multica/server/pkg/db/queries/comment.sql

-- name: ListCommentsForIssue :many
-- The NEWEST $3 comments for an issue, returned in chronological order.
--
-- Same shape and same reason as ListActivitiesForIssue: the inner query takes
-- the window with the keyset ordering so the cap discards the OLDEST rows, and
-- the outer query restores the ascending contract callers rely on. The ordering
-- of the inner window is satisfied by idx_comment_issue_keyset (migration 068);
-- the outer query sorts that bounded window back into chronological order.
--
-- A newest-N window is a suffix of the timeline, and unlike a prefix it is NOT
-- closed under "parent of": a reply is always newer than its parent, so an old
-- thread root can fall outside the window while a fresh reply to it stays
-- inside. Callers that render threads must close the parent chains afterwards —
-- see completeCommentThreads (MUL-5492).
--
-- The cap is still purely defensive here — issue p99 is ~30 comments and the max
-- ever observed in prod is ~1.1k — but "defensive" is not a reason to drop the
-- newest rows when it does fire (MUL-5492).
SELECT * FROM (
    SELECT * FROM comment
    WHERE issue_id = $1 AND workspace_id = $2
    ORDER BY created_at DESC, id DESC
    LIMIT $3
) AS recent
ORDER BY created_at ASC, id ASC;

-- name: ListCommentsByIDsForIssue :many
-- The subset of @ids that exists within this issue and workspace.
--
-- Used to walk parent chains one level at a time (see completeCommentThreads).
-- Deliberately NOT a recursive CTE: an earlier revision walked parent_id upward
-- in SQL, which had no depth bound — a deep chain could pull tens of thousands of
-- ancestors back and defeat the whole point of the row cap — and its recursive
-- branch matched on parent_id alone, so a stray cross-workspace parent reference
-- would have dragged another tenant's comments into the response. Both tenant
-- columns are required here on every level, and the caller owns the budget.
SELECT * FROM comment
WHERE id = ANY(@ids::uuid[])
  AND issue_id = @issue_id
  AND workspace_id = @workspace_id
ORDER BY created_at ASC, id ASC;

-- name: ListChildCommentsForParents :many
-- Fetch one descendant level for a bounded set of parent comments. Each parent
-- maps back to its thread root in the Go breadth-first walk, avoiding one query
-- per root.
--
-- Both tenant predicates apply at every level. row_limit is always a probe
-- limit owned by completeCommentThreads; it prevents a wide level from defeating
-- the response budget. through_at/through_id pin the walk to the newest row in
-- the original window, so replies created concurrently are left to realtime
-- delivery instead of making the multi-query snapshot internally inconsistent.
SELECT * FROM comment
WHERE parent_id = ANY(@parent_ids::uuid[])
  AND issue_id = @issue_id
  AND workspace_id = @workspace_id
  AND (created_at, id) <= (@through_at::timestamptz, @through_id::uuid)
ORDER BY parent_id ASC, created_at ASC, id ASC
LIMIT @row_limit;

-- name: ListCommentsSinceForIssue :many
-- Comments created strictly after $3 in chronological order, capped at $4.
-- Powers the CLI's `--since` agent-polling flow.
SELECT * FROM comment
WHERE issue_id = $1 AND workspace_id = $2 AND created_at > $3
ORDER BY created_at ASC, id ASC
LIMIT $4;

-- name: ListRootCommentsForIssue :many
-- Top-level comments only, in issue chronological order, each annotated with
-- per-thread orientation stats: reply_count (number of descendants) and
-- last_activity_at (MAX(created_at) over the whole subtree). This powers
-- `comment list --roots-only` so agents can not only orient around the global
-- discussion but also triage which thread to drill into (biggest / most
-- recently active) before fetching any specific reply thread.
--
-- `selected_roots` takes the newest @row_limit roots. The final SELECT restores
-- chronological order, so the defensive cap drops the oldest roots without
-- changing the wire order. The recursive `membership` walk only expands those
-- selected threads rather than every thread in the issue, and labels every
-- descendant with its root so the stats stay correct at any reply depth.
WITH RECURSIVE selected_roots AS (
    SELECT c.id, c.created_at
    FROM comment c
    WHERE c.issue_id = @issue_id
      AND c.workspace_id = @workspace_id
      AND c.parent_id IS NULL
    ORDER BY c.created_at DESC, c.id DESC
    LIMIT @row_limit
),
membership(id, root_id, comment_created_at) AS (
    SELECT sr.id, sr.id AS root_id, sr.created_at
    FROM selected_roots sr
    UNION ALL
    SELECT c.id, m.root_id, c.created_at
    FROM comment c
    JOIN membership m ON c.parent_id = m.id
    WHERE c.issue_id = @issue_id
      AND c.workspace_id = @workspace_id
),
thread_stats AS (
    SELECT root_id,
           (COUNT(*) - 1)::int AS reply_count,
           MAX(comment_created_at)::timestamptz AS last_activity_at
    FROM membership
    GROUP BY root_id
)
SELECT c.id, c.issue_id, c.author_type, c.author_id, c.content, c.type,
       c.created_at, c.updated_at, c.parent_id, c.workspace_id,
       c.resolved_at, c.resolved_by_type, c.resolved_by_id,
       c.source_task_id, c.quick_action_id, c.revision, c.deleted_at,
       ts.reply_count AS reply_count,
       ts.last_activity_at AS last_activity_at
FROM selected_roots sr
JOIN comment c ON c.id = sr.id
JOIN thread_stats ts ON ts.root_id = sr.id
ORDER BY c.created_at ASC, c.id ASC;

