alexrichterxyz / sirca_asx_itch_parser

A high-performance parser for ASX ITCH messages provided by SIRCA.

High-performance parser for ASX ITCH feed messages provided by SIRCA

This is a highly optimized library for parsing ASX ITCH messages provided by SIRCA. The ASX ITCH feed provides traders with a view of the entire market including full order book activity and trades of thousands of assets. The library is written in C++ (17 or higher).

Some features include:

• full parsing support for all 14 message types outlined in the message format specification
• efficient parsing of timestamps with nanosecond precision
• conversion of hex-string order numbers to 64-bit and 96-bit numeric representations

Efficiency

The code parses each message (line) sequentially. Bellow is an example of 5 messages. Each message consists comma-separated components.

2024-01-04 09:57:24.293086295,D,6c5ae041:00064240,70759,S,IGO,230016
2024-01-04 09:57:24.293992811,Z,70860,4662,7597,52100,56000,49500,5,500,NUF,230017
2024-01-04 09:57:24.293992811,D,6c5ae041:00060344,70860,S,NUF,230018
2024-01-04 09:57:24.295321875,D,6c5ae041:00064300,70915,B,RIO,230019
2024-01-04 09:57:24.295506674,A,6c5ae041:0006430b,70915,B,56,25,1342400,0,2,RIO,230020

I make use of numerous optimizations including:

• pre-parsing functions: shared message components (e.g. the message type, timestamp, and symbol) can be parsed directly from the message string without having to instantiate a message object.
• reduced memory allocation: support for std::string_view as opposed to std::string as the data type for messages avoids memory allocation when splitting the comma-separated message components. Preliminary tests have shown this can more than double performance.
• memory preallocation: the maximum number of message components is known and can be used to pre-allocate sufficient memory for storing the components of any message type.
• lazy processing: message components are only fully parsed once they are needed.

Example

This example shows how to parse messages from sample.csv and output the symbol, nanosecond timestamp, price, and quantity of all insertion orders (without participants).

#include <fstream>
#include <iostream>
#include "include/asxitch.hpp"

int main() {
    std::ifstream file("sample.csv");

    if (!file.is_open()) {
        std::cerr << "Could not open the file." << std::endl;
        return 1;
    }

    // string to hold message contents
    std::string message_str;
    
    // loop through each line
    while(std::getline(file, message_str)) { 

        // extract message type before instantiating message object
        const asxitch::message_type_t type = asxitch::get_message_type(message_str);

        // ignore all messages that are not "add orders"
        if(type != asxitch::message_type_t::ADD_ORDER_NO_PARTICIPANT) {
            continue;
        }

        // convert message string to string_view to improve performance 
        const std::string_view message_view(message_str); 

        // create message from string view
        const asxitch::message_add_order_no_participant message(message_view);
        
        // output parsed message payload
        std::cout
        << "Add order for symbol " << message.symbol()
        << " at time " << message.timestamp().count()
        << " at price " << message.price()
        << " with quantity " << message.quantity() << '\n';
        
    }

    file.close();

    return 0;
}

Todos

• add doxygen-style documentation comments
• add tests for all message types
• support for decimal output of the custom asxitch::int96_t type