Function run_git_backfill

src/sessions_cmd.rs:196–269

(
    project_id: Option<String>,
    project_path: Option<String>,
    since: Option<String>,
    limit_sessions: usize,
    dry_run: bool,
)

const GIT_BACKFILL_ANALYTICS_LIMIT: usize = 500_000;

async fn run_git_backfill(
    project_id: Option<String>,
    project_path: Option<String>,
    since: Option<String>,
    limit_sessions: usize,
    dry_run: bool,
) -> tracedecay::errors::Result<()> {
    use tracedecay::sessions::git_correlation::{
        run_backfill, BackfillOptions, SystemGit, DEFAULT_SPAN_MERGE_GAP_SECS,
    };

    let project_root = resolve_cli_project_root(None, project_id, project_path).await?;
    let since_ts = resolve_backfill_since(since.as_deref())?;

    let session_db = tracedecay::sessions::cursor::open_project_session_db(&project_root)
        .await
        .ok_or_else(|| tracedecay::errors::TraceDecayError::Config {
            message: format!(
                "could not open project session database for {}",
                project_root.display()
            ),
        })?;

    // Global analytics rows are a finer-grained timestamp signal. Missing or
    // unreadable analytics is non-fatal: the backfill still runs on session
    // timestamps and the reflog. Scope the query to this project and the
    // backfill window: only rows for the target project within [since_ts, now]
    // are ever consumed, so pulling every event from every project (these
    // stores reach tens of GB) would be pure waste.
    let analytics_events = match tracedecay::global_db::GlobalDb::open().await {
        Some(global) => global
            .query_analytics_events(&tracedecay::global_db::AnalyticsEventQuery {
                project_id: Some(tracedecay::global_db::GlobalDb::canonical_project_key(
                    &project_root,
                )),
                since: Some(since_ts),
                limit: GIT_BACKFILL_ANALYTICS_LIMIT,
                ..Default::default()
            })
            .await
            .unwrap_or_default(),
        None => Vec::new(),
    };

    let opts = BackfillOptions {
        since: since_ts,
        limit_sessions,
        merge_gap_secs: DEFAULT_SPAN_MERGE_GAP_SECS,
        max_commits_per_repo: 5_000,
        dry_run,
    };

    let git = SystemGit;
    let stats = run_backfill(&session_db, &analytics_events, &git, &opts)
        .await
        .map_err(|err| tracedecay::errors::TraceDecayError::Config {
            message: format!("git backfill failed: {err}"),
        })?;