-- name: ListRootCommentsSinceForIssue :many
-- Top-level comments created strictly after @since, each annotated with the
-- same reply_count / last_activity_at stats as ListRootCommentsForIssue. The
-- @since filter narrows which roots are returned; the stats are still computed
-- over each selected thread's full subtree (so a freshly created root with no
-- replies reports reply_count 0 and last_activity_at = its own created_at).
-- selected_roots applies the @since + @row_limit cut up front so the recursive
-- membership walk only touches the subtrees of the roots we actually return.
WITH RECURSIVE selected_roots AS (
    SELECT c.id, c.created_at
    FROM comment c
    WHERE c.issue_id = @issue_id
      AND c.workspace_id = @workspace_id
      AND c.parent_id IS NULL
      AND c.created_at > @since
    ORDER BY c.created_at ASC, c.id ASC
    LIMIT @row_limit
),
membership(id, root_id, comment_created_at) AS (
    SELECT sr.id, sr.id AS root_id, sr.created_at
    FROM selected_roots sr
    UNION ALL
    SELECT c.id, m.root_id, c.created_at
    FROM comment c
    JOIN membership m ON c.parent_id = m.id
    WHERE c.issue_id = @issue_id
      AND c.workspace_id = @workspace_id
),
thread_stats AS (
    SELECT root_id,
           (COUNT(*) - 1)::int AS reply_count,
           MAX(comment_created_at)::timestamptz AS last_activity_at
    FROM membership
    GROUP BY root_id
)
SELECT c.id, c.issue_id, c.author_type, c.author_id, c.content, c.type,
       c.created_at, c.updated_at, c.parent_id, c.workspace_id,
       c.resolved_at, c.resolved_by_type, c.resolved_by_id,
       c.source_task_id, c.quick_action_id, c.revision, c.deleted_at,
       ts.reply_count AS reply_count,
       ts.last_activity_at AS last_activity_at
FROM selected_roots sr
JOIN comment c ON c.id = sr.id
JOIN thread_stats ts ON ts.root_id = sr.id
ORDER BY c.created_at ASC, c.id ASC;

