feat(#2635): consolidate todos + reminders into one hyperhive-state.sqlite

hive-agent/src/db_migrate.rs

@@ -0,0 +1,207 @@
+//! One-time on-boot migration from the pre-db-consolidation per-concern
+//! sqlite files (`hyperhive-todos.sqlite`, `hyperhive-reminders.sqlite`)
+//! into the single consolidated `hyperhive-state.sqlite` (mara: "not yet
+//! another sqlite ... add it to one of the existing ones"). Idempotent + best
+//! effort: `run` is called once in `main.rs`, before anything opens the
+//! consolidated db — a fresh harness boot with no legacy files (or one that
+//! already migrated) is a silent no-op.
+//!
+//! Two legacy shapes fold in:
+//! - `hyperhive-todos.sqlite` already has exactly the `todos` table the
+//!   consolidated db wants and nothing touches it during the rename window
+//!   (this runs before the todo/reminder subsystems start), so a plain
+//!   filesystem rename is correct and cheap — no sqlite involved.
+//! - `hyperhive-reminders.sqlite`'s `reminders` table then gets copied in via
+//!   `ATTACH DATABASE` + `INSERT INTO ... SELECT`, preserving row ids (a
+//!   reminder id an agent saw in a `get_loose_ends` call just before the
+//!   migrating boot must still `cancel` correctly after it), and the legacy
+//!   file is removed once the copy lands.
+
+use std::path::Path;
+
+use anyhow::{Context, Result};
+use rusqlite::{Connection, params};
+
+/// Fold the legacy `legacy_todos`/`legacy_reminders` files (if present) into
+/// the single consolidated `state_db` path. No-op if `state_db` already
+/// exists — legacy files are only ever consulted on the first boot after
+/// the db-consolidation upgrade.
+///
+/// # Errors
+///
+/// Propagates filesystem rename / sqlite attach-copy failures. Callers
+/// treat this as best-effort (log and continue) — `Todos`/`Reminders::open`
+/// against `state_db` still work standalone either way (todos would simply
+/// start empty rather than inheriting the legacy rows). Retry-next-boot
+/// only holds for failures before `state_db` exists (the todos rename, or
+/// the reminders copy itself): once the copy has landed, `state_db` exists,
+/// so a `remove_file(legacy_reminders)` failure afterwards leaves an
+/// orphaned (already-copied, now-inert) legacy file behind rather than
+/// retrying — the rows are safe, but the leftover file won't self-clean.
+pub fn run(state_db: &Path, legacy_todos: &Path, legacy_reminders: &Path) -> Result<()> {
+    if state_db.exists() {
+        return Ok(());
+    }
+    if legacy_todos.exists() {
+        std::fs::rename(legacy_todos, state_db).with_context(|| {
+            format!(
+                "rename legacy todos db {} -> {}",
+                legacy_todos.display(),
+                state_db.display()
+            )
+        })?;
+    }
+    if legacy_reminders.exists() {
+        // Ensure `state_db` exists and has the `reminders` schema — either
+        // the rename above already brought a todos-only db across (no
+        // `reminders` table yet), or there was no legacy todos file at all
+        // and this creates `state_db` fresh.
+        crate::reminders::Reminders::open(state_db)
+            .context("apply reminders schema to consolidated db")?;
+        migrate_reminders_rows(state_db, legacy_reminders)
+            .context("copy legacy reminders rows into consolidated db")?;
+        std::fs::remove_file(legacy_reminders).with_context(|| {
+            format!("remove legacy reminders db {}", legacy_reminders.display())
+        })?;
+    }
+    Ok(())
+}
+
+/// Copy every row of the legacy `reminders` table into `state_db`'s table
+/// of the same name, preserving `id`. Uses a scratch `ATTACH DATABASE` so
+/// the copy is one sqlite statement rather than a row-by-row read/insert
+/// loop.
+fn migrate_reminders_rows(state_db: &Path, legacy_reminders: &Path) -> Result<()> {
+    let conn = Connection::open(state_db)
+        .with_context(|| format!("open consolidated db {}", state_db.display()))?;
+    conn.execute(
+        "ATTACH DATABASE ?1 AS legacy",
+        params![legacy_reminders.to_string_lossy()],
+    )
+    .context("attach legacy reminders db")?;
+    let copied = conn
+        .execute_batch(
+            "INSERT INTO reminders (id, message, file_path, due_at, created_at, sent_at) \
+             SELECT id, message, file_path, due_at, created_at, sent_at FROM legacy.reminders;",
+        )
+        .context("copy legacy reminders rows");
