refactor: 5 architectural improvements (Tasks 2-5 + Task 6 fix)

- feat(domain): Hashtag value object with canonical extract() — unifies two
  divergent private implementations; fields pre-compute raw/normalized/url_slug/ap_name

- feat(presentation): Deps<S: FromAppState> extractor — each handler now
  declares its exact dependency surface; AppState unchanged; handlers
  become unit-testable without mocking all 20 deps

- refactor(feed): replace 5 flat FeedRepository methods with FeedQuery/FeedScope
  — single query() method; SQL shared logic lives once; adding feed types
  no longer requires 5 edits

- refactor(activitypub): ActivityPubRepository + OutboundFederationPort moved
  out of domain::ports into activitypub-base::ap_ports — domain crate no
  longer knows about AP IDs, inboxes, or actor URLs

- fix(outbox): OutboxRelay now opens a per-row transaction so FOR UPDATE
  SKIP LOCKED actually holds the lock during publish + mark_delivered

crates/domain/src/hashtag.rs

@@ -0,0 +1,139 @@
+use std::collections::HashSet;
+
+/// A hashtag extracted from content.
+#[derive(Debug, Clone, PartialEq, Eq, Hash)]
+pub struct Hashtag {
+    /// Original casing, e.g. "Rust"
+    pub raw: String,
+    /// Lowercased, e.g. "rust" — used for DB lookups
+    pub normalized: String,
+    /// "tags/rust" — callers prepend base_url
+    pub url_slug: String,
+    /// "#rust" — used directly in AP tag array
+    pub ap_name: String,
+}
+
+/// Extract hashtags from content using a char-by-char scan.
+///
+/// Rules:
+/// - Tag starts after a bare `#` followed immediately by an alphanumeric char.
+/// - Tag chars: `[A-Za-z0-9_]`.
+/// - Deduplicated case-insensitively; first occurrence wins.
+/// - Returned in order of first appearance.
+pub fn extract(content: &str) -> Vec<Hashtag> {
+    let mut seen: HashSet<String> = HashSet::new();
+    let mut tags: Vec<Hashtag> = Vec::new();
+    let mut chars = content.char_indices().peekable();
+
+    while let Some((_, c)) = chars.next() {
+        if c == '#'
+            && chars
+                .peek()
+                .map(|(_, nc)| nc.is_alphanumeric())
+                .unwrap_or(false)
+        {
+            let raw: String = chars
+                .by_ref()
+                .take_while(|(_, nc)| nc.is_alphanumeric() || *nc == '_')
+                .map(|(_, nc)| nc)
+                .collect();
+
+            if raw.is_empty() {
+                continue;
+            }
+
+            let normalized = raw.to_lowercase();
+            if seen.insert(normalized.clone()) {
+                tags.push(Hashtag {
+                    url_slug: format!("tags/{}", normalized),
+                    ap_name: format!("#{}", normalized),
+                    raw,
+                    normalized,
+                });
+            }
+        }
+    }
+
+    tags
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    fn names(tags: &[Hashtag]) -> Vec<&str> {
+        tags.iter().map(|h| h.normalized.as_str()).collect()
+    }
+
+    #[test]
+    fn basic() {
+        let tags = extract("Hello #world and #Rust!");
+        assert_eq!(names(&tags), ["world", "rust"]);
+    }
+
+    #[test]
+    fn fields() {
+        let tags = extract("#Rust");
+        assert_eq!(tags.len(), 1);
+        let h = &tags[0];
+        assert_eq!(h.raw, "Rust");
+        assert_eq!(h.normalized, "rust");
+        assert_eq!(h.url_slug, "tags/rust");
+        assert_eq!(h.ap_name, "#rust");
+    }
+
+    #[test]
+    fn dedup_case_insensitive() {
+        let tags = extract("#rust #Rust #RUST");
+        assert_eq!(names(&tags), ["rust"]);
+        assert_eq!(tags[0].raw, "rust"); // first occurrence wins
+    }
+
+    #[test]
+    fn deduplicates_non_adjacent() {
+        // The old algorithm used Vec::dedup() which only removes adjacent duplicates.
+        // Using HashSet silently fixed this bug. This test documents the fix.
+        let tags = extract("#a #b #a");
+        assert_eq!(tags.len(), 2);
+        assert_eq!(tags[0].normalized, "a");
+        assert_eq!(tags[1].normalized, "b");
+    }
+
+    #[test]
+    fn mid_word_extracted() {
+        // `text#tag` — `#` not preceded by whitespace is still matched by the
+        // char-by-char scan (the old algorithm didn't require whitespace before `#`).
+        // This test documents the authoritative behaviour: mid-word tags ARE extracted.
+        let tags = extract("text#tag");
+        assert_eq!(names(&tags), ["tag"]);
+    }
+
+    #[test]
+    fn hash_only_ignored() {
+        assert!(extract("# lone hash").is_empty());
+    }
+
+    #[test]
+    fn trailing_punctuation_excluded() {
+        // punctuation after tag terminates the tag, not included
+        let tags = extract("#rust.");
+        assert_eq!(names(&tags), ["rust"]);
+    }
+
+    #[test]
+    fn underscore_allowed() {
+        let tags = extract("#hello_world");
+        assert_eq!(names(&tags), ["hello_world"]);
+    }
+
+    #[test]
+    fn empty_content() {
+        assert!(extract("").is_empty());
+    }
+
+    #[test]
+    fn order_of_appearance() {
+        let tags = extract("#b #a #c");
+        assert_eq!(names(&tags), ["b", "a", "c"]);
+    }
+}