-- name: ListThreadCommentsForIssuePaged :many
-- Resolves @anchor_id to its thread root, recursively expands every descendant,
-- and returns the root + only the @reply_limit most recent replies (per the
-- (created_at, id) composite key). When @has_cursor=TRUE only replies with
-- (created_at, id) < (@before_at, @before_id) are eligible — that is the
-- cursor for scrolling *within* a thread.
--
-- Root is unconditional: it is included regardless of @reply_limit (even 0)
-- and regardless of the cursor. A reader landing on a long thread needs the
-- root for the "what is this thread about" context, even if every reply has
-- been paginated past.
--
-- Reply selection happens DESC (newest replies first) so the cursor walks
-- toward older replies; the outer SELECT then re-sorts the combined output
-- ASC so the body stays chronological (oldest → newest), matching every
-- other comment list path.
WITH RECURSIVE root_of AS (
    SELECT c.id, c.parent_id
    FROM comment c
    WHERE c.id = @anchor_id AND c.issue_id = @issue_id AND c.workspace_id = @workspace_id
    UNION ALL
    SELECT p.id, p.parent_id
    FROM comment p
    JOIN root_of r ON p.id = r.parent_id
    WHERE p.issue_id = @issue_id AND p.workspace_id = @workspace_id
),
thread_root AS (
    SELECT id FROM root_of WHERE parent_id IS NULL LIMIT 1
),
descendants AS (
    SELECT c.id, c.issue_id, c.author_type, c.author_id, c.content, c.type,
           c.created_at, c.updated_at, c.parent_id, c.workspace_id,
           c.resolved_at, c.resolved_by_type, c.resolved_by_id,
           c.source_task_id, c.quick_action_id, c.revision, c.deleted_at
    FROM comment c
    JOIN thread_root tr ON c.id = tr.id
    UNION
    SELECT c.id, c.issue_id, c.author_type, c.author_id, c.content, c.type,
           c.created_at, c.updated_at, c.parent_id, c.workspace_id,
           c.resolved_at, c.resolved_by_type, c.resolved_by_id,
           c.source_task_id, c.quick_action_id, c.revision, c.deleted_at
    FROM comment c
    JOIN descendants d ON c.parent_id = d.id
    WHERE c.issue_id = @issue_id AND c.workspace_id = @workspace_id
),
reply_page AS (
    SELECT d.id, d.issue_id, d.author_type, d.author_id, d.content, d.type,
           d.created_at, d.updated_at, d.parent_id, d.workspace_id,
           d.resolved_at, d.resolved_by_type, d.resolved_by_id,
           d.source_task_id, d.quick_action_id, d.revision, d.deleted_at
    FROM descendants d
    WHERE d.id NOT IN (SELECT id FROM thread_root)
      AND (
          @has_cursor::boolean = FALSE
          OR (d.created_at, d.id) < (@before_at::timestamptz, @before_id::uuid)
      )
    ORDER BY d.created_at DESC, d.id DESC
    LIMIT @reply_limit
)
SELECT id, issue_id, author_type, author_id, content, type,
       created_at, updated_at, parent_id, workspace_id,
       resolved_at, resolved_by_type, resolved_by_id,
       source_task_id, quick_action_id, revision, deleted_at
FROM (
    SELECT d.id, d.issue_id, d.author_type, d.author_id, d.content, d.type,
           d.created_at, d.updated_at, d.parent_id, d.workspace_id,
           d.resolved_at, d.resolved_by_type, d.resolved_by_id,
           d.source_task_id, d.quick_action_id, d.revision, d.deleted_at
    FROM descendants d
    JOIN thread_root tr ON d.id = tr.id
    UNION ALL
    SELECT id, issue_id, author_type, author_id, content, type,
           created_at, updated_at, parent_id, workspace_id,
           resolved_at, resolved_by_type, resolved_by_id,
           source_task_id, quick_action_id, revision, deleted_at
    FROM reply_page
) combined
ORDER BY created_at ASC, id ASC;

-- name: ListRecentThreadCommentsForIssue :many
-- Returns the N most recently active threads (root + every descendant) rather
-- than the N most recent rows. A thread's "last activity" is MAX(created_at)
-- over its whole subtree; threads are ranked by (last_activity_at DESC,
-- root_id DESC) and the top N are expanded.
--
-- Why thread-grouped instead of row-recent: with row-recent the newest 20
-- comments can come from 8 different threads — the agent sees 8 unrelated
-- tails. With thread-grouped the agent sees N complete conversational arcs,
-- which matches how a human reads an issue (#2340).
--
-- Response ordering:
--   threads:     (thread_last_activity_at ASC, root_id ASC)
--   in-thread:   (created_at ASC, id ASC)
-- So the oldest-active thread appears first and the most recently-active
-- thread is at the tail, closest to "now" in an agent prompt.
--
-- Cursor scrolls back through threads. When @has_cursor=TRUE only threads
-- with (last_activity_at, root_id) < (@before_at, @before_id) are eligible.
-- The cursor is a THREAD cursor — both values identify a thread (its last
-- activity timestamp and its root comment id), not a single row.
--
-- The recursive `membership` CTE labels each comment with its thread root by
-- walking down from every root. It does not assume any maximum nesting depth,
-- which preserves correctness even if the schema ever allows reply-of-reply
-- (the agent path in TaskService.createAgentComment collapses to root today,
-- but the user-facing CreateComment handler does not enforce it).
WITH RECURSIVE membership(id, root_id, comment_created_at) AS (
    -- Each root maps to itself.
    SELECT c.id, c.id AS root_id, c.created_at
    FROM comment c
    WHERE c.issue_id = @issue_id
      AND c.workspace_id = @workspace_id
      AND c.parent_id IS NULL
    UNION ALL
    -- Each descendant inherits its parent's root_id.
    SELECT c.id, m.root_id, c.created_at
    FROM comment c
    JOIN membership m ON c.parent_id = m.id
    WHERE c.issue_id = @issue_id
      AND c.workspace_id = @workspace_id
),
thread_stats AS (
    SELECT root_id, MAX(comment_created_at)::timestamptz AS last_activity_at
    FROM membership
    GROUP BY root_id
),
picked AS (
    SELECT ts.root_id, ts.last_activity_at
    FROM thread_stats ts
    WHERE (
        @has_cursor::boolean = FALSE
        OR (ts.last_activity_at, ts.root_id) < (@before_at::timestamptz, @before_id::uuid)
    )
    ORDER BY ts.last_activity_at DESC, ts.root_id DESC
    LIMIT @thread_limit
)
SELECT c.id, c.issue_id, c.author_type, c.author_id, c.content, c.type,
       c.created_at, c.updated_at, c.parent_id, c.workspace_id,
       c.resolved_at, c.resolved_by_type, c.resolved_by_id,
       c.source_task_id, c.quick_action_id, c.revision, c.deleted_at,
       p.root_id AS thread_root_id,
       p.last_activity_at AS thread_last_activity_at
FROM picked p
JOIN membership m ON m.root_id = p.root_id
JOIN comment c ON c.id = m.id
ORDER BY p.last_activity_at ASC, p.root_id ASC, c.created_at ASC, c.id ASC;

