jamesge commented on a change in pull request #972: Redis server protocol
URL: https://github.com/apache/incubator-brpc/pull/972#discussion_r353117211
##########
File path: src/brpc/policy/redis_protocol.cpp
##########
@@ -52,62 +61,348 @@ struct InputResponse : public InputMessageBase {
}
};
-// "Message" = "Response" as we only implement the client for redis.
-ParseResult ParseRedisMessage(butil::IOBuf* source, Socket* socket,
- bool /*read_eof*/, const void* /*arg*/) {
- if (source->empty()) {
- return MakeParseError(PARSE_ERROR_NOT_ENOUGH_DATA);
+static const char** ParseArgs(const RedisReply& message) {
+ if (!message.is_array() || message.size() == 0) {
+ LOG(WARNING) << "request message is not array or size equals to zero";
+ return NULL;
}
- // NOTE(gejun): PopPipelinedInfo() is actually more contended than what
- // I thought before. The Socket._pipeline_q is a SPSC queue pushed before
- // sending and popped when response comes back, being protected by a
- // mutex. Previously the mutex is shared with Socket._id_wait_list. When
- // 200 bthreads access one redis-server, ~1.5s in total is spent on
- // contention in 10-second duration. If the mutex is separated, the time
- // drops to ~0.25s. I further replaced PeekPipelinedInfo() with
- // GivebackPipelinedInfo() to lock only once(when receiving response)
- // in most cases, and the time decreases to ~0.14s.
- PipelinedInfo pi;
- if (!socket->PopPipelinedInfo(&pi)) {
- LOG(WARNING) << "No corresponding PipelinedInfo in socket";
- return MakeParseError(PARSE_ERROR_TRY_OTHERS);
+ const char** args = (const char**)
+ malloc(sizeof(const char*) * (message.size() + 1 /* NULL */));
+ for (size_t i = 0; i < message.size(); ++i) {
+ if (!message[i].is_string()) {
+ LOG(WARNING) << "request message[" << i << "] is not array";
+ free(args);
+ return NULL;
+ }
+ args[i] = message[i].c_str();
}
+ args[message.size()] = NULL;
+ return args;
+}
+
+// One redis command corresponding to one ConsumeTaskDone. Whenever user
+// has completed the process of handling command and call done->Run()
+// (read redis.h for more details), RedisConnContext::Flush() will be
+// called and flush the response to client by the order that commands arrive.
+class ConsumeTaskDone;
+
+// This class is as parsing_context in socket.
+class RedisConnContext : public SharedObject
+ , public Destroyable {
+public:
+ RedisConnContext()
+ : handler_continue(NULL)
+ , message_count(0) {}
+ ~RedisConnContext();
+ // @Destroyable
+ void Destroy();
+
+ int Init();
+ // Push `done` to a queue which is read by Flush().
+ void Push(ConsumeTaskDone* done);
+ void Flush();
+ void ClearSentDones();
+
+ SocketId socket_id;
+ RedisService::CommandMap command_map;
+ // If user starts a transaction, handler_continue indicates the
+ // first handler pointer that triggers the transaction.
+ RedisCommandHandler* handler_continue;
+ // The redis command are parsed and pushed into this queue
+ bthread::ExecutionQueueId<ConsumeTaskDone*> queue;
+
+ RedisReply parsing_message;
+ butil::Arena arena;
+ int64_t message_count;
+private:
+ void AddSentDone(ConsumeTaskDone* done);
- do {
- InputResponse* msg =
static_cast<InputResponse*>(socket->parsing_context());
- if (msg == NULL) {
- msg = new InputResponse;
- socket->reset_parsing_context(msg);
+ bool _writing = false;
+ butil::Mutex _mutex;
+ std::queue<ConsumeTaskDone*> _dones;
+
+ butil::Mutex _dones_sent_mutex;
+ std::queue<ConsumeTaskDone*> _dones_sent;
+};
+
+class ConsumeTaskDone : public google::protobuf::Closure {
+public:
+ ConsumeTaskDone()
+ : _ready(false)
+ , output_message(&arena)
+ , ctx(NULL) {}
+
+ void Run() override;
+ bool IsReady() { return _ready.load(butil::memory_order_acquire); }
+
+private:
+ butil::atomic<bool> _ready;
+
+public:
+ RedisReply input_message;
+ RedisReply output_message;
+ RedisConnContext* ctx;
+ butil::IOBuf sendbuf;
+ butil::Arena arena;
+};
+
+int ConsumeTask(RedisConnContext* ctx, ConsumeTaskDone* done) {
+ ClosureGuard done_guard(done);
+ done->ctx = ctx;
+ ctx->Push(done);
Review comment:
这里的ownership出问题了，done必然是uniquely owned，这里同时被done_guard和ctx
own，比如下面return -1时ctx中的done就野指针了。
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@brpc.apache.org
For additional commands, e-mail: dev-h...@brpc.apache.org