crates/domain/src/lib.rs

@@ -1,5 +1,6 @@
 pub mod errors;
 pub mod events;
+pub mod hashtag;
 pub mod models;
 pub mod ports;
 pub mod value_objects;

crates/domain/src/ports.rs

@@ -317,37 +317,43 @@ impl<
 {
 }
 
+#[derive(Debug, Clone)]
+pub enum FeedScope {
+    Home { following_ids: Vec<UserId> },
+    Public,
+    Tag { tag_name: String },
+    User { user_id: UserId },
+    Search { query: String },
+}
+
+#[derive(Debug, Clone)]
+pub struct FeedQuery {
+    pub scope: FeedScope,
+    pub page: PageParams,
+    pub viewer_id: Option<UserId>,
+}
+
+impl FeedQuery {
+    pub fn home(viewer_id: UserId, following_ids: Vec<UserId>, page: PageParams) -> Self {
+        Self { scope: FeedScope::Home { following_ids }, page, viewer_id: Some(viewer_id) }
+    }
+    pub fn public(page: PageParams, viewer_id: Option<UserId>) -> Self {
+        Self { scope: FeedScope::Public, page, viewer_id }
+    }
+    pub fn tag(tag_name: impl Into<String>, page: PageParams, viewer_id: Option<UserId>) -> Self {
+        Self { scope: FeedScope::Tag { tag_name: tag_name.into() }, page, viewer_id }
+    }
+    pub fn user(user_id: UserId, page: PageParams, viewer_id: Option<UserId>) -> Self {
+        Self { scope: FeedScope::User { user_id }, page, viewer_id }
+    }
+    pub fn search(query: impl Into<String>, page: PageParams, viewer_id: Option<UserId>) -> Self {
+        Self { scope: FeedScope::Search { query: query.into() }, page, viewer_id }
+    }
+}
+
 #[async_trait]
 pub trait FeedRepository: Send + Sync {
-    async fn home_feed(
-        &self,
-        following_ids: &[UserId],
-        page: &PageParams,
-        viewer_id: Option<&UserId>,
-    ) -> Result<Paginated<FeedEntry>, DomainError>;
-    async fn public_feed(
-        &self,
-        page: &PageParams,
-        viewer_id: Option<&UserId>,
-    ) -> Result<Paginated<FeedEntry>, DomainError>;
-    async fn search(
-        &self,
-        query: &str,
-        page: &PageParams,
-        viewer_id: Option<&UserId>,
-    ) -> Result<Paginated<FeedEntry>, DomainError>;
-    async fn tag_feed(
-        &self,
-        tag_name: &str,
-        page: &PageParams,
-        viewer_id: Option<&UserId>,
-    ) -> Result<Paginated<FeedEntry>, DomainError>;
-    async fn user_feed(
-        &self,
-        user_id: &UserId,
-        page: &PageParams,
-        viewer_id: Option<&UserId>,
-    ) -> Result<Paginated<FeedEntry>, DomainError>;
+    async fn query(&self, q: &FeedQuery) -> Result<Paginated<FeedEntry>, DomainError>;
 }
 
 #[async_trait]
@@ -368,166 +374,6 @@ pub trait SearchPort: Send + Sync {
     ) -> Result<Paginated<User>, DomainError>;
 }
 