-- name: CountComments :one
SELECT count(*) FROM comment
WHERE issue_id = $1 AND workspace_id = $2;

-- name: CountNewCommentsSince :one
-- Counts comments on an issue created strictly after @since, ACROSS THE WHOLE
-- ISSUE (every thread, not just the triggering one). Excludes the triggering
-- comment itself (@anchor_id — its body is already injected into the prompt)
-- and any authored by the given agent (@author_id), so a chatty agent does not
-- inflate its own new-comment count. The agent is steered to read the
-- triggering thread first (see BuildNewCommentsHint), but the count is
-- issue-wide so it knows the full catch-up volume. Feeds the daemon claim
-- response without shipping comment bodies.
SELECT count(*) FROM comment
WHERE issue_id = @issue_id
  AND workspace_id = @workspace_id
  AND created_at > @since
  AND id <> @anchor_id
  AND deleted_at IS NULL
  AND NOT (author_type = 'agent' AND author_id = @author_id);

-- name: GetLatestMemberCommentForIssueSince :one
-- MUL-4195 completion reconciliation: the newest MEMBER-authored comment on an
-- issue created strictly after @since (a run's started_at). Used when a task
-- completes to detect deliberate user input that landed while the agent was
-- busy — or that was merged into the running task after its context was
-- already built — so a single follow-up run can be scheduled for it. Restricted
-- to author_type = 'member' on purpose: only human input earns the guaranteed
-- follow-up, which preserves the existing anti-loop guarantees (agent replies,
-- acknowledgements, and self-triggers never qualify). Returns pgx.ErrNoRows
-- when nothing newer exists, i.e. the run already covered the latest input.
SELECT * FROM comment
WHERE issue_id = @issue_id
  AND author_type = 'member'
  AND created_at > @since
  AND deleted_at IS NULL
ORDER BY created_at DESC
LIMIT 1;

