Struct FileJsonSink 

pub struct FileJsonSink {
    pub path: PathBuf,
    pub file: Arc<Mutex<File>>,
}

File-backed sink. Records are appended to path as JSONL.

The file is opened once at construction with O_APPEND (or the platform equivalent) and writes serialise through an inner std::sync::Mutex. Each emit performs write_all + flush so a process crash loses at most the in-flight record. See module docs for the latency rationale.

Fields

path: PathBuf

Path the sink was constructed with — retained for diagnostics.

file: Arc<Mutex<File>>

The append-only file handle. std::sync::Mutex (not tokio::sync::Mutex) because the critical section is sync I/O with no .await point.

PERF: wrapped in an Arc so AuditSink::emit can hand a cheap clone of the handle to a tokio::task::spawn_blocking worker and offload the blocking write_all + flush off the async response path (see the emit impl below). The Arc is the minimal change that lets the &self-borrowed mutex outlive the request future on the blocking pool.

Implementations

impl FileJsonSink

pub fn open(path: impl Into<PathBuf>) -> Result<Self>

Open path for append, creating it if missing. Returns an io error if the OS refuses (typically permissions or a missing parent directory).

Trait Implementations

impl AuditSink for FileJsonSink

fn emit(&self, record: &AuditRecord)

Persist record. Errors are intentionally absorbed: an audit failure must never bubble up and corrupt a successful API call. Implementations should log internally via tracing::error! instead of propagating.

impl Debug for FileJsonSink

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.