From 258346e3d6ce03407c5a4b810ca5a3f669979122 Mon Sep 17 00:00:00 2001
From: Nulo <git@nulo.in>
Date: Wed, 13 Nov 2024 09:48:23 -0300
Subject: [PATCH] WIP: metadata scrapped

---
 rust/src/api/main.rs                      | 39 ++++++++++++++++++++
 sepa/db/schema.ts                         | 21 +++++++++++
 sepa/scripts/refresh-scrapped-metadata.ts | 43 +++++++++++++++++++++++
 3 files changed, 103 insertions(+)
 create mode 100644 sepa/scripts/refresh-scrapped-metadata.ts

diff --git a/rust/src/api/main.rs b/rust/src/api/main.rs
index 90816f8..902fbe4 100644
--- a/rust/src/api/main.rs
+++ b/rust/src/api/main.rs
@@ -266,6 +266,44 @@ async fn search(State(pool): State<SqlitePool>, Path(query): Path<String>) -> im
     Json(results)
 }
 
+#[derive(sqlx::FromRow, Debug, Serialize)]
+struct Metadata {
+    ean: String,
+    fetched_at: chrono::DateTime<Utc>,
+    precio_centavos: Option<i64>,
+    in_stock: Option<bool>,
+    url: String,
+    name: Option<String>,
+    image_url: Option<String>,
+}
+
+async fn dump_latest_metadata(State(pool): State<SqlitePool>) -> impl IntoResponse {
+    let precios = sqlx::query!("
+    SELECT p.id, p.ean, p.name, p.image_url, p.url, p.precio_centavos, p.in_stock, p.fetched_at
+    FROM precios p
+    INNER JOIN (
+        SELECT ean, MAX(fetched_at) as max_fetched_at
+        FROM precios
+        GROUP BY ean
+    ) latest ON p.ean = latest.ean AND p.fetched_at = latest.max_fetched_at
+    WHERE p.name IS NOT NULL")
+    .fetch_all(&pool)
+    .await
+    .unwrap()
+    .into_iter()
+    .map(|r| Metadata {
+        ean: r.ean,
+        fetched_at: DateTime::from_timestamp(r.fetched_at, 0).unwrap(),
+        image_url: r.image_url,
+        name: r.name,
+        in_stock: r.in_stock.map(|x| x == 1),
+        precio_centavos: r.precio_centavos,
+        url: r.url,
+    })
+    .collect_vec();
+    Json(precios)
+}
+
 async fn get_info(State(pool): State<SqlitePool>) -> impl IntoResponse {
     #[derive(Serialize)]
     struct Info {
@@ -321,6 +359,7 @@ async fn main() {
         .route("/api/0/ean/:ean/history", get(get_product_history))
         .route("/api/0/info", get(get_info))
         .route("/api/0/search/:query", get(search))
+        .route("/api/0/internal/latest-metadata", get(dump_latest_metadata))
         .with_state(pool);
 
     let listener = tokio::net::TcpListener::bind("0.0.0.0:8000").await.unwrap();
diff --git a/sepa/db/schema.ts b/sepa/db/schema.ts
index efd4aba..192375e 100644
--- a/sepa/db/schema.ts
+++ b/sepa/db/schema.ts
@@ -11,6 +11,8 @@ import {
   index,
   pgMaterializedView,
   pgView,
+  timestamp,
+  boolean,
 } from "drizzle-orm/pg-core";
 
 export const datasets = pgTable(
@@ -211,3 +213,22 @@ export const productos_descripcion_index = pgTable(
     ).using("gin", sql`to_tsvector('spanish', ${table.productos_descripcion})`),
   })
 );
+
+// véase scripts/refresh-scrapped-metadata.ts
+export const productos_metadata_scrapped = pgTable(
+  "productos_metadata_scrapped",
+  {
+    ean: bigint("ean", { mode: "bigint" }),
+    fetchedAt: timestamp("fetched_at").notNull(),
+    precioCentavos: integer("precio_centavos"),
+    inStock: boolean("in_stock"),
+    url: text("url").notNull(),
+    name: text("name"),
+    imageUrl: text("image_url"),
+  },
+  (table) => ({
+    productos_metadata_scrapped_ean_idx: index(
+      "productos_metadata_scrapped_ean_idx"
+    ).on(table.ean),
+  })
+);
diff --git a/sepa/scripts/refresh-scrapped-metadata.ts b/sepa/scripts/refresh-scrapped-metadata.ts
new file mode 100644
index 0000000..4f529dd
--- /dev/null
+++ b/sepa/scripts/refresh-scrapped-metadata.ts
@@ -0,0 +1,43 @@
+/**
+ * este script actualiza la base de datos "nueva" a partir de una base de datos
+ * generada por el scraper "viejo" de preciazo, que scrapea los sitios de los supermercados.
+ *
+ * solo guarda los últimos metadatos de cada producto.
+ *
+ * se le pasa la base de datos SQLite del scraper como parametro.
+ */
+
+import { drizzle } from "drizzle-orm/postgres-js";
+import postgres from "postgres";
+import * as schema from "../db/schema";
+import { Database } from "bun:sqlite";
+
+if (!process.argv[2]) {
+  console.error("falta pasar la base de datos del scraper como parametro");
+  process.exit(1);
+}
+
+const db = drizzle(postgres(), {
+  schema,
+  logger: true,
+});
+using scraperDb = new Database(process.argv[2], {
+  strict: true,
+  readonly: true,
+});
+
+const precios = scraperDb.query(`
+  SELECT p.id, p.ean, p.name, p.image_url, p.url, p.precio_centavos, p.in_stock, p.fetched_at
+  FROM precios p
+  INNER JOIN (
+      SELECT ean, MAX(fetched_at) as max_fetched_at
+      FROM precios
+      GROUP BY ean
+  ) latest ON p.ean = latest.ean AND p.fetched_at = latest.max_fetched_at
+  WHERE p.name IS NOT NULL
+`);
+
+// @ts-expect-error bun 1.1.30 has outdated types, it's fixed in main branch
+for (const row of precios.iterate()) {
+  console.log(row);
+}