-- name: ListReconcilableCommentsForIssueSince :many
-- MUL-4195 / MUL-4304 completion reconciliation: every MEMBER- or AGENT-authored
-- comment on an issue created strictly after @since (the completing run's
-- created_at anchor), plus every id in its planned trigger/coalesced batch.
-- The one platform-authored exception is a delegated-failure recovery signal:
-- author_type=system, type=progress_update, source_task_id set. Such a signal
-- can be registered after a coordinator was claimed and must be replayed when
-- that run completes; the handler routes it through the dedicated recovery
-- path instead of generic comment/mention routing.
-- Planned ids matter for retry children because their input comments predate
-- the child's created_at; if one could not be embedded at claim time it still
-- needs reconciliation. The handler excludes only delivered_comment_ids, then
-- replays the remainder through the normal trigger pipeline oldest first.
--
-- Author-type scope (MUL-4304): originally restricted to author_type = 'member'.
-- That left a gap — an explicit agent→agent @mention (agent A comments
-- `@agent B`) that landed while B already had a DISPATCHED task was dropped by
-- the create-time enqueue path (merge only folds into a QUEUED task, so a
-- dispatched target hits the merge-miss + active-task continue) and then never
-- compensated here, because agent-authored comments were excluded. We now also
-- return 'agent' comments so those explicit mentions can be replayed.
--
-- The handler rechecks current routing and permissions. For agent authors it
-- accepts explicit mentions, plus worker-to-assigned-leader replies already
-- recorded in this run's planned inputs. Timestamp-only implicit agent replies
-- remain excluded. Replays are scoped to the completing agent, never a fan-out.
-- Ordered ASC so later comments coalesce onto the follow-up created by the first.
SELECT c.* FROM comment c
WHERE c.issue_id = @issue_id
  AND (c.id = ANY(@planned_comment_ids::uuid[])
       OR comment_thread_root_id(c.id) = sqlc.narg('comment_thread_id')::uuid)
  -- A deleted comment is no longer input, even when replies keep its row.
  AND c.deleted_at IS NULL
  -- A comment that steered this agent's running turn belongs to that turn,
  -- regardless of delivery status: failed delivery must not become an automatic
  -- new run. The same comment still reconciles normally for any other agent
  -- it addressed without steering.
  AND NOT EXISTS (
      SELECT 1 FROM task_supplement s
      JOIN agent_task_queue bound ON bound.id = s.task_id
      WHERE s.comment_id = c.id AND s.workspace_id = c.workspace_id
        AND bound.agent_id = @agent_id
  )
  -- An explicit member/agent cancellation acknowledges the planned input for
  -- this agent only for completing tasks already in flight at cancellation;
  -- automatic cancellations deliberately remain replayable.
  AND NOT EXISTS (
      SELECT 1
      FROM agent_task_queue cancelled
      WHERE cancelled.issue_id = c.issue_id
        AND cancelled.agent_id = @agent_id
        AND cancelled.status = 'cancelled'
        AND cancelled.cancelled_by_type IN ('member', 'agent')
        AND cancelled.cancelled_by_id IS NOT NULL
        AND cancelled.completed_at IS NOT NULL
        AND cancelled.completed_at >= @since
        AND c.updated_at <= cancelled.completed_at
        AND (
            cancelled.trigger_comment_id = c.id
            OR c.id = ANY(cancelled.coalesced_comment_ids)
        )
  )
  -- Nor does a comment whose author chose not to start this agent for it.
  AND NOT COALESCE(@agent_id::uuid = ANY(c.suppressed_agent_ids), false)
  AND (
      (
          c.author_type IN ('member', 'agent')
          AND (c.created_at > @since OR c.id = ANY(@planned_comment_ids::uuid[]))
      )
      OR (
          c.author_type = 'system'
          AND c.type = 'progress_update'
          AND c.source_task_id IS NOT NULL
          AND c.id = ANY(@planned_comment_ids::uuid[])
      )
  )
ORDER BY c.created_at ASC, c.id ASC;

-- name: GetComment :one
SELECT * FROM comment
WHERE id = $1;

-- name: GetCommentInWorkspace :one
SELECT * FROM comment
WHERE id = $1 AND workspace_id = $2;

-- name: GetThreadRoot :one
-- Returns the thread-root comment for @comment_id by walking parent_id up to
-- the row whose parent_id IS NULL. For a root comment it returns that comment
-- itself. Used when callers need thread-level behavior while parent_id remains
-- the exact direct parent of a reply. Cycle-safe under the PK constraint (a
-- comment cannot be its own ancestor).
WITH RECURSIVE root_of AS (
    SELECT c.id, c.parent_id
    FROM comment c
    WHERE c.id = @comment_id AND c.workspace_id = @workspace_id
    UNION ALL
    SELECT p.id, p.parent_id
    FROM comment p
    JOIN root_of r ON p.id = r.parent_id
)
SELECT c.* FROM comment c
WHERE c.id = (SELECT id FROM root_of WHERE parent_id IS NULL LIMIT 1);