+    // DETACH regardless of the copy's outcome so the connection doesn't
+    // leak the attachment on an early return via `?` inside this fn.
+    let detach = conn.execute("DETACH DATABASE legacy", []);
+    copied?;
+    detach.context("detach legacy reminders db")?;
+    Ok(())
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::{reminders::Reminders, todos::Todos};
+
+    #[test]
+    fn no_legacy_files_is_a_noop() {
+        let dir = tempfile::tempdir().unwrap();
+        let state = dir.path().join("hyperhive-state.sqlite");
+        let todos = dir.path().join("hyperhive-todos.sqlite");
+        let reminders = dir.path().join("hyperhive-reminders.sqlite");
+        run(&state, &todos, &reminders).unwrap();
+        assert!(!state.exists(), "nothing to migrate, nothing created");
+    }
+
+    #[test]
+    fn state_db_already_present_short_circuits() {
+        let dir = tempfile::tempdir().unwrap();
+        let state = dir.path().join("hyperhive-state.sqlite");
+        let legacy_todos = dir.path().join("hyperhive-todos.sqlite");
+        std::fs::write(&legacy_todos, b"should be left alone").unwrap();
+        Todos::open(&state).unwrap();
+        run(
+            &state,
+            &legacy_todos,
+            &dir.path().join("hyperhive-reminders.sqlite"),
+        )
+        .unwrap();
+        assert_eq!(
+            std::fs::read(&legacy_todos).unwrap(),
+            b"should be left alone",
+            "already-migrated boot must not touch a leftover legacy file"
+        );
+    }
+
+    #[test]
+    fn renames_legacy_todos_into_state_db() {
+        let dir = tempfile::tempdir().unwrap();
+        let legacy_todos = dir.path().join("hyperhive-todos.sqlite");
+        let state = dir.path().join("hyperhive-state.sqlite");
+        {
+            let store = Todos::open(&legacy_todos).unwrap();
+            store
+                .upsert("matrix", Some("!a:x"), "1 unread", None)
+                .unwrap();
+        }
+        run(
+            &state,
+            &legacy_todos,
+            &dir.path().join("hyperhive-reminders.sqlite"),
+        )
+        .unwrap();
+        assert!(!legacy_todos.exists(), "renamed away");
+        let migrated = Todos::open(&state).unwrap();
+        let rows = migrated.list(None).unwrap();
+        assert_eq!(rows.len(), 1);
+        assert_eq!(rows[0].summary, "1 unread");
+    }
+
+    #[test]
+    fn copies_legacy_reminders_preserving_ids() {
+        let dir = tempfile::tempdir().unwrap();
+        let legacy_reminders = dir.path().join("hyperhive-reminders.sqlite");
+        let state = dir.path().join("hyperhive-state.sqlite");
+        let orig_id = {
+            let store = Reminders::open(&legacy_reminders).unwrap();
+            store.store("check on x", None, 1000).unwrap()
+        };
+        run(
+            &state,
+            &dir.path().join("hyperhive-todos.sqlite"),
+            &legacy_reminders,
+        )
+        .unwrap();
+        assert!(!legacy_reminders.exists(), "removed after copy");
+        let migrated = Reminders::open(&state).unwrap();
+        let pending = migrated.list_pending().unwrap();
+        assert_eq!(pending.len(), 1);
+        assert_eq!(pending[0].id, orig_id, "id preserved across the copy");
+        assert_eq!(pending[0].due_at, 1000);
+    }
+
+    #[test]
+    fn folds_both_legacy_files_into_one_consolidated_db() {
+        let dir = tempfile::tempdir().unwrap();
+        let legacy_todos = dir.path().join("hyperhive-todos.sqlite");
+        let legacy_reminders = dir.path().join("hyperhive-reminders.sqlite");
+        let state = dir.path().join("hyperhive-state.sqlite");
+        {
+            Todos::open(&legacy_todos)
+                .unwrap()
+                .upsert("bash", None, "task done", None)
+                .unwrap();
+            Reminders::open(&legacy_reminders)
+                .unwrap()
+                .store("ping", None, 500)
+                .unwrap();
+        }
+        run(&state, &legacy_todos, &legacy_reminders).unwrap();
+        assert!(!legacy_todos.exists());
+        assert!(!legacy_reminders.exists());
+        assert_eq!(Todos::open(&state).unwrap().list(None).unwrap().len(), 1);
+        assert_eq!(
+            Reminders::open(&state)
+                .unwrap()
+                .list_pending()
+                .unwrap()
+                .len(),
+            1
+        );
+    }
+}

hive-agent/src/main.rs

@@ -10,6 +10,7 @@
 //! before lib + bin were collapsed into one) plus the serve loop.
 
