#cache #memoization #lru

yanked near-cached

Near fork of https://github.com/jaemk/cached

Uses old Rust 2015

0.8.0 Aug 21, 2019

#51 in #memoization

MIT license

35KB
648 lines

Near fork of https://github.com/jaemk/cached crate with some optimizations.


lib.rs:

Build Status crates.io docs

Caching structures and simplified function memoization

cached provides implementations of several caching structures as well as a handy macro for defining memoized functions.

Memoized functions defined using cached! macros are thread-safe with the backing function-cache wrapped in mutex. The function-cache is not locked for the duration of the function's execution, so initial (on an empty cache) concurrent calls of long-running functions with the same arguments will each execute fully and each overwrite the memoized value as they complete. This mirrors the behavior of Python's functools.lru_cache.

See cached::stores docs for details about the cache stores available.

Defining memoized functions using cached!

cached! defined functions will have their results cached using the function's arguments as a key (or a specific expression when using cached_key!). When a cached! defined function is called, the function's cache is first checked for an already computed (and still valid) value before evaluating the function body.

Due to the requirements of storing arguments and return values in a global cache:

  • Function return types must be owned and implement Clone
  • Function arguments must either be owned and implement Hash + Eq + Clone OR the cached_key! macro must be used to convert arguments into an owned + Hash + Eq + Clone type.
  • Arguments and return values will be cloned in the process of insertion and retrieval.
  • cached! functions should not be used to produce side-effectual results!
  • cached! functions cannot live directly under impl blocks since cached! expands to a once_cell initialization and a funtion definition.

NOTE: Any custom cache that implements cached::Cached can be used with the cached macros in place of the built-ins.

See examples for basic usage and an example of implementing a custom cache-store.

cached! and cached_key! Usage & Options:

There are several options depending on how explicit you want to be. See below for a full syntax breakdown.

1.) Using the shorthand will use an unbounded cache.

#[macro_use] extern crate cached;

/// Defines a function named `fib` that uses a cache named `FIB`
cached!{
FIB;
fn fib(n: u64) -> u64 = {
if n == 0 || n == 1 { return n }
fib(n-1) + fib(n-2)
}
}

2.) Using the full syntax requires specifying the full cache type and providing an instance of the cache to use. Note that the cache's key-type is a tuple of the function argument types. If you would like fine grained control over the key, you can use the cached_key! macro. The following example uses a SizedCache (LRU):

#[macro_use] extern crate cached;

use std::thread::sleep;
use std::time::Duration;
use cached::SizedCache;

/// Defines a function `compute` that uses an LRU cache named `COMPUTE` which has a
/// size limit of 50 items. The `cached!` macro will implicitly combine
/// the function arguments into a tuple to be used as the cache key.
cached!{
COMPUTE: SizedCache<(u64, u64), u64> = SizedCache::with_size(50);
fn compute(a: u64, b: u64) -> u64 = {
sleep(Duration::new(2, 0));
return a * b;
}
}

3.) The cached_key macro functions identically, but allows you to define the cache key as an expression.

#[macro_use] extern crate cached;

use std::thread::sleep;
use std::time::Duration;
use cached::SizedCache;

/// Defines a function named `length` that uses an LRU cache named `LENGTH`.
/// The `Key = ` expression is used to explicitly define the value that
/// should be used as the cache key. Here the borrowed arguments are converted
/// to an owned string that can be stored in the global function cache.
cached_key!{
LENGTH: SizedCache<String, usize> = SizedCache::with_size(50);
Key = { format!("{}{}", a, b) };
fn length(a: &str, b: &str) -> usize = {
let size = a.len() + b.len();
sleep(Duration::new(size as u64, 0));
size
}
}

4.) The cached_result and cached_key_result macros function similarly to cached and cached_key respectively but the cached function needs to return Result (or some type alias like io::Result). If the function returns Ok(val) then val is cached, but errors are not. Note that only the success type needs to implement Clone, not the error type. When using cached_result and cached_key_result, the cache type cannot be derived and must always be explicitly specified.

#[macro_use] extern crate cached;

use cached::UnboundCache;

/// Cache the successes of a function.
/// To use `cached_key_result` add a key function as in `cached_key`.
cached_result!{
MULT: UnboundCache<(u64, u64), u64> = UnboundCache::new(); // Type must always be specified
fn mult(a: u64, b: u64) -> Result<u64, ()> = {
if a == 0 || b == 0 {
return Err(());
} else {
return Ok(a * b);
}
}
}

Syntax

The common macro syntax is:

cached_key!{
CACHE_NAME: CacheType = CacheInstance;
Key = KeyExpression;
fn func_name(arg1: arg_type, arg2: arg_type) -> return_type = {
// do stuff like normal
return_type
}
}

Where:

  • CACHE_NAME is the unique name used to hold a static ref to the cache
  • CacheType is the full type of the cache
  • CacheInstance is any expression that yields an instance of CacheType to be used as the cache-store, followed by ;
  • When using the cached_key! macro, the "Key" line must be specified. This line must start with the literal tokens Key = , followed by an expression that evaluates to the key, followed by ;
  • fn func_name(arg1: arg_type) -> return_type is the same form as a regular function signature, with the exception that functions with no return value must be explicitly stated (e.g. fn func_name(arg: arg_type) -> ())
  • The expression following = is the function body assigned to func_name. Note, the function body can make recursive calls to its cached-self (func_name).

Fine grained control using cached_control!

The cached_control! macro allows you to provide expressions that get plugged into key areas of the memoized function. While the cached and cached_result variants are adequate for most scenarios, it can be useful to have the ability to customize the macro's functionality.

#[macro_use] extern crate cached;

use cached::UnboundCache;

/// The following usage plugs in expressions to make the macro behave like
/// the `cached_result!` macro.
cached_control!{
CACHE: UnboundCache<String, String> = UnboundCache::new();

// Use an owned copy of the argument `input` as the cache key
Key = { input.to_owned() };

// If a cached value exists, it will bind to `cached_val` and
// a `Result` will be returned containing a copy of the cached
// evaluated body. This will return before the function body
// is executed.
PostGet(cached_val) = { return Ok(cached_val.clone()) };

// The result of executing the function body will be bound to
// `body_result`. In this case, the function body returns a `Result`.
// We match on the `Result`, returning an early `Err` if the function errored.
// Otherwise, we pass on the function's result to be cached.
PostExec(body_result) = {
match body_result {
Ok(v) => v,
Err(e) => return Err(e),
}
};

// When inserting the value into the cache we bind
// the to-be-set-value to `set_value` and give back a copy
// of it to be inserted into the cache
Set(set_value) = { set_value.clone() };

// Before returning, print the value that will be returned
Return(return_value) = {
println!("{}", return_value);
Ok(return_value)
};

fn can_fail(input: &str) -> Result<String, String> = {
let len = input.len();
if len < 3 { Ok(format!("{}-{}", input, len)) }
else { Err("too big".to_string()) }
}
}

Dependencies

~1.5MB
~24K SLoC