-- name: CreateComment :one
-- A new comment counts as activity on its issue, so the same statement bumps
-- the parent issue's updated_at and last_activity_at. The touch is a leading data-modifying CTE and
-- the INSERT selects the issue/workspace back out of it, which makes the two
-- inseparable and gives two query-level guarantees:
--   * atomicity — the insert and the timestamp bump commit or roll back
--     together, so an issue is never left with a stale updated_at after a
--     comment persists; and
--   * tenant integrity — the comment can only be created against an issue that
--     actually exists in the given workspace. A mismatched (issue, workspace)
--     pair matches 0 rows in the CTE, the dependent INSERT then selects nothing,
--     and the :one query returns pgx.ErrNoRows. A wrong workspace can therefore
--     never leave a mis-attributed comment or a silently un-touched issue.
-- Centralizing this here means every comment entrypoint inherits both
-- guarantees regardless of what a caller passes. The "Updated date" sort and
-- the daemon GC TTL both read updated_at, so this consistency is load-bearing.
WITH touched_issue AS (
    UPDATE issue SET
        updated_at = now(),
        revision = revision + 1,
        last_activity_at = GREATEST(COALESCE(last_activity_at, updated_at), now())
    WHERE issue.id = sqlc.arg(issue_id) AND issue.workspace_id = sqlc.arg(workspace_id)
    RETURNING issue.id, issue.workspace_id, issue.revision
), inserted_comment AS (
    INSERT INTO comment (issue_id, workspace_id, author_type, author_id, content, type, parent_id, source_task_id, quick_action_id, via_plugin_id, suppressed_agent_ids, id)
    SELECT ti.id, ti.workspace_id, sqlc.arg(author_type), sqlc.arg(author_id), sqlc.arg(content), sqlc.arg(type), sqlc.narg(parent_id), sqlc.narg(source_task_id), sqlc.narg(quick_action_id), sqlc.narg(via_plugin_id), sqlc.narg(suppressed_agent_ids)::uuid[], COALESCE(sqlc.narg('id')::uuid, gen_random_uuid())
    FROM touched_issue ti
    RETURNING *
)
SELECT inserted_comment.*, touched_issue.revision AS issue_revision
FROM inserted_comment
JOIN touched_issue ON touched_issue.id = inserted_comment.issue_id;

-- name: GetDelegatedFailureRecoveryComment :one
-- The failed task row is locked by the caller before this lookup/insert pair,
-- making (source issue, failed task) a durable idempotency key without a new
-- hot-table index or schema migration. Platform recovery signals are the only
-- system-authored progress updates that carry source_task_id.
SELECT * FROM comment
WHERE issue_id = @issue_id
  AND workspace_id = @workspace_id
  AND author_type = 'system'
  AND type = 'progress_update'
  AND source_task_id = @source_task_id
ORDER BY created_at ASC, id ASC
LIMIT 1;

-- name: GetDelegatedFailureRecoveryExhaustionComment :one
-- The failed task and newest recovery-attempt row are locked by the caller
-- before this lookup/insert pair. Keeping exhaustion as a separate system
-- comment preserves the original recovery signal while making the automatic
-- stop visible in the issue timeline.
SELECT * FROM comment
WHERE issue_id = @issue_id
  AND workspace_id = @workspace_id
  AND author_type = 'system'
  AND type = 'system'
  AND source_task_id = @source_task_id
ORDER BY created_at ASC, id ASC
LIMIT 1;

-- name: UpdateComment :one
WITH locked_issue AS MATERIALIZED (
    -- Keep the global issue -> child lock order used by issue teardown. The
    -- aggregate below still yields one row when the parent was concurrently
    -- deleted, preserving best-effort edits of an orphaned comment.
    SELECT issue.id
    FROM issue
    JOIN comment ON comment.issue_id = issue.id
                AND comment.workspace_id = issue.workspace_id
    WHERE comment.id = $1
    FOR UPDATE OF issue
), issue_fence AS MATERIALIZED (
    -- The aggregate always emits one row. Consuming locked_count in target's
    -- tautological predicate creates a real data dependency: locked_issue must
    -- acquire the owner lock before target can lock the comment. MATERIALIZED
    -- prevents folding/re-evaluation; it does not itself establish lock order.
    SELECT count(*) AS locked_count FROM locked_issue
), target AS MATERIALIZED (
    SELECT comment.*,
           ROW(comment.content, comment.source_task_id) IS DISTINCT FROM
               ROW($2, sqlc.narg(source_task_id)::uuid) AS did_change
    FROM comment
    CROSS JOIN issue_fence
    WHERE comment.id = $1
      AND issue_fence.locked_count >= 0
      -- A deleted comment's tombstone is not editable.
      AND comment.deleted_at IS NULL
      AND (sqlc.narg('expected_revision')::bigint IS NULL OR revision = sqlc.narg('expected_revision')::bigint)
      AND (
        sqlc.narg('content_base')::text IS NULL
        OR content IS NOT DISTINCT FROM sqlc.narg('content_base')::text
        OR content IS NOT DISTINCT FROM $2
      )
    FOR UPDATE OF comment
), updated_comment AS (
    UPDATE comment SET
        content = $2,
        source_task_id = sqlc.narg(source_task_id)::uuid,
        -- New text re-decides which agents it must not start, under the same
        -- revision check as the text itself.
        suppressed_agent_ids = CASE WHEN target.content IS DISTINCT FROM $2
                                    THEN sqlc.narg(suppressed_agent_ids)::uuid[]
                                    ELSE comment.suppressed_agent_ids END,
        revision = comment.revision + CASE WHEN target.did_change THEN 1 ELSE 0 END,
        updated_at = CASE WHEN target.did_change THEN now() ELSE comment.updated_at END
    FROM target
    WHERE comment.id = target.id
    RETURNING comment.id, comment.issue_id, comment.author_type, comment.author_id,
              comment.content, comment.type, comment.created_at, comment.updated_at,
              comment.parent_id, comment.workspace_id, comment.resolved_at,
              comment.resolved_by_type, comment.resolved_by_id, comment.source_task_id,
              comment.quick_action_id, comment.via_plugin_id, comment.revision,
              comment.deleted_at, target.did_change
), touched_issue AS (
    UPDATE issue
    SET revision = issue.revision + 1,
        last_activity_at = GREATEST(COALESCE(issue.last_activity_at, issue.updated_at), now())
    FROM updated_comment
    WHERE updated_comment.did_change
      AND issue.id = updated_comment.issue_id
      AND issue.workspace_id = updated_comment.workspace_id
    RETURNING issue.id, issue.revision
)
SELECT updated_comment.id, updated_comment.issue_id, updated_comment.author_type,
       updated_comment.author_id, updated_comment.content, updated_comment.type,
       updated_comment.created_at, updated_comment.updated_at, updated_comment.parent_id,
       updated_comment.workspace_id, updated_comment.resolved_at,
       updated_comment.resolved_by_type, updated_comment.resolved_by_id,
       updated_comment.source_task_id, updated_comment.quick_action_id,
       updated_comment.via_plugin_id, updated_comment.revision,
       updated_comment.deleted_at,
       COALESCE((SELECT revision FROM touched_issue), 0)::bigint AS issue_revision
FROM updated_comment;