-/// AP-protocol endpoints for a locally-stored user (local or interned remote).
-#[derive(Debug, Clone)]
-pub struct ActorApUrls {
-    pub ap_id: String,
-    pub inbox_url: String,
-}
-
-/// A local thought ready for AP serialization, with the author's username
-/// pre-joined so the handler can build AP URLs without a second query.
-#[derive(Debug, Clone)]
-pub struct OutboxEntry {
-    pub thought: crate::models::thought::Thought,
-    pub author_username: Username,
-}
-
-#[async_trait]
-pub trait ActivityPubRepository: Send + Sync {
-    // ── Outbox (local → remote) ──────────────────────────────────────
-
-    /// All public local thoughts for this actor. Used for outbox totals
-    /// and full-collection delivery.
-    async fn outbox_entries_for_actor(
-        &self,
-        user_id: &UserId,
-    ) -> Result<Vec<OutboxEntry>, DomainError>;
-
-    /// Cursor page of public local thoughts, newest-first, before `before`.
-    /// Used for OrderedCollectionPage responses.
-    async fn outbox_page_for_actor(
-        &self,
-        user_id: &UserId,
-        before: Option<chrono::DateTime<chrono::Utc>>,
-        limit: usize,
-    ) -> Result<Vec<OutboxEntry>, DomainError>;
-
-    // ── Remote actor resolution ──────────────────────────────────────
-
-    /// Find the local UserId for a remote actor by its AP URL.
-    async fn find_remote_actor_id(&self, actor_ap_url: &str)
-        -> Result<Option<UserId>, DomainError>;
-
-    /// Ensure a remote actor placeholder exists; create one if absent.
-    /// Idempotent — safe to call multiple times with the same URL.
-    async fn intern_remote_actor(&self, actor_ap_url: &str) -> Result<UserId, DomainError>;
-
-    /// Update display_name and avatar_url for an already-interned remote actor.
-    async fn update_remote_actor_display(
-        &self,
-        user_id: &UserId,
-        display_name: Option<&str>,
-        avatar_url: Option<&str>,
-    ) -> Result<(), DomainError>;
-
-    // ── Inbox processing (remote → local) ───────────────────────────
-
-    /// Persist an incoming remote Note. Idempotent on ap_id.
-    #[allow(clippy::too_many_arguments)]
-    async fn accept_note(
-        &self,
-        ap_id: &str,
-        author_id: &UserId,
-        content: &str,
-        published: chrono::DateTime<chrono::Utc>,
-        sensitive: bool,
-        content_warning: Option<String>,
-        visibility: &str,
-        in_reply_to: Option<&str>,
-    ) -> Result<(), DomainError>;
-
-    /// Apply an Update to a previously accepted remote Note.
-    async fn apply_note_update(&self, ap_id: &str, new_content: &str) -> Result<(), DomainError>;
-
-    /// Remove a specific remote Note (Delete activity). Only touches
-    /// remotely-originated thoughts.
-    async fn retract_note(&self, ap_id: &str) -> Result<(), DomainError>;
-
-    /// Remove all Notes from a remote actor (actor-level Delete/Tombstone).
-    async fn retract_actor_notes(&self, actor_ap_url: &str) -> Result<(), DomainError>;
-
-    // ── Node-level stats ─────────────────────────────────────────────
-
-    /// Total locally-authored thought count for NodeInfo responses.
-    async fn count_local_notes(&self) -> Result<u64, DomainError>;
-
-    /// Return the ActivityPub object URL for a thought, if one is stored.
-    /// Returns None for local thoughts (caller constructs URL from base_url + thought_id).
-    async fn get_thought_ap_id(
-        &self,
-        thought_id: &ThoughtId,
-    ) -> Result<Option<String>, DomainError>;
-
-    /// Return the AP actor URL and inbox URL for a user, if stored.
-    /// Returns None for users that have not been federated.
-    async fn get_actor_ap_urls(&self, user_id: &UserId)
-        -> Result<Option<ActorApUrls>, DomainError>;
-}
-
-#[async_trait]
-pub trait OutboundFederationPort: Send + Sync {
-    /// Fan out a new local Note to all accepted followers.
-    async fn broadcast_create(
-        &self,
-        author_user_id: &UserId,
-        thought: &Thought,
-        author_username: &str,
-        in_reply_to_url: Option<&str>,
-    ) -> Result<(), DomainError>;
-
-    /// Fan out a Delete tombstone for a now-deleted local Note.
-    /// `thought_ap_id` is pre-constructed by the caller because the thought
-    /// has already been deleted from the DB when this fires.
-    async fn broadcast_delete(
-        &self,
-        author_user_id: &UserId,
-        thought_ap_id: &str,
-    ) -> Result<(), DomainError>;
-
-    /// Fan out an Update(Note) for an edited local thought.
-    async fn broadcast_update(
-        &self,
-        author_user_id: &UserId,
-        thought: &Thought,
-        author_username: &str,
-        in_reply_to_url: Option<&str>,
-    ) -> Result<(), DomainError>;
-
-    /// Fan out an Announce(object_ap_id) for a boost.
-    async fn broadcast_announce(
-        &self,
-        booster_user_id: &UserId,
-        object_ap_id: &str,
-    ) -> Result<(), DomainError>;
-
-    /// Fan out an Undo(Announce) to followers when a boost is removed.
-    async fn broadcast_undo_announce(
-        &self,
-        booster_user_id: &UserId,
-        object_ap_id: &str,
-    ) -> Result<(), DomainError>;
-
-    /// Send a Like activity to a remote thought author's inbox.
-    /// Only called when a LOCAL user likes a REMOTE thought (one with an ap_id).
-    async fn broadcast_like(
-        &self,
-        liker_user_id: &UserId,
-        object_ap_id: &str,
-        author_inbox_url: &str,
-    ) -> Result<(), DomainError>;
-
-    /// Send Undo(Like) to a remote thought author's inbox.
-    async fn broadcast_undo_like(
-        &self,
-        liker_user_id: &UserId,
-        object_ap_id: &str,
-        author_inbox_url: &str,
-    ) -> Result<(), DomainError>;
-
-    /// Fan out an Update(Actor) to all accepted followers after a profile change.
-    async fn broadcast_actor_update(&self, user_id: &UserId) -> Result<(), DomainError>;
-}
 