 mod client;
+mod db_migrate;
 mod events;
 mod forge_notify;
 mod harness_state;
@@ -467,6 +468,19 @@ async fn serve_main<S: Surface>(socket: &Path, poll_ms: u64) -> Result<()> {
     // never observes a "closed" state. Opened before the todo socket
     // below so the same `Arc` can be handed to its request dispatch
     // (reminder ops share the todo socket/listener).
+    // Fold the pre-consolidation per-concern db files into the single
+    // `hyperhive-state.sqlite`, if this is the first boot since the
+    // upgrade — see `db_migrate` module docs. Best-effort: a failure here
+    // just leaves the legacy file(s) in place to retry next boot, so the
+    // subsystems below still start (`Todos`/`Reminders::open` create the
+    // schema fresh if nothing migrated across).
+    if let Err(e) = db_migrate::run(
+        &paths::state_db(),
+        &paths::legacy_todos_db(),
+        &paths::legacy_reminders_db(),
+    ) {
+        tracing::warn!(error = ?e, "legacy db_migrate failed — continuing with fresh/partial state db");
+    }
     let (reminder_tx, reminder_rx) = tokio::sync::mpsc::unbounded_channel();
     let reminder_store = match reminders::Reminders::open(&paths::reminders_db()) {
         Ok(store) => Some(Arc::new(store)),

hive-agent/src/paths.rs

@@ -40,19 +40,52 @@ pub fn harness_dir() -> PathBuf {
     hive_sh4re::paths::harness_dir()
 }
 
-/// Harness-local todo store (loose-ends v2). A dedicated sqlite db under
-/// the harness dir — the todos are mutable per-agent state the harness
-/// owns, kept out of the append-only `hyperhive-events.sqlite` sink.
+/// Consolidated harness-local state db — currently todos + reminders, one
+/// table each — mutable per-agent state the harness owns, kept out of the
+/// append-only `hyperhive-events.sqlite` sink. Per mara's call ("not yet
+/// another sqlite! todos, reminders, questions should be like three tiny
+/// tables in one 500kb sqlite"), this file is the shared home for all
+/// loose-ends-v2 stores; each store's `open()` only applies its own
+/// `CREATE TABLE IF NOT EXISTS`, so opening multiple stores against the
+/// same path is safe (distinct table names, no schema collision). A
+/// questions mirror table is the planned third tenant (a following
+/// increment), not part of this schema yet.
+/// [`todos_db`] and [`reminders_db`] both resolve here — kept as separate
+/// fns (rather than every call site reading `state_db` directly) so each
+/// store's callers still say what they mean.
+///
+/// Before this consolidation, todos and reminders lived in their own
+/// `hyperhive-todos.sqlite` / `hyperhive-reminders.sqlite` files; a
+/// one-time boot migration (`db_migrate::run`) folds those into this path
+/// the first time a harness boots after the upgrade.
+#[must_use]
+pub fn state_db() -> PathBuf {
+    harness_dir().join("hyperhive-state.sqlite")
+}
+
+/// Harness-local todo store (loose-ends v2) path — see [`state_db`].
 #[must_use]
 pub fn todos_db() -> PathBuf {
+    state_db()
+}
+
+/// Harness-local reminder store path — see [`state_db`].
+#[must_use]
+pub fn reminders_db() -> PathBuf {
+    state_db()
+}
+
+/// Legacy pre-consolidation todos db path, consulted only by
+/// [`crate::db_migrate`] on the first boot after the upgrade.
+#[must_use]
+pub fn legacy_todos_db() -> PathBuf {
     harness_dir().join("hyperhive-todos.sqlite")
 }
 
-/// Harness-local reminder store. Same rationale as [`todos_db`] — mutable
-/// per-agent state the harness owns, kept out of the append-only events
-/// sink.
+/// Legacy pre-consolidation reminders db path, consulted only by
+/// [`crate::db_migrate`] on the first boot after the upgrade.
 #[must_use]
-pub fn reminders_db() -> PathBuf {
+pub fn legacy_reminders_db() -> PathBuf {
     harness_dir().join("hyperhive-reminders.sqlite")
 }