audit: drop the audit_links table (#416)

drizzle-pg/0015_sticky_dagger.sql

@@ -0,0 +1 @@
+DROP TABLE "audit_links" CASCADE;

drizzle-pg/meta/_journal.json

@@ -106,6 +106,13 @@
       "when": 1784588248793,
       "tag": "0014_solid_centennial",
       "breakpoints": true
+    },
+    {
+      "idx": 15,
+      "version": "7",
+      "when": 1784610564142,
+      "tag": "0015_sticky_dagger",
+      "breakpoints": true
     }
   ]
 }

drizzle/0038_happy_steel_serpent.sql

@@ -0,0 +1 @@
+DROP TABLE `audit_links`;

drizzle/meta/_journal.json

@@ -267,6 +267,13 @@
       "when": 1784588247803,
       "tag": "0037_small_caretaker",
       "breakpoints": true
+    },
+    {
+      "idx": 38,
+      "version": "6",
+      "when": 1784610563184,
+      "tag": "0038_happy_steel_serpent",
+      "breakpoints": true
     }
   ]
 }

src/db/audit.schema.ts

@@ -123,36 +123,8 @@ export const auditPages = sqliteTable(
   (table) => [index("audit_pages_audit_url_idx").on(table.auditId, table.url)],
 );
 
-// One row per unique (source page, target URL) link edge. Currently only
-// internal edges are stored (see AuditRepository); isInternal stays so
-// external-link checks can start writing rows without a migration.
-export const auditLinks = sqliteTable(
-  "audit_links",
-  {
-    id: text("id").primaryKey(),
-    auditId: text("audit_id")
-      .notNull()
-      .references(() => audits.id, { onDelete: "cascade" }),
-    sourcePageId: text("source_page_id")
-      .notNull()
-      .references(() => auditPages.id, { onDelete: "cascade" }),
-    sourceUrl: text("source_url").notNull(),
-    targetUrl: text("target_url").notNull(),
-    anchor: text("anchor"),
-    isInternal: integer("is_internal", { mode: "boolean" })
-      .notNull()
-      .default(true),
-    isNofollow: integer("is_nofollow", { mode: "boolean" })
-      .notNull()
-      .default(false),
-  },
-  (table) => [
-    index("audit_links_audit_target_idx").on(table.auditId, table.targetUrl),
-    // Cascade path from audit_pages deletes; without this every page delete
-    // seq-scans the (large) links table.
-    index("audit_links_source_page_id_idx").on(table.sourcePageId),
-  ],
-);
+// Link edges live in the per-audit AuditScratchpad Durable Object for the
+// duration of the crawl; they are never persisted to the app DB.
 
 // One row per (issue type, affected page)
 export const auditIssues = sqliteTable(

src/db/pg/audit.schema.ts

@@ -122,32 +122,8 @@ export const auditPages = pgTable(
   (table) => [index("audit_pages_audit_url_idx").on(table.auditId, table.url)],
 );
 
-// One row per unique (source page, target URL) link edge. Currently only
-// internal edges are stored (see AuditRepository); isInternal stays so
-// external-link checks can start writing rows without a migration.
-export const auditLinks = pgTable(
-  "audit_links",
-  {
-    id: text("id").primaryKey(),
-    auditId: text("audit_id")
-      .notNull()
-      .references(() => audits.id, { onDelete: "cascade" }),
-    sourcePageId: text("source_page_id")
-      .notNull()
-      .references(() => auditPages.id, { onDelete: "cascade" }),
-    sourceUrl: text("source_url").notNull(),
-    targetUrl: text("target_url").notNull(),
-    anchor: text("anchor"),
-    isInternal: boolean("is_internal").notNull().default(true),
-    isNofollow: boolean("is_nofollow").notNull().default(false),
-  },
-  (table) => [
-    index("audit_links_audit_target_idx").on(table.auditId, table.targetUrl),
-    // Cascade path from audit_pages deletes; without this every page delete
-    // seq-scans the (large) links table.
-    index("audit_links_source_page_id_idx").on(table.sourcePageId),
-  ],
-);
+// Link edges live in the per-audit AuditScratchpad Durable Object for the
+// duration of the crawl; they are never persisted to the app DB.
 
 // One row per (issue type, affected page)
 export const auditIssues = pgTable(

src/db/schema.ts

@@ -77,7 +77,6 @@ export const {
   backlinkSnapshots,
   audits,
   auditPages,
-  auditLinks,
   auditIssues,
   auditLighthouseResults,
   samSessions,