 #[async_trait]
 pub trait FederationSchedulerPort: Send + Sync {

crates/domain/src/testing.rs

@@ -39,8 +39,6 @@ pub struct TestStore {
     pub actor_ap_ids: Arc<Mutex<HashMap<String, UserId>>>,
     /// ThoughtId → AP object URL (used by get_thought_ap_id)
     pub thought_ap_ids: Arc<Mutex<HashMap<ThoughtId, String>>>,
-    /// UserId → ActorApUrls (used by get_actor_ap_urls)
-    pub actor_ap_urls: Arc<Mutex<HashMap<UserId, ActorApUrls>>>,
 }
 
 #[async_trait]
@@ -706,63 +704,7 @@ impl RemoteActorConnectionRepository for TestStore {
 
 #[async_trait]
 impl FeedRepository for TestStore {
-    async fn home_feed(
-        &self,
-        _ids: &[UserId],
-        _p: &PageParams,
-        _v: Option<&UserId>,
-    ) -> Result<Paginated<FeedEntry>, DomainError> {
-        Ok(Paginated {
-            items: vec![],
-            total: 0,
-            page: 1,
-            per_page: 20,
-        })
-    }
-    async fn public_feed(
-        &self,
-        _p: &PageParams,
-        _v: Option<&UserId>,
-    ) -> Result<Paginated<FeedEntry>, DomainError> {
-        Ok(Paginated {
-            items: vec![],
-            total: 0,
-            page: 1,
-            per_page: 20,
-        })
-    }
-    async fn search(
-        &self,
-        _q: &str,
-        _p: &PageParams,
-        _v: Option<&UserId>,
-    ) -> Result<Paginated<FeedEntry>, DomainError> {
-        Ok(Paginated {
-            items: vec![],
-            total: 0,
-            page: 1,
-            per_page: 20,
-        })
-    }
-    async fn tag_feed(
-        &self,
-        _tag_name: &str,
-        _page: &PageParams,
-        _viewer_id: Option<&UserId>,
-    ) -> Result<Paginated<FeedEntry>, DomainError> {
-        Ok(Paginated {
-            items: vec![],
-            total: 0,
-            page: 1,
-            per_page: 20,
-        })
-    }
-    async fn user_feed(
-        &self,
-        _user_id: &UserId,
-        _page: &PageParams,
-        _viewer_id: Option<&UserId>,
-    ) -> Result<Paginated<FeedEntry>, DomainError> {
+    async fn query(&self, _q: &crate::ports::FeedQuery) -> Result<Paginated<FeedEntry>, DomainError> {
         Ok(Paginated {
             items: vec![],
             total: 0,
@@ -801,109 +743,6 @@ impl SearchPort for TestStore {
     }
 }
 
