#compression #data-stream #financial-data #data-encoding #data-exchange #market-data #data-processing

fastlib

FAST (FIX Adapted for STreaming protocol) is a space and processing efficient encoding method for message oriented data streams

5 unstable releases

0.3.1 Jul 29, 2024
0.3.0 Jul 23, 2024
0.2.1 Jul 11, 2024
0.2.0 Jul 6, 2024
0.1.0 Jul 3, 2024

#705 in Parser implementations

Download history 236/week @ 2024-07-24 52/week @ 2024-07-31 6/week @ 2024-09-11 4/week @ 2024-09-18 7/week @ 2024-09-25 2/week @ 2024-10-02

277 downloads per month

MIT license

325KB
8K SLoC

FIX/FAST Protocol Decoder/Encoder

Crates.io Build Status License

FAST (FIX Adapted for STreaming protocol) is a space and processing efficient encoding method for message oriented data streams.

The FAST protocol has been developed as part of the FIX Market Data Optimization Working Group. FAST data compression algorithm is designed to optimize electronic exchange of financial data, particularly for high volume, low latency data dissemination. It significantly reduces bandwidth requirements and latency between sender and receiver. FAST works especially well at improving performance during periods of peak message rates.

Technical Specification: https://www.fixtrading.org/standards/fast-online/
Supported version: 1.x.1

Usage

Add to your Cargo.toml:

[dependencies]
fastlib = "0.3"

Serialize/Deserialize using serde

For templates defined in XML, e.g.:

<?xml version="1.0" encoding="UTF-8" ?>
<templates xmlns="http://www.fixprotocol.org/ns/fast/td/1.1">
    <template name="MsgHeader">
        <uInt32 id="34" name="MsgSeqNum"/>
        <uInt64 id="52" name="SendingTime"/>
    </template>
    <template id="1" name="MDHeartbeat">
        <templateRef name="MsgHeader"/>
    </template>
    <template id="2" name="MDLogout">
        <templateRef name="MsgHeader"/>
        <string id="58" name="Text" presence="optional"/>
    </template>
</templates>

Define the message types in Rust:

use serde::{Serialize, Deserialize};

#[derive(Serialize, Deserialize)]
enum Message {
    MDHeartbeat(Heartbeat),
    MDLogout(Logout),
}

#[derive(Serialize, Deserialize)]
struct MsgHeader {
    #[serde(rename = "MsgSeqNum")]
    msg_seq_num: u32,
    #[serde(rename = "SendingTime")]
    sending_time: u64,
}

#[derive(Serialize, Deserialize)]
#[serde(rename_all = "PascalCase")]
struct Heartbeat {
    #[serde(flatten)]
    msg_header: MsgHeader,
}

#[derive(Serialize, Deserialize)]
#[serde(rename_all = "PascalCase")]
struct Logout {
    #[serde(flatten)]
    msg_header: MsgHeader,
    text: Option<String>,
}

Some implementation guidelines:

  • <templates> must be implemented as enum;
  • <decimal> can be deserialized to f64 or fastlib::Decimal (if you need to preserve original scale);
  • <byteVector> is a Vec<u8> and must be prefixed with #[serde(with = "serde_bytes")];
  • <sequence> is a Vec<SequenceItem>, where SequenceItem is a struct;
  • <group> is a nested struct;
  • fields with optional presence are Option<...>;
  • static template reference can be plain fields from the template or flattened struct,
  • dynamic template references must be Box<Message> with #[serde(rename = "templateRef:N")], where N is a 0-based index of the <teplateRef> in its group.

To deserialize a message call fastlib::from_vec, fastlib::from_bytes or from_stream:

use fastlib::Decoder;

// Create a decoder from XML templates.
let mut decoder = Decoder::new_from_xml(include_str!("templates.xml"))?;

// Raw data that contains one message.
let raw_data: Vec<u8> = vec![ ... ];

// Deserialize a message.
let msg: Message = fastlib::from_vec(&mut decoder, raw_data)?;

To serialize a message call fastlib::to_vec, fastlib::to_bytes or to_stream:

use fastlib::Encoder;

// Create a encoder from XML templates.
let mut encoder = Encoder::new_from_xml(include_str!("templates.xml"))?;

// Message to serialize.
let msg = Message::MDHeartbeat{
    Heartbeat {
        ...
    }
};

// Serialize a message.
let raw: Vec<u8> = fastlib::to_vec(&mut encoder, &msg)?;

Decode to JSON

use fastlib::Decoder;
use fastlib::JsonMessageFactory;

// Create a decoder from XML templates.
let mut decoder = Decoder::new_from_xml(include_str!("templates.xml"))?;

// Raw data that contains one message.
let raw_data: Vec<u8> = vec![ ... ];

// Create a JSON message factory.
let mut msg = JsonMessageFactory::new();

// Decode the message.
decoder.decode_vec(raw_data, &mut msg)?;

println!("{}", msg.json);

Decode using own message factory

Make a new struct that implements fastlib::MessageFactory trait:

use fastlib::{MessageFactory, Value};

pub struct MyMessageFactory {
}

impl MessageFactory for MyMessageFactory {
    // ... your implementation here ...
}

You have to implement callbacks that will be called during message decoding:

pub trait MessageFactory {
    // Process template id
    fn start_template(&mut self, id: u32, name: &str);
    fn stop_template(&mut self);
    
    // Process field value
    fn set_value(&mut self, id: u32, name: &str, value: Option<Value>);
    
    // Process sequence
    fn start_sequence(&mut self, id: u32, name: &str, length: u32);
    fn start_sequence_item(&mut self, index: u32);
    fn stop_sequence_item(&mut self);
    fn stop_sequence(&mut self);
    
    // Process group
    fn start_group(&mut self, name: &str);
    fn stop_group(&mut self);
    
    // Process template ref
    fn start_template_ref(&mut self, name: &str, dynamic: bool);
    fn stop_template_ref(&mut self);
}

For examples see implementation for fastlib::text::TextMessageFactory or fastlib::text::JsonMessageFactory but more likely you will want to construct you own message structs.

Then create a decoder from templates XML file and decode a message:

use fastlib::Decoder;

// Create a decoder from XML templates.
let mut decoder = Decoder::new_from_xml(include_str!("templates.xml"))?;

// Raw data that contains one message.
let raw_data: Vec<u8> = vec![ ... ];

// Create a message factory.
let mut msg = MyMessageFactory{};

// Decode the message.
decoder.decode_vec(raw_data, &mut msg)?;

Examples

License

This project is licensed under the MIT license.

Dependencies

~2.3–3MB
~52K SLoC