In the past few weeks I downloaded pcap data from NYSE and learned how to parse the pcap files using C++.
I have been able to strip out all the network headers (UDP, IP headers) and get to the payload. I have started writing code to process the payload.
The method I have used so far is to define various structs for each message type. For e.g.
struct msg1 {
uint16_t msgSize; // 2 bytes
uint16_t msgType; // 2 bytes
uint16_t time; // 4 bytes
uint16_t timeNS; // 4 bytes
uint8_t productID; //1 byte
uint8_t channelID; // 1 byte
};
struct msg34 {
uint16_t msgSize; // 2 byte
uint16_t msgType; // 2 byte
uint32_t time; // 4 byte
uint32_t timeNS; // 4 byte
uint32_t symbolIndex; // 4 byte
uint32_t symbolSeqNum; // 4 byte
u_char secStatus; // 1 byte
u_char haltCondition; // 1 byte
uint32_t reserved; // 4 bytes
uint32_t price_1; // 4 bytes
uint32_t price_2; // 4 bytes
u_char ssr; // 1 byte
uint32_t ssrVol; // 4 bytes
uint32_t time; // 4 bytes
u_char ssrState; // 1 byte
u_char mktState; //1 byte
u_char sessionState; // 1 byte
};
Next, once I determine the appropriate message type from the first 2 bytes of the payload, I use reinterpret_cast to initialize the struct and store the payload data as follows:
const uint16_t* msgType = reinterpret_cast<const uint16_t*>(packet+2);
switch(*msgType) {
case 34: {
const msg34* msg34_ = reinterpret_cast<const msg34*>(packet);
std::cout << *msg34_;
break;
}
}
packet here is of type const u_char* that points to the payload which is in binary format.
Some people did not approve the use of reinterpret_cast to initialize the struct citing that it unsafe. I would like to get some feedback on an alternative method to directly initialize a struct with binary data accessed by the const u_char* pointer named packet.
Besides, I would also like to get some advice on how I could better structure the different message types for better code readability and runtime performance. Right now I have to define a struct for each message type and as you can imagine, my struct definition file is pretty long. Would it be better to define the messages using classes (abstract and real)?
Thank you in advance for any and all help.
P.S.: If anyone needs any help getting started on a similar project, feel free to reach out and I will share whatever code I have.
