Exponential data-transfer can be changed to linear

[rpriyadarshi]

Since we are accumulating data transferred in result string, the find becomes exponential as we always search the concatinated string from beginning.

Client file linuxtcpsocketclient.cpp:

void LinuxTcpSocketClient::SendRPCMessage(const std::string& message, std::string& result) throw (JsonRpcException) {
    ...
		{
			string tmp;
			tmp.append(buffer, nbytes);
			result.append(buffer,nbytes);
		}
	} while(result.find(DELIMITER_CHAR) == string::npos);
    ...
}

Server file linuxtcpsocketserver.cpp:

void* LinuxTcpSocketServer::GenerateResponse(void *p_data) {
...
		{
			request.append(buffer,nbytes);
		}
	} while(request.find(DELIMITER_CHAR) == string::npos);
...
}

Change the while loop to:
} while(memchr(buffer, DELIMITER_CHAR, nbytes) == NULL);
Delete the tmp string as well

			string tmp;
			tmp.append(buffer, nbytes);

With these changes, I did a round trip transfer of 10^7 characters in 0.8 seconds instead of 108 seconds.

[loreilei]

In which version did you spot this ?

This is indeed completely useless to have this

string tmp; tmp.append(buffer, nbytes);
It add useless copy.

I did not spot it in the master branch libjson-rpc-cpp/src/jsonrpccpp/client/connectors/linuxtcpsocketclient.cpp

but it is still present in the windows implementation ( libjson-rpc-cpp/src/jsonrpccpp/client/connectors/windowstcpsocketclient.cpp).

I see in the linux tcp socket connector and the unix domain socket connector that the stream reader is used, I guess it would be nice to also use it in the windows tcpsocket connector (unless it uses linux features, I haven't checked that).

PS: My bad it seems this useless copy was introduced by me since it is in tag v0.7.0 version of file...

[rpriyadarshi]

There has been quite a bit of refactoring since I got the code last year. Looks like the exponential data issue is still there. The target will be a cumulative string that would be searched from the beginning each time: (common/streamreader.cpp)

bool StreamReader::Read(std::string &target, int fd, char delimiter) {
  ssize_t bytesRead;
  do {
    bytesRead = read(fd, this->buffer, buffersize);
    if (bytesRead < 0) {
      return false;
    } else {
      target.append(buffer, static_cast<size_t>(bytesRead));
    }
  } while (target.find(delimiter) == string::npos && bytesRead > 0);

  target.pop_back();
  return true;
}

[loreilei]

So a fix could be add a counter of how many bytes we read and use it as a starting position for the find call.