4 releases (2 breaking)

0.4.1 Jul 12, 2024
0.4.0 Jan 5, 2024
0.3.0 Jan 5, 2024
0.1.0 Jan 4, 2024

#168 in Memory management

Download history 9/week @ 2024-06-17 65/week @ 2024-06-24 23/week @ 2024-07-01 156/week @ 2024-07-08 130/week @ 2024-07-15 67/week @ 2024-07-22 83/week @ 2024-07-29 60/week @ 2024-08-05 17/week @ 2024-08-12 54/week @ 2024-08-19 86/week @ 2024-08-26 30/week @ 2024-09-02 14/week @ 2024-09-09 39/week @ 2024-09-16 58/week @ 2024-09-23 21/week @ 2024-09-30

132 downloads per month

Apache-2.0

22KB
322 lines

resourcetrack

Yet-another resource tracking tool.

About

A trim, low-dependency tool for counting things. This project does not use unsafe code. It only depends on std.

Safety, clarity, flexibility, and performance are favored in descending order. Counters are plain structs holding an Arc to an atomic usize. This means there is a memory cost to the tracking. 0-overhead is not a goal here.

Details

Resourcetrack supports static categories with your own names:

use resourcetrack::new_registry;

#[derive(Clone, Debug, PartialEq, Eq, Hash)]
enum MyCategories {
    Miscellaneous,
    Specific,
}

let registry = new_registry::<MyCategories>();
let specific_category_tracker = registry.category(MyCategories::Specific);

Resourcetrack does not explicitly synchronize categorized resource counters.

let _count_sentinel: Count = category_tracker.track(); // non-blocking, on both track and drop

Counters are a plain struct, and you can compose them onto your expensive business objects for automatic count management. If you do this, you can consider using lazy_static for this Registry to wrap up the counts inside of your constructor function. Ideally you'd cascade the lazy_static into category Trackers too!

struct ExpensiveResource {
    payload: String,
    _phantom_count: resourcetrack::tracked::Count,
}

When you need to get the counts, for logging or metrics or whatever, just read them.

let registry = new_registry::<MyCategories>();
{
    let _counter_1 = registry.category(MyCategories::Specific).track();
    assert_eq!(vec![(MyCategories::Specific, 1)], registry.read_counts::<Vec<_>>(), "1 specific instance");
    let _counter_2 = registry.category(MyCategories::Specific).track();
    assert_eq!(vec![(MyCategories::Specific, 2)], registry.read_counts::<Vec<_>>(), "2 specific instances");
}

assert_eq!(vec![(MyCategories::Specific, 0)], registry.read_counts::<Vec<_>>(), "both dropped");

You can track sized resources, where their size changes. To stay sane, you should probably limit yourself to either using track() or track_sized() for a given category. You can mix counts and sizes within a registry though, no problem! Complete example:

use resourcetrack::tracked;

// Set up your statically knowable categories
#[derive(Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash)]
enum MyCategories {
    ResourceCount,
    ResourceWeight,
}

// Here is an example of a tracked business object. Both size and weight are tracked.
struct TrackedVector {
    internal: Vec<String>,
    _count_sentinel: tracked::Count,
    weight: tracked::Size,
}
impl TrackedVector {
    pub fn push(&mut self, next: String) {
        self.weight.add(next.len());
    }
}

// Static setup - this should be in some shared lazy static scope.
let registry = resourcetrack::new_registry::<MyCategories>();
let resource_counts = registry.category(MyCategories::ResourceCount);
let resource_weights = registry.category(MyCategories::ResourceWeight);

let mut v = TrackedVector { // This should be wrapped into TrackedVector::new() in your application
    internal: Default::default(),
    _count_sentinel: resource_counts.track(),
    weight: resource_weights.track_size(0),
};
v.push("hello".to_string());
//!
let mut counts = registry.read_counts::<Vec<_>>();
counts.sort();
assert_eq!(
    vec![
        (MyCategories::ResourceCount, 1),
        (MyCategories::ResourceWeight, 5),
    ],
    counts,
)

No runtime deps