ISPN-14671 HSCAN #11051
ISPN-14671 HSCAN #11051
Conversation
server/resp/src/main/java/org/infinispan/server/resp/commands/hash/HSCAN.java
Show resolved
Hide resolved
server/core/src/main/java/org/infinispan/server/iteration/HashMapIterationManager.java
Outdated
Show resolved
Hide resolved
| private CompletionStage<RespRequestHandler> iterate(Resp3Handler handler, IterationManager manager, String cursor, | ||
| IterationArguments arguments) { | ||
| if (manager == null) { | ||
| ByteBufferUtils.stringToByteBuf("*2\r\n$1\r\n0\r\n*0\r\n", handler.allocator()); |
There was a problem hiding this comment.
better to give a meaningful name to this bytes seq
| IterableIterationResult.Status status = result.getStatusCode(); | ||
| if (status == IterableIterationResult.Status.InvalidIteration) { | ||
| // Let's just return 0 | ||
| ByteBufferUtils.stringToByteBuf("*2\r\n$1\r\n0\r\n*0\r\n", handler.allocator()); |
| @Override | ||
| protected CacheStream<CacheEntry<Object, Object>> baseStream(AdvancedCache cache, IterationInitializationContext ctx) { | ||
| if (ctx == null) return EMPTY_STREAM; | ||
| if (!(ctx instanceof HashMapInitializationContext)) throw new IllegalStateException("Illegal context for iteration"); |
There was a problem hiding this comment.
how about turn ctx into a property of the specialized class, remove it from the method params and avoid this check?
There was a problem hiding this comment.
Yeah, this was something I was trying at first. I took this approach because the iteration manager is shared between all clients IIUC. So it would need a way to synchronize between retrieving the entry from the cache and invoking start. All this (trying to synchronize or add a ctx) could be avoided if the start method receives the source stream or something similar. Let me think more about this.
|
|
||
| Map<Object, Object> source = ((HashMapInitializationContext) ctx).source; | ||
| AbstractLocalCacheStream.StreamSupplier<CacheEntry<Object, Object>, Stream<CacheEntry<Object, Object>>> ss = new HashMapStreamSupplier(source); | ||
| return new LocalCacheStream<>(ss, false, SecurityActions.getCacheComponentRegistry(cache)); |
There was a problem hiding this comment.
I'm wondering if we really need a full cache stream even to iterate on an inner struct
There was a problem hiding this comment.
This one also was a small paint point. I think this one might be necessary as we can use the existing code to apply the filters. But let me think about this, too.
* Create an iterator manager that handles a HashMap source.
https://issues.redhat.com/browse/ISPN-14671
Create an iterator manager that handles a
Mapsource. The downside is we have the complete entry loaded in memory to iterate, and the entry could be huge.