Crate snowcloud

source ·
Expand description

Snowcloud

a small library for implementing custom ids based on timestamps, static ids, and sequence counters. the module provides 2 types of generators, a thread safe and non thread safe version. they allow for different types of waiting for ids if you want specific behavior. the snowflakes generated by a generator are composed of 3 sections and are described below. small example of how to create a generator and create a snowflake is shown below.

// 43 bit timestamp, 8 bit primary id, 12 bit sequence
type MyCloud = snowcloud::SingleThread<43, 8, 12>;

// 2023/03/23 9:00:00 in milliseconds, timestamps will start from this
// date
const START_TIME: u64 = 1679587200000;
// primary id could be a machine/node id for example
const PRIMARY_ID: i64 = 1;

let mut cloud = MyCloud::new(PRIMARY_ID, START_TIME)
    .expect("failed to create MyCloud");
let flake = cloud.next_id()
    .expect("failed to create snowflake");

println!("{}", flake.id());

Behavior

MultiThread is a thread safe implementation for sharing between threads on a system. it uses an Arc Mutex to handle sharing the sequence count and prev_time. the only time it will block is when acquiring the mutex and will not wait if a valid snowflake cannot be acquired. if a generator is unable to create a snowflake because the max sequence number has been reached an error will be returned providing an estimated duration to the next millisecond. how you want to wait can be decided by the user.

SingleThread is similar in most aspects to MultiThread expect next_id is a mutating call and sequence count with prev_time are not stored in an Arc Mutext. THIS IS NOT THREAD SAFE.

Traits

to help with using a generator in other situations, traits are provided and implemented for the base types (more can be added later if necessary/desired).

  • IdGenerator describes the basic layout of an id generator. requiring an Error, Id, and Output type to be specified along with the next_id method.
  • IdGeneratorMut is similar to IdGenerator except the next_id call allows for mutating the object
  • NextAvailId describes an object that is capable of returing a duraiton to the next available millisecond.

current use case would be for allowing different types of waiting for the next available id. see blocking_next_id for example implementation.

Snowflake Bits

the format of the snowflake is as follows:

 01111111111111111111111111111111111111111111 - 11111111 - 111111111111
 |                                          |   |      |   |          |
64                                         21  20     13  12          1
                                    timestamp          |              |
                                              primary id              |
                                                               sequence

the bit values for each segment can be specified by the user if you need/want certain ranges. there is currently no check to ensure that the values specified equal 63 bits (at least none that I can thing of).

Timestamp

the timestamp is in millisconds and is based from a specific start date that you can specify. the start date must be in the future of UNIX_EPOCH and cannot be a date in the future, now >= start_time >= UNIX_EPOCH. internally, a snowcloud will use SystemTime to get the timestamp and the convert to the necessary values.

// the current example date is 2023/03/23 9:00:00.
const VALID_START_DATE: u64 =   1679587200000;

// if a date that is after the current system time is provided the
// snowcloud will return an error. 2077/10/23 9:00:00
const INVALID_START_DATE: u64 = 3402205200000;

below is a table with various bit values and how many years you can get out of a timestamp. you will probably get diminishing returns with lower bit values if this is to be used over a long duration of time.

bitsmax valueyears
438796093022207278
424398046511103139
41219902325555169
40109951162777534
3954975581388717
382748779069438
371374389534714
36687194767352
35343597383671

Primary Id

the primary id is static to the instance of a generator. if using a generator across multiple machines then it could be the id of the machine. this would allow for those machines to create unique snowflakes that will not conflict with each other as long as the primary id is different between them.

Sequence

the sequence number is used to indicate how many records can be generated within a single millisecond. when the max sequence value is reached in a generator, the user will have to wait for the next available millisecond to get a new snowflake. the higher the bit value the more records that can be created each millisecond. the sequence number will always start with 1 when creating a generator.

De/Serialize

snowflakes support serde Serialize and Deserialize to i64 with an addtional option to de/serailize to a string with i64_string_id

Modules

  • base traits for implementing helper code
  • methods for waiting on the next available id from a snowcloud

Structs

Enums

  • possible errors for Snowclouds/Snowflakes