Struct IndexManager 

pub struct IndexManager {
    pub schema_version: u32,
    /* private fields */
}

Manages the SQLite persistence layer.

Owns one write Connection (used by the writer task) and exposes read_connection() to open additional per-thread read-only connections.

Fields

schema_version: u32

Current schema version after migration.

Implementations

impl IndexManager

pub fn open(path: &str) -> Result<Self, TuiError>

Open (or create) the SQLite index at path with WAL mode.

Uses SQLITE_OPEN_NO_MUTEX — thread safety is handled by Rust’s ownership model (Req 19.9). Runs schema migrations on every open (Req 19.4).

pub fn open_in_memory() -> Result<Self, TuiError>

Open an in-memory database — useful for tests and as a fallback when the on-disk file is corrupt or inaccessible (Req 19.5).

pub fn migrate(&self) -> Result<u32, TuiError>

Run pending schema migrations sequentially, preserving existing data.

Detects the current version from schema_meta (0 if the table doesn’t yet exist), applies each migration whose target version is higher, and updates schema_version afterward. Each migration runs inside its own SAVEPOINT so a failure leaves the database at a clean prior version rather than in an unknown intermediate state (Req 19.4).

pub fn read_connection(&self) -> Result<Connection, TuiError>

Open a fresh read-only Connection for use in a caller thread.

Each call returns an independent Connection instance, satisfying the requirement that every thread or task gets its own connection (Req 19.8, 19.9).

In-memory managers share the write connection for reads (the in-memory database is not accessible from a second connection object).

pub fn write_conn(&self) -> &Connection

Borrow the write connection for direct use (e.g. from tests or the writer task before the channel is set up).

pub fn is_scan_stale(&self, workspace_path: &str, fs_mtime: i64) -> bool

Returns true if the cached scan for workspace_path is older than fs_mtime (Unix seconds), meaning a re-scan is needed.

Returns true when there is no cached entry (first scan).

pub fn load_cached_scan_paths(&self) -> Vec<String>

Return all workspace paths that have cached scan entries.

pub fn load_drift_history( &self, ) -> Vec<(String, String, String, Option<String>)>

Load all drift-history rows as (workspace_path, asset_id, drift_type, resolved_at) tuples, ordered by id. resolved_at is None for still-active drift.

pub fn load_coverage_scores(&self) -> Vec<(String, f64)>

Load cached per-workspace coverage scores as (workspace_path, score) tuples, ordered by workspace path.