2 stable releases

Uses old Rust 2015

1.1.0 Feb 26, 2016
1.0.0 Feb 25, 2016

#43 in #dag

MIT license

6KB
84 lines

Ackr

A Rust implementation of the Storm acking algorithm allowing to track the state of arbitrary tuples in a DAG with a static ~20 bytes of memory needed.

Getting Started

Install it with Cargo:

[dependencies]
ackr = "1.0.0"

And include the crate in your src/lib.rs

extern crate ackr;

use ackr::{Tuple, Source, Task, Ackr};

Creating a new Source

Each Ackr can track a number of DAG's of tuples and their state.

let mut ackr = Ackr::new();
ackr.insert(Source(1), Task(1));

Tasks are arbitrary and are simply a wrapper around u32s, but they have the ability to track the origin of the acks.

Source's are simply the source tuple id, as u64s. If you have the following DAG:

A(1)
|
B(2)

Each tuple should be associated a random id. A or the Source would be the first tuple we insert and thus need to ack later on for it to be considered "completed".

assert!(ackr.has_completed(Source(1)));

Because we use the initial Source as the first tuple, has_completed will return false because we haven't acked it.

ackr.ack(Source(1), Tuple(1));
assert!(ackr.has_completed(Source(1)));

This will now return true because we have acked the Source/Tuple of 1.

Acking Tuples

Inserting and acking are basically the same thing. Acking a tuple once will act as an insert and acking it twice will remove it, effectively "completing" the tuple.

However, to make it clearer, there are two distinct APIs.

ackr.insert(Source(1), Tuple(2));

Going back to the previous DAG example of A(1) -> B(2), we're inserting the next tuple of the DAG.

ackr.ack(Source(1), Tuple(2));

Now we have acked/completed the tuple.

Once all the tuples (including the source dag) have been acked, has_completed will return true.

License

MIT

No runtime deps