Current section

Files

Jump to
glimit src glimit.gleam
Raw

src/glimit.gleam

//// This module provides a distributed rate limiter that can be used to limit the
//// number of requests or function calls per second for a given identifier.
////
//// A single actor is used to assign one rate limiter actor per identifier. The
//// rate limiter actor then uses a Token Bucket algorithm to determine if a
//// request or function call should be allowed to proceed. A separate process is
//// polling the rate limiters to remove full buckets to reduce unnecessary memory
//// usage.
////
//// The rate limits are configured using the following two options:
////
//// - `per_second`: The rate of new available tokens per second. Think of this
//// as the steady state rate limit.
//// - `burst_limit`: The maximum number of available tokens. Think of this as
//// the burst rate limit. The default value is the `per_second` rate limit.
////
//// The rate limiter can be applied to a function or handler using the `apply`
//// function, which returns a new function that checks the rate limit before
//// calling the original function.
////
//// # Example
////
//// ```gleam
//// import glimit
////
//// let limiter =
//// glimit.new()
//// |> glimit.per_second(10)
//// |> glimit.burst_limit(100)
//// |> glimit.identifier(fn(request) { request.ip })
//// |> glimit.on_limit_exceeded(fn(_request) { "Rate limit reached" })
////
//// let handler =
//// fn(_request) { "Hello, world!" }
//// |> glimit.apply(limiter)
//// ```
////
//// # Multi-argument functions
////
//// `apply` wraps a single-argument function `fn(a) -> b`. To rate-limit a
//// function with multiple arguments, use `apply2`, `apply3`, or `apply4`:
////
//// ```gleam
//// let limiter =
//// glimit.new()
//// |> glimit.per_second(10)
//// |> glimit.identifier(fn(args: #(String, String)) { args.0 })
//// |> glimit.on_limit_exceeded(fn(_args) { too_many_requests() })
////
//// let limited_handle =
//// handle
//// |> glimit.apply2(limiter)
////
//// limited_handle("user_123", "upload")
//// ```
////
import gleam/option.{type Option, None, Some}
import gleam/result
import glimit/rate_limiter
import glimit/registry.{type RateLimiterRegistryActor}
/// A rate limiter.
///
pub type RateLimiter(a, b, id) {
RateLimiter(
rate_limiter_registry: RateLimiterRegistryActor(id),
on_limit_exceeded: fn(a) -> b,
identifier: fn(a) -> id,
)
}
/// A builder for configuring the rate limiter.
///
pub type RateLimiterBuilder(a, b, id) {
RateLimiterBuilder(
per_second: Option(fn(id) -> Int),
burst_limit: Option(fn(id) -> Int),
identifier: Option(fn(a) -> id),
on_limit_exceeded: Option(fn(a) -> b),
)
}
/// Create a new rate limiter builder.
///
pub fn new() -> RateLimiterBuilder(a, b, id) {
RateLimiterBuilder(
per_second: None,
burst_limit: None,
identifier: None,
on_limit_exceeded: None,
)
}
/// Set the rate of new available tokens per second.
///
/// Note that this is not the maximum number of requests that can be made in a single
/// second, but the rate at which tokens are added to the bucket. Think of this as the
/// steady state rate limit, while the `burst_limit` function sets the maximum number of
/// available tokens (or the burst rate limit).
///
/// This value is also used as the default value for the `burst_limit` function.
///
/// # Example
///
/// ```gleam
/// import glimit
///
/// let limiter =
/// glimit.new()
/// |> glimit.per_second(10)
/// ```
///
pub fn per_second(
limiter: RateLimiterBuilder(a, b, id),
limit: Int,
) -> RateLimiterBuilder(a, b, id) {
RateLimiterBuilder(..limiter, per_second: Some(fn(_) { limit }))
}
/// Set the rate limit per second, based on the identifier.
///
/// # Example
///
/// ```gleam
/// import glimit
///
/// let limiter =
/// glimit.new()
/// |> glimit.identifier(fn(request) { request.user_id })
/// |> glimit.per_second_fn(fn(user_id) {
/// db.get_rate_limit(user_id)
/// })
/// ```
///
pub fn per_second_fn(
limiter: RateLimiterBuilder(a, b, id),
limit_fn: fn(id) -> Int,
) -> RateLimiterBuilder(a, b, id) {
RateLimiterBuilder(..limiter, per_second: Some(limit_fn))
}
/// Set the maximum number of available tokens.
///
/// The maximum number of available tokens is the maximum number of requests that can be
/// made in a single second. The default value is the same as the rate limit per second.
///
/// # Example
///
/// ```gleam
/// import glimit
///
/// let limiter =
/// glimit.new()
/// |> glimit.per_second(10)
/// |> glimit.burst_limit(100)
/// ```
///
pub fn burst_limit(
limiter: RateLimiterBuilder(a, b, id),
burst_limit: Int,
) -> RateLimiterBuilder(a, b, id) {
RateLimiterBuilder(..limiter, burst_limit: Some(fn(_) { burst_limit }))
}
/// Set the maximum number of available tokens, based on the identifier.
///
/// # Example
///
/// ```gleam
/// import glimit
///
/// let limiter =
/// glimit.new()
/// |> glimit.identifier(fn(request) { request.user_id })
/// |> glimit.per_second(10)
/// |> glimit.burst_limit_fn(fn(user_id) {
/// db.get_burst_limit(user_id)
/// })
/// ```
///
pub fn burst_limit_fn(
limiter: RateLimiterBuilder(a, b, id),
burst_limit_fn: fn(id) -> Int,
) -> RateLimiterBuilder(a, b, id) {
RateLimiterBuilder(..limiter, burst_limit: Some(burst_limit_fn))
}
/// Set the handler to be called when the rate limit is reached.
///
/// # Example
///
/// ```gleam
/// import glimit
///
/// let limiter =
/// glimit.new()
/// |> glimit.per_second(10)
/// |> glimit.on_limit_exceeded(fn(_request) { "Rate limit reached" })
/// ```
///
pub fn on_limit_exceeded(
limiter: RateLimiterBuilder(a, b, id),
on_limit_exceeded: fn(a) -> b,
) -> RateLimiterBuilder(a, b, id) {
RateLimiterBuilder(..limiter, on_limit_exceeded: Some(on_limit_exceeded))
}
/// Set the identifier function to be used to identify the rate limit.
///
/// # Example
///
/// ```gleam
/// import glimit
///
/// let limiter =
/// glimit.new()
/// |> glimit.identifier(fn(request) { request.ip })
/// ```
///
pub fn identifier(
limiter: RateLimiterBuilder(a, b, id),
identifier: fn(a) -> id,
) -> RateLimiterBuilder(a, b, id) {
RateLimiterBuilder(..limiter, identifier: Some(identifier))
}
/// Build the rate limiter.
///
/// Note that using `apply` will already build the rate limiter, so this function is
/// only useful if you want to build the rate limiter manually and apply it to multiple
/// functions.
///
/// To apply the resulting rate limiter to a function or handler, use the `apply_built`
/// function.
///
pub fn build(
config: RateLimiterBuilder(a, b, id),
) -> Result(RateLimiter(a, b, id), String) {
use per_second <- result.try(case config.per_second {
Some(per_second) -> Ok(per_second)
None -> Error("`per_second` rate limit is required")
})
let burst_limit = case config.burst_limit {
Some(burst_limit) -> burst_limit
None -> per_second
}
use rate_limiter_registry <- result.try(
registry.new(per_second, burst_limit)
|> result.map_error(fn(_) { "Failed to start rate limiter registry" }),
)
use identifier <- result.try(case config.identifier {
Some(identifier) -> Ok(identifier)
None -> Error("`identifier` function is required")
})
use on_limit_exceeded <- result.try(case config.on_limit_exceeded {
Some(on_limit_exceeded) -> Ok(on_limit_exceeded)
None -> Error("`on_limit_exceeded` function is required")
})
Ok(RateLimiter(
rate_limiter_registry: rate_limiter_registry,
on_limit_exceeded: on_limit_exceeded,
identifier: identifier,
))
}
/// Apply the rate limiter to a request handler or function.
///
/// Panics if the rate limiter registry cannot be started or if the `identifier`
/// function or `on_limit_exceeded` function is missing.
///
pub fn apply(
func: fn(a) -> b,
config: RateLimiterBuilder(a, b, id),
) -> fn(a) -> b {
let limiter = case build(config) {
Ok(limiter) -> limiter
Error(message) -> panic as message
}
apply_built(func, limiter)
}
/// Apply the rate limiter to a request handler or function.
///
/// This function is useful if you want to build the rate limiter manually using the
/// `build` function.
///
pub fn apply_built(
func: fn(a) -> b,
limiter: RateLimiter(a, b, id),
) -> fn(a) -> b {
fn(input: a) -> b {
let identifier = limiter.identifier(input)
case limiter.rate_limiter_registry |> registry.get_or_create(identifier) {
Ok(rate_limiter) -> {
case rate_limiter |> rate_limiter.hit {
Ok(Nil) -> func(input)
Error(rate_limiter.RateLimited) -> limiter.on_limit_exceeded(input)
Error(rate_limiter.Unavailable) -> func(input)
}
}
// Fail open — if rate limiting infrastructure fails, let requests through
Error(_) -> func(input)
}
}
}
/// Apply the rate limiter to a 2-argument function.
///
/// The config's `identifier` and `on_limit_exceeded` receive a `#(a, b)` tuple.
///
pub fn apply2(
func: fn(a, b) -> c,
config: RateLimiterBuilder(#(a, b), c, id),
) -> fn(a, b) -> c {
let wrapped =
fn(args: #(a, b)) -> c { func(args.0, args.1) }
|> apply(config)
fn(a: a, b: b) -> c { wrapped(#(a, b)) }
}
/// Apply the rate limiter to a 3-argument function.
///
/// The config's `identifier` and `on_limit_exceeded` receive a `#(a, b, c)` tuple.
///
pub fn apply3(
func: fn(a, b, c) -> d,
config: RateLimiterBuilder(#(a, b, c), d, id),
) -> fn(a, b, c) -> d {
let wrapped =
fn(args: #(a, b, c)) -> d { func(args.0, args.1, args.2) }
|> apply(config)
fn(a: a, b: b, c: c) -> d { wrapped(#(a, b, c)) }
}
/// Apply the rate limiter to a 4-argument function.
///
/// The config's `identifier` and `on_limit_exceeded` receive a `#(a, b, c, d)` tuple.
///
pub fn apply4(
func: fn(a, b, c, d) -> e,
config: RateLimiterBuilder(#(a, b, c, d), e, id),
) -> fn(a, b, c, d) -> e {
let wrapped =
fn(args: #(a, b, c, d)) -> e { func(args.0, args.1, args.2, args.3) }
|> apply(config)
fn(a: a, b: b, c: c, d: d) -> e { wrapped(#(a, b, c, d)) }
}