-- name: BumpCommentRevision :one
UPDATE comment
SET revision = revision + 1,
    updated_at = now()
WHERE id = @id
  AND workspace_id = @workspace_id
RETURNING *;

-- name: HasAgentCommentedSince :one
SELECT EXISTS (
    SELECT 1 FROM comment
    WHERE issue_id = @issue_id
      AND author_type = 'agent'
      AND author_id = @author_id
      AND created_at >= @since
      AND deleted_at IS NULL
) AS commented;

-- name: HasAgentRepliedInThread :one
-- Returns true if the given agent has posted a reply in the thread rooted at
-- the specified parent comment. Used to detect agent participation in a
-- member-started thread so that follow-up member replies still trigger the agent.
SELECT count(*) > 0 AS has_replied FROM comment
WHERE parent_id = @parent_id AND author_type = 'agent' AND author_id = @agent_id;

-- name: LockCommentForDelete :one
-- First statement of the comment delete transaction (#8296). Defense-in-depth:
-- workspace_id is a SQL-layer tenant guard. See DeleteIssue.
--
-- It takes the aggregate owner's lock before the comment's, so deleting cannot
-- deadlock with issue teardown (which takes the same issue -> comment order).
-- The issue lock is also what freezes the thread: every comment insert goes
-- through CreateComment, whose first step updates this issue row, so no reply
-- can be added to the issue until the delete transaction ends. The caller's
-- later statements start from a fresh snapshot taken after the locks, which is
-- what makes their reply check exact.
--
-- A tombstone is excluded: deleting an already-deleted comment finds nothing.
WITH locked_issue AS MATERIALIZED (
    SELECT issue.id
    FROM issue
    JOIN comment ON comment.issue_id = issue.id
                AND comment.workspace_id = issue.workspace_id
    WHERE comment.id = @id AND comment.workspace_id = @workspace_id
    FOR UPDATE OF issue
), issue_fence AS MATERIALIZED (
    -- The consumed locked_count below is the ordering fence: the issue lock is
    -- acquired before the comment lock. MATERIALIZED only prevents
    -- folding/re-evaluation and is not, by itself, a lock-order guarantee.
    SELECT count(*) AS locked_count FROM locked_issue
)
SELECT comment.*
FROM comment
CROSS JOIN issue_fence
WHERE comment.id = @id AND comment.workspace_id = @workspace_id
  AND comment.deleted_at IS NULL
  AND issue_fence.locked_count >= 0
FOR UPDATE OF comment;

