Change ID generation strategy (#69200)

Previously, IDs were generated by hashing and taking the 4 least significant bits, and iteratively taking 4 more bits into account until an available ID was found. Now, the same is done but with digits, i.e. take the first digit of the hash, then first 2, then first 3, etc until an available ID is found.
vercel · Aug 27, 2024 · 59183b9 · 59183b9
1 parent f133f28
commit 59183b9
Show file tree

Hide file tree

Showing 2 changed files with 15 additions and 10 deletions.
diff --git a/turbopack/crates/turbopack-ecmascript/src/global_module_id_strategy.rs b/turbopack/crates/turbopack-ecmascript/src/global_module_id_strategy.rs
@@ -190,20 +190,25 @@ pub async fn process_module(
         return Ok(());
     }
 
-    let mut masked_hash = full_hash & 0xF;
-    let mut mask = 0xF;
-    while used_ids.contains(&masked_hash) {
-        if mask == 0xFFFFFFFFFFFFFFFF {
-            return Err(anyhow::anyhow!("This is a... 64-bit hash collision?"));
+    let mut trimmed_hash = full_hash % 10;
+    let mut power = 10;
+    while used_ids.contains(&trimmed_hash) {
+        if power >= u64::MAX / 10 {
+            // We don't want to take the next power as it would overflow
+            if used_ids.contains(&full_hash) {
+                return Err(anyhow::anyhow!("This is a... 64-bit hash collision?"));
+            }
+            trimmed_hash = full_hash;
+            break;
         }
-        mask = (mask << 4) | 0xF;
-        masked_hash = full_hash & mask;
+        power *= 10;
+        trimmed_hash = full_hash % power;
     }
 
-    let hashed_module_id = ModuleId::String(masked_hash.to_string().into());
+    let hashed_module_id = ModuleId::String(trimmed_hash.to_string().into());
 
     id_map.insert(ident_str, hashed_module_id.cell());
-    used_ids.insert(masked_hash);
+    used_ids.insert(trimmed_hash);
 
     Ok(())
 }
diff --git a/turbopack/crates/turbopack-ecmascript/src/references/external_module.rs b/turbopack/crates/turbopack-ecmascript/src/references/external_module.rs
@@ -245,7 +245,7 @@ impl EcmascriptChunkItem for CachedExternalModuleChunkItem {
     }
 }
 
-/// A module that only has an ident and no content. This is used
+/// A module that only has an ident and no content nor references. It is used
 /// to include a module's ident in the module graph before the module
 /// itself is resolved, as is the case with NextServerComponentModule's
 /// "client modules" and "client modules ssr".