encontré el sku en el json");
   return productSkuJson[1].ean;
 }
-function eanFromDynamicYieldScript(dom: Window): string {
-  const scriptEl = dom.window.document.querySelector(
-    `script[src^="//st.dynamicyield.com/st?"]`
-  );
-  if (!scriptEl || !(scriptEl instanceof dom.window.HTMLScriptElement))
-    throw new Error("no encuentro el script de dynamicyield");
-
-  const url = new URL(scriptEl.src);
-  const ctx = url.searchParams.get("ctx");
-  if (!ctx) throw new Error("no hay ctx");
-  return JSON.parse(ctx).data[0];
-}
 
 export function getCarrefourProduct(html: string | Buffer): Precioish {
   const dom = parseHTML(html);
 
   const precioCentavos = priceFromMeta(dom);
+  const inStock = stockFromMeta(dom);
 
   const ean = eanFromSeedState(dom);
 
+  let name, imageUrl;
+  try {
     const ld = getProductJsonLd(dom);
-  const name = ld.name;
-  const imageUrl = ld.image;
-  const inStock =
-    ld.offers.offers[0].availability === "http://schema.org/InStock";
+    name = ld.name;
+    imageUrl = ld.image;
+  } catch (error) {
+    if (inStock) {
+      throw error;
+    } else {
+      // algunas paginas sin stock no tienen json ld
+    }
+  }
 
   return {
     name,

scraper/scrap.ts

@@ -37,9 +37,8 @@ export async function parseWarc(path: string) {
     errors: { error: any; warcRecordId: string; path: string }[];
   } = { done: 0, errors: [] };
 
-  const warc = Bun.spawn(["zstd", "-do", "/dev/stdout", path], {
-    stderr: "ignore",
-  }).stdout;
+  const proc = Bun.spawn(["zstdcat", "-d", path], {});
+  const warc = proc.stdout;
   // TODO: tirar error si falla zstd
 
   const parser = new WARCParser(warc);
@@ -53,6 +52,12 @@ export async function parseWarc(path: string) {
         console.debug(`skipped ${warcRecordId}`);
         continue;
       }
+      if (record.httpHeaders?.statusCode !== 200) {
+        console.debug(
+          `skipped ${warcRecordId} because status=${record.httpHeaders?.statusCode} (!=200)`
+        );
+        continue;
+      }
       // TODO: sobreescribir si existe el mismo record-id pero con version mas bajo?
 
       const html = await record.contentText();
@@ -99,5 +104,9 @@ export async function parseWarc(path: string) {
     }
   }
 
+  if ((await proc.exited) !== 0) {
+    throw new Error("zstd tiró un error");
+  }
+
   return progress;
 }