product images added to the catalogue

scratch/productimages/main.go

@@ -0,0 +1,151 @@
+// Backfills products.productimages for everything imported before the column
+// existed.
+//
+// The import used to keep `Images[0]` and discard the rest, so roughly six
+// thousand products carry one photo where the global catalogue has up to ten.
+// Re-importing would work but rewrites price, category and stock links; this
+// only fills the gallery.
+//
+// The images are re-listed rather than reconstructed. Filenames are
+// `image_000.jpeg`, `image_001.jpg` … — the extension varies within a single
+// product, so the sibling URLs cannot be derived from the first one. The
+// `image_id` is taken out of the stored URL and the same S3 listing the API
+// uses is asked for the rest, which is why this boots the app's own
+// connections instead of talking to Postgres alone.
+//
+// Dry by default; pass --apply to write.
+//
+//	go run ./scratch/productimages            # report only
+//	go run ./scratch/productimages --apply    # write
+//	go run ./scratch/productimages --apply --tenant 1087
+package main
+
+import (
+	"encoding/json"
+	"flag"
+	"fmt"
+	"log"
+	"strings"
+
+	"nearle/db"
+
+	"github.com/joho/godotenv"
+)
+
+type row struct {
+	Productid    int
+	Tenantid     int
+	Productbrand string
+	Productimage string
+}
+
+func main() {
+	apply := flag.Bool("apply", false, "write the changes; otherwise report only")
+	tenant := flag.Int("tenant", 0, "restrict to one tenant")
+	flag.Parse()
+
+	_ = godotenv.Load()
+	db.Connect()
+
+	if db.ImageStore == nil {
+		log.Fatal("the image store is not configured (USE_S3 / S3_*), so there is nothing to list images from")
+	}
+
+	q := `
+		SELECT productid, tenantid,
+		       COALESCE(productbrand,'') AS productbrand,
+		       COALESCE(productimage,'') AS productimage
+		FROM products
+		WHERE COALESCE(productimage,'') <> ''
+		  AND COALESCE(productimages::text,'') IN ('', 'null', '[]')`
+	args := []interface{}{}
+	if *tenant > 0 {
+		q += ` AND tenantid = ?`
+		args = append(args, *tenant)
+	}
+	q += ` ORDER BY productid`
+
+	var rows []row
+	if err := db.DB.Raw(q, args...).Scan(&rows).Error; err != nil {
+		log.Fatal(err)
+	}
+
+	fmt.Printf("candidates: %d\n\n", len(rows))
+
+	var filled, single, unparsed, nolist int
+	for _, r := range rows {
+		brand, imageID := parseImageRef(r.Productimage)
+		if brand == "" || imageID == "" {
+			unparsed++
+			continue
+		}
+		// The stored brand is authoritative where it exists; the URL is only a
+		// fallback, because a product renamed in the back office keeps its
+		// original image path.
+		if b := strings.ToLower(strings.TrimSpace(r.Productbrand)); b != "" {
+			brand = b
+		}
+
+		images := db.GetImages(brand, imageID)
+		if len(images) == 0 {
+			nolist++
+			continue
+		}
+		if len(images) == 1 {
+			// Nothing gained — leave the column null rather than writing a
+			// one-element array that says the same as productimage.
+			single++
+			continue
+		}
+
+		encoded, err := json.Marshal(images)
+		if err != nil {
+			log.Printf("product %d: encoding failed: %v", r.Productid, err)
+			continue
+		}
+
+		if *apply {
+			if err := db.DB.Exec(
+				`UPDATE products SET productimages = ?::jsonb, updated = NOW() WHERE productid = ?`,
+				string(encoded), r.Productid).Error; err != nil {
+				log.Printf("product %d: update failed: %v", r.Productid, err)
+				continue
+			}
+		}
+		filled++
+		if filled <= 10 {
+			fmt.Printf("  product %-7d %-10s %-34s %d images\n", r.Productid, brand, imageID, len(images))
+		}
+	}
+
+	fmt.Printf(`
+would fill (>1 image) : %d
+only one image        : %d
+image_id unparseable  : %d
+nothing listed in S3  : %d
+`, filled, single, unparsed, nolist)
+
+	if !*apply {
+		fmt.Println("\ndry run — nothing written. Re-run with --apply.")
+	}
+}
+
+// parseImageRef pulls the brand and image_id out of a stored image URL.
+//
+//	https://…/daily/brands/nestle/nestle_maggi_615a9fa9/image_000.jpeg
+//	                        ^brand  ^image_id
+//
+// Returns empty strings for anything that does not match, so a hand-uploaded
+// image URL is skipped rather than guessed at.
+func parseImageRef(url string) (brand, imageID string) {
+	const marker = "/daily/brands/"
+	i := strings.Index(url, marker)
+	if i < 0 {
+		return "", ""
+	}
+	parts := strings.Split(strings.Trim(url[i+len(marker):], "/"), "/")
+	if len(parts) < 3 {
+		return "", ""
+	}
+	return strings.ToLower(parts[0]), parts[1]
+}