-- name: LockLiveComment :one
-- Locks a live comment ahead of a write to one of its children — reactions and
-- attachments — in the order every comment mutation shares: issue, then
-- comment, then child. A comment deleted before or while this waits reads as
-- absent (the row lock re-checks deleted_at), so no child can land on its
-- tombstone. FOR NO KEY UPDATE conflicts with the delete transaction's FOR
-- UPDATE but not with a reply's FOR KEY SHARE, so it never blocks replies.
WITH locked_issue AS MATERIALIZED (
    SELECT issue.id
    FROM issue
    JOIN comment ON comment.issue_id = issue.id
                AND comment.workspace_id = issue.workspace_id
    WHERE comment.id = @id AND comment.workspace_id = @workspace_id
    FOR NO KEY UPDATE OF issue
), issue_fence AS MATERIALIZED (
    SELECT count(*) AS locked_count FROM locked_issue
)
SELECT comment.*
FROM comment
CROSS JOIN issue_fence
WHERE comment.id = @id AND comment.workspace_id = @workspace_id
  AND comment.deleted_at IS NULL
  AND issue_fence.locked_count >= 0
FOR NO KEY UPDATE OF comment;

-- name: CommentHasReplies :one
SELECT EXISTS (
    SELECT 1 FROM comment
    WHERE parent_id = @id AND workspace_id = @workspace_id
) AS has_replies;

-- name: TombstoneComment :one
-- Deletes a comment that still has replies by clearing it in place. The row,
-- its id and its parent_id stay so every reply keeps its direct parent; the
-- body goes, and so does any resolution, since a deleted comment cannot be a
-- thread's conclusion. The caller removes attachments and reactions in the
-- same transaction.
UPDATE comment SET
    content = '',
    deleted_at = now(),
    resolved_at = NULL,
    resolved_by_type = NULL,
    resolved_by_id = NULL,
    revision = revision + 1,
    updated_at = now()
WHERE id = @id AND workspace_id = @workspace_id
  AND deleted_at IS NULL
RETURNING *;

-- name: DeleteLeafComment :one
-- Removes a comment that has no replies. The NOT EXISTS guard keeps this
-- statement from ever reaching the legacy parent_id cascade: a comment with a
-- reply is tombstoned instead, never deleted with its replies.
DELETE FROM comment
WHERE comment.id = @id AND comment.workspace_id = @workspace_id
  AND NOT EXISTS (SELECT 1 FROM comment child WHERE child.parent_id = comment.id)
RETURNING comment.id, comment.parent_id;

-- name: DeleteReplylessCommentTombstone :one
-- Removes a tombstone whose last reply is gone. The delete transaction calls
-- this for each ancestor in turn, so a chain of placeholders never outlives
-- the replies it kept attached.
DELETE FROM comment
WHERE comment.id = @id AND comment.workspace_id = @workspace_id
  AND comment.deleted_at IS NOT NULL
  AND NOT EXISTS (SELECT 1 FROM comment child WHERE child.parent_id = comment.id)
RETURNING comment.id, comment.parent_id;

-- name: TouchIssueForCommentDelete :one
-- A delete counts as activity on its issue, like CreateComment and
-- UpdateComment. Runs inside the delete transaction after the comment rows
-- changed, so a delete that lost its race never touches the issue.
UPDATE issue
SET revision = revision + 1,
    last_activity_at = GREATEST(COALESCE(last_activity_at, updated_at), now())
WHERE id = @issue_id AND workspace_id = @workspace_id
RETURNING revision;

-- name: ResolveComment :one
-- Idempotent: re-resolving keeps the original resolved_at + resolver. Always
-- returns the row so the handler can surface the canonical state — unless the
-- comment has been deleted, whose tombstone cannot be a resolution.
UPDATE comment SET
    resolved_at = COALESCE(resolved_at, now()),
    resolved_by_type = COALESCE(resolved_by_type, $2),
    resolved_by_id = COALESCE(resolved_by_id, $3),
    revision = revision + CASE WHEN resolved_at IS NULL THEN 1 ELSE 0 END,
    updated_at = CASE WHEN resolved_at IS NULL THEN now() ELSE updated_at END
WHERE id = $1
  AND deleted_at IS NULL
RETURNING *;

-- name: ClearOtherThreadResolutions :many
-- Single-resolution invariant: a thread has at most one resolved comment.
-- Resolving @target_id makes it the sole resolution, so this clears resolved_at
-- on every OTHER currently-resolved comment in the same thread (the root of
-- @target_id plus every descendant). The handler runs this in the SAME tx as
-- ResolveComment so the replace is atomic — a crash can never leave two
-- resolutions or zero. Scope is the thread only (id IN descendants AND
-- id <> @target_id), never the whole issue. Returns each cleared row so the
-- handler can emit a comment:unresolved event per row; