Struct S3State 

pub struct S3State {
    pub account_id: String,
    pub region: String,
    pub buckets: BTreeMap<String, S3Bucket>,
    pub notification_events: Vec<S3NotificationEvent>,
    pub body_cache: Option<Arc<BodyCache>>,
    pub object_lambda_responses: BTreeMap<String, ObjectLambdaResponse>,
    pub access_points: BTreeMap<String, S3AccessPoint>,
}

Fields

account_id: String

region: String

buckets: BTreeMap<String, S3Bucket>

notification_events: Vec<S3NotificationEvent>

body_cache: Option<Arc<BodyCache>>

object_lambda_responses: BTreeMap<String, ObjectLambdaResponse>

Object Lambda responses keyed by request token.

access_points: BTreeMap<String, S3AccessPoint>

Implementations

impl S3State

pub fn new(account_id: &str, region: &str) -> Self

pub fn set_body_cache(&mut self, cache: Arc<BodyCache>)

pub fn reset(&mut self)

pub fn read_body_uncached(body: &BodyRef) -> Result<Bytes>

Read the full body referenced by a BodyRef without touching the BodyCache or any S3State. Because it borrows nothing from state, callers can read part bodies after dropping the global S3 lock — used by CompleteMultipartUpload to assemble a multi-GB object off-lock instead of serializing every other S3 operation behind the assembly (bug-audit 2026-05-28, 4.7). Disk bodies are read straight from their path.

pub fn read_body(&self, body: &BodyRef) -> Result<Bytes>

Read the full body referenced by a BodyRef, consulting the persistent BodyCache when one is configured.

pub fn read_body_range( &self, body: &BodyRef, offset: u64, len: u64, ) -> Result<Bytes>

Read a byte range from the body without loading the full object into memory. Memory bodies are sliced directly; disk bodies are seek+read’d. Ranges bypass the body cache (the cache stores whole objects only).

Trait Implementations

impl AccountState for S3State

fn new_for_account(account_id: &str, region: &str, _endpoint: &str) -> Self

Create a fresh, empty state for the given account.

fn inherit_from(&mut self, sibling: &Self)

Called after a new account state is created via MultiAccountState::get_or_create, with a reference to an existing sibling state. Services can override this to propagate shared resources (e.g. body caches) to the new state.