Struct AtomicSnowflakeGenerator 

pub struct AtomicSnowflakeGenerator<ID, T>
where
    ID: SnowflakeId<Ty = u64>,
    T: TimeSource<ID::Ty>,
{ /* private fields */ }

Available on crate feature snowflake and crate feature atomic and target_has_atomic=64 only.

A lock-free Snowflake ID generator suitable for multi-threaded environments.

This generator stores the Snowflake state in an AtomicU64, allowing safe shared use across threads.

Features

• ✅ Thread-safe
• ❌ Safely implement any SnowflakeId layout

Caveats

This implementation uses an AtomicU64 internally, so it only supports ID layouts where the underlying type is u64. You cannot use layouts with larger or smaller representations (i.e., ID::Ty must be u64).

Recommended When

• You’re in a multi-threaded environment
• Fair access is sacrificed for higher throughput

See Also

• BasicSnowflakeGenerator
• LockSnowflakeGenerator

Implementations

impl<ID, T> AtomicSnowflakeGenerator<ID, T>

where ID: SnowflakeId<Ty = u64>, T: TimeSource<ID::Ty>,

pub fn new(machine_id: ID::Ty, time: T) -> Self

Available on crate features snowflake or ulid only.

Creates a new AtomicSnowflakeGenerator initialized with the current time and a given machine ID.

This constructor sets the initial timestamp and sequence to zero, and uses the provided time to fetch the current time during ID generation. It is the recommended way to create a new atomic generator for typical use cases.

Parameters

• machine_id: A unique identifier for the node or instance generating IDs. This value will be encoded into every generated ID.
• time: A TimeSource implementation (e.g., MonotonicClock) that determines how timestamps are generated.

Returns

A new AtomicSnowflakeGenerator ready to produce unique, time-ordered IDs.

Example

use ferroid::{AtomicSnowflakeGenerator, IdGenStatus, SnowflakeTwitterId, TWITTER_EPOCH, MonotonicClock};
     
let generator = AtomicSnowflakeGenerator::new(0, MonotonicClock::with_epoch(TWITTER_EPOCH));

let id: SnowflakeTwitterId = loop {
    match generator.next_id() {
        IdGenStatus::Ready { id } => break id,
        IdGenStatus::Pending { .. } => core::hint::spin_loop(),
    }
};

pub fn from_components( timestamp: ID::Ty, machine_id: ID::Ty, sequence: ID::Ty, time: T, ) -> Self

Available on crate features snowflake or ulid only.

Creates a new ID generator from explicit component values.

This constructor is primarily useful for advanced use cases such as restoring state from persistent storage or controlling the starting point of the generator manually.

Parameters

• timestamp: The initial timestamp component (usually in milliseconds)
• machine_id: The machine or worker identifier
• sequence: The initial sequence number
• time: A TimeSource implementation used to fetch the current time

Returns

A new generator instance preloaded with the given state.

⚠️ Note

In typical use cases, you should prefer Self::new to let the generator initialize itself from the current time.

pub fn next_id(&self) -> IdGenStatus<ID>

Available on crate features snowflake or ulid only.

Attempts to generate the next available ID.

Returns a new, time-ordered, unique ID if generation succeeds. If the generator is temporarily exhausted (e.g., the sequence is full and the time has not advanced, or CAS fails), it returns IdGenStatus::Pending.

Panics

This method currently has no fallible code paths, but may panic if an internal error occurs in future implementations. For explicitly fallible behavior, use Self::try_next_id instead.

Example

use ferroid::{AtomicSnowflakeGenerator, IdGenStatus, SnowflakeTwitterId, TWITTER_EPOCH, MonotonicClock};

let generator = AtomicSnowflakeGenerator::new(0, MonotonicClock::with_epoch(TWITTER_EPOCH));

let id: SnowflakeTwitterId = loop {
    match generator.next_id() {
        IdGenStatus::Ready { id } => break id,
        IdGenStatus::Pending { .. } => std::thread::yield_now(),
    }
};

pub fn try_next_id(&self) -> Result<IdGenStatus<ID>>

Available on crate features snowflake or ulid only.

A fallible version of Self::next_id that returns a Result.

This method attempts to generate the next ID based on the current time and internal state. If successful, it returns IdGenStatus::Ready with a newly generated ID. If the generator is temporarily exhausted or CAS fails, it returns IdGenStatus::Pending. If an internal failure occurs (e.g., a time source or lock error), it returns an error.

Returns

• Ok(IdGenStatus::Ready { id }): A new ID is available
• Ok(IdGenStatus::Pending { yield_for }): The time to wait (in milliseconds) before trying again
• Err(_): infallible for this generator

Errors

• This method currently does not return any errors and always returns Ok. It is marked as fallible to allow for future extensibility

Example

use ferroid::{AtomicSnowflakeGenerator, ToU64, IdGenStatus, SnowflakeTwitterId, TWITTER_EPOCH, MonotonicClock};

let generator = AtomicSnowflakeGenerator::new(0, MonotonicClock::with_epoch(TWITTER_EPOCH));

// Attempt to generate a new ID
let id: SnowflakeTwitterId = loop {
    match generator.try_next_id() {
        Ok(IdGenStatus::Ready { id }) => break id,
        Ok(IdGenStatus::Pending { yield_for }) => {
            std::thread::sleep(core::time::Duration::from_millis(yield_for.to_u64()));
        }
        Err(_) => unreachable!(),
    }
};

Trait Implementations

impl<ID, T> SnowflakeGenerator<ID, T> for AtomicSnowflakeGenerator<ID, T>

where ID: SnowflakeId<Ty = u64>, T: TimeSource<ID::Ty>,

Available on crate features snowflake or ulid only.

type Err = Infallible

fn new(machine_id: ID::Ty, time: T) -> Self

fn next_id(&self) -> IdGenStatus<ID>

Returns the next available ID

fn try_next_id(&self) -> Result<IdGenStatus<ID>, Self::Err>

A fallible version of Self::next_id that returns a Result.