-#[async_trait]
-impl ActivityPubRepository for TestStore {
-    async fn outbox_entries_for_actor(
-        &self,
-        _uid: &UserId,
-    ) -> Result<Vec<crate::ports::OutboxEntry>, DomainError> {
-        Ok(vec![])
-    }
-    async fn outbox_page_for_actor(
-        &self,
-        _uid: &UserId,
-        _before: Option<chrono::DateTime<chrono::Utc>>,
-        _limit: usize,
-    ) -> Result<Vec<crate::ports::OutboxEntry>, DomainError> {
-        Ok(vec![])
-    }
-    async fn find_remote_actor_id(
-        &self,
-        actor_ap_url: &str,
-    ) -> Result<Option<UserId>, DomainError> {
-        Ok(self.actor_ap_ids.lock().unwrap().get(actor_ap_url).cloned())
-    }
-    async fn intern_remote_actor(&self, actor_ap_url: &str) -> Result<UserId, DomainError> {
-        if let Some(uid) = self.find_remote_actor_id(actor_ap_url).await? {
-            return Ok(uid);
-        }
-        let uid = UserId::new();
-        let handle = url::Url::parse(actor_ap_url)
-            .map(|u| u.path().trim_start_matches('/').replace('/', "_"))
-            .unwrap_or_else(|_| format!("remote_{}", &uid.to_string()[..8]));
-        let user = crate::models::user::User {
-            id: uid.clone(),
-            username: Username::from_trusted(handle.clone()),
-            email: Email::from_trusted(format!("{}@remote", uid)),
-            password_hash: PasswordHash("".into()),
-            display_name: None,
-            bio: None,
-            avatar_url: None,
-            header_url: None,
-            custom_css: None,
-            local: false,
-            created_at: chrono::Utc::now(),
-            updated_at: chrono::Utc::now(),
-        };
-        self.users.lock().unwrap().push(user);
-        self.actor_ap_ids
-            .lock()
-            .unwrap()
-            .insert(actor_ap_url.to_string(), uid.clone());
-        Ok(uid)
-    }
-    async fn update_remote_actor_display(
-        &self,
-        _user_id: &UserId,
-        _display_name: Option<&str>,
-        _avatar_url: Option<&str>,
-    ) -> Result<(), DomainError> {
-        Ok(())
-    }
-    async fn accept_note(
-        &self,
-        _ap_id: &str,
-        _author_id: &UserId,
-        _content: &str,
-        _published: chrono::DateTime<chrono::Utc>,
-        _sensitive: bool,
-        _content_warning: Option<String>,
-        _visibility: &str,
-        _in_reply_to: Option<&str>,
-    ) -> Result<(), DomainError> {
-        Ok(())
-    }
-    async fn apply_note_update(&self, _ap_id: &str, _new_content: &str) -> Result<(), DomainError> {
-        Ok(())
-    }
-    async fn retract_note(&self, _ap_id: &str) -> Result<(), DomainError> {
-        Ok(())
-    }
-    async fn retract_actor_notes(&self, _actor_ap_url: &str) -> Result<(), DomainError> {
-        Ok(())
-    }
-    async fn count_local_notes(&self) -> Result<u64, DomainError> {
-        Ok(self
-            .thoughts
-            .lock()
-            .unwrap()
-            .iter()
-            .filter(|t| t.local)
-            .count() as u64)
-    }
-    async fn get_thought_ap_id(
-        &self,
-        thought_id: &ThoughtId,
-    ) -> Result<Option<String>, DomainError> {
-        Ok(self.thought_ap_ids.lock().unwrap().get(thought_id).cloned())
-    }
-    async fn get_actor_ap_urls(
-        &self,
-        user_id: &UserId,
-    ) -> Result<Option<ActorApUrls>, DomainError> {
-        Ok(self.actor_ap_urls.lock().unwrap().get(user_id).cloned())
-    }
-}
 
 #[async_trait]
 impl FederationSchedulerPort for TestStore {
@@ -964,31 +803,6 @@ impl OutboxWriter for NoOpOutboxWriter {
     }
 }
 
-#[cfg(test)]
-mod ap_repo_tests {
-    use super::*;
-    use crate::value_objects::UserId;
-
-    #[tokio::test]
-    async fn test_store_outbox_returns_empty() {
-        let store = TestStore::default();
-        let result = store
-            .outbox_entries_for_actor(&UserId::new())
-            .await
-            .unwrap();
-        assert!(result.is_empty());
-    }
-
-    #[tokio::test]
-    async fn test_store_intern_creates_placeholder() {
-        let store = TestStore::default();
-        let url = "https://example.com/users/alice";
-        let id1 = store.intern_remote_actor(url).await.unwrap();
-        let id2 = store.intern_remote_actor(url).await.unwrap();
-        assert_eq!(id1, id2, "intern must be idempotent");
-    }
-}
-
 #[cfg(test)]
 mod federation_port_tests {
     use super::*;