belliottsmith commented on code in PR #106:
URL: https://github.com/apache/cassandra-accord/pull/106#discussion_r1692254593


##########
accord-core/src/main/java/accord/local/SafeCommandStore.java:
##########
@@ -214,58 +214,64 @@ private void updateMaxConflicts(Command prev, Command 
updated)
         commandStore().updateMaxConflicts(prev, updated);
     }
 
-    private void updateCommandsForKey(Command prev, Command updated)
+    private void updateCommandsForKey(Command prev, Command next)
     {
-        if (!CommandsForKey.needsUpdate(prev, updated))
+        if (!CommandsForKey.needsUpdate(prev, next))
             return;
 
-        TxnId txnId = updated.txnId();
-        Keys keys;
-        if (txnId.domain().isKey() && txnId.kind().isGloballyVisible())
+        TxnId txnId = next.txnId();
+        if (CommandsForKey.manages(txnId)) updateManagedCommandsForKey(this, 
prev, next);
+        if (!CommandsForKey.managesExecution(txnId) && 
next.hasBeen(Status.Stable) && !next.hasBeen(Status.Truncated) && 
!prev.hasBeen(Status.Stable))
+            updateUnmanagedExecutionCommandsForKey(this, next);
+    }
+
+    private static void updateManagedCommandsForKey(SafeCommandStore 
safeStore, Command prev, Command next)
+    {
+        TxnId txnId = next.txnId();
+        Keys keys = (Keys)next.keysOrRanges();
+        if (keys == null || next.hasBeen(Status.Truncated)) keys = 
(Keys)prev.keysOrRanges();
+        if (keys == null)
+            return;
+
+        // TODO (required): additionalKeysOrRanges may not be being handled 
entirely correctly here, though it may not matter.
+        //    Once committed without a given key, we should be effectively 
erasing the command from that CFK
+        PreLoadContext context = PreLoadContext.contextFor(txnId, keys, 
COMMANDS);
+        // TODO (expected): execute immediately for any keys we already have 
loaded, and save only those we haven't for later

Review Comment:
   Because it's better to perform the update immediately, where possible? 
Especially as any continuation goes to the back of the processing queue, which 
might cause everything to take a lot longer (e.g. there may be some command 
that runs next that cannot now execute because it's waiting for the last 
command to update CFK, so we now have at least two additional continuations and 
queue delays)



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to