Struct AppState 

pub struct AppState {
    pub sites: Arc<[Site]>,
    pub client: Arc<Client>,
    pub scans: Arc<RwLock<HashMap<ScanId, ScanHandle>>>,
    pub scan_tasks: Arc<RwLock<HashMap<ScanId, JoinHandle<()>>>>,
    pub scan_capacity: usize,
    pub scans_dir: Option<Arc<PathBuf>>,
}

State shared across all axum handlers.

Cheap to clone — every field is an Arc or a small primitive. axum requires State<T> to be Clone, hence this design.

Fields

sites: Arc<[Site]>

Pre-filtered site list (registry + workspace flags applied at startup). Held as an Arc<[Site]> to avoid re-cloning the 2.5k-entry vector on every scan dispatch.

client: Arc<Client>

Shared HTTP client (connection pool, throttle, etc.).

scans: Arc<RwLock<HashMap<ScanId, ScanHandle>>>

In-flight + recently-finished scans, keyed by ID.

scan_tasks: Arc<RwLock<HashMap<ScanId, JoinHandle<()>>>>

Running-scan task handles, keyed by ScanId. Lets the refilter endpoint cancel an in-flight scan via JoinHandle::abort before spawning a successor with the new filter. Entries are removed when their scan finishes naturally (the task’s last act before returning) or when the eviction policy reaps them alongside the ScanHandle.

scan_capacity: usize

Maximum number of scans retained in memory. Beyond this, the oldest finished scan is evicted on the next insertion (a tiny LRU — we never need more than ~dozens of recent scans in a human-driven web session).

scans_dir: Option<Arc<PathBuf>>

Directory where finished scans are persisted as JSON. None disables persistence (used by tests and ephemeral runs).

Implementations

impl AppState

pub fn new(sites: Vec<Site>, client: Client, scan_capacity: usize) -> Self

Build initial state from a registry + a pre-built HTTP client.

The full registry is filtered with the supplied predicate; the result is materialised into an Arc<[Site]> once so handler dispatch is a pointer copy. Persistence is off by default — chain Self::with_scans_dir to enable.

pub fn from_registry( registry: &Registry, client: Client, scan_capacity: usize, ) -> Self

Convenience: build state from a Registry using the “no filter, NSFW excluded” default. The web UI exposes per-scan filters anyway, so the initial site list is the full non-NSFW set.

pub fn with_scans_dir(self, dir: PathBuf) -> Self

Enable on-disk persistence of finished scans under dir. Files are written as <scan_id>.json after each scan completes; startup reads them back so history survives server restarts.

pub async fn insert_scan(&self, id: ScanId, handle: ScanHandle)

Insert a fresh scan handle, evicting the oldest finished entry (or the oldest entry overall, if none has finished) when we are at capacity.

pub async fn register_scan_task(&self, id: ScanId, task: JoinHandle<()>)

Register an in-flight scan task. The handle is stored so the refilter endpoint can abort it before starting a successor.

pub async fn forget_scan_task(&self, id: &ScanId)

Remove an in-flight scan task entry. Used at the end of crate::scan::run so the map doesn’t accumulate completed tasks.

pub async fn abort_scan(&self, id: &ScanId) -> bool

Abort the running task for id (if any). Returns true when an abort signal was actually sent; false when no live task was recorded (already finished, or never started). Doesn’t wait for the task to observe the abort — JoinHandle::abort is non-blocking and the caller continues immediately.

pub async fn get_scan(&self, id: &ScanId) -> Option<ScanHandle>

Look up a scan by ID, cloning the handle (cheap — Arc inside).

Trait Implementations

impl Clone for AppState

fn clone(&self) -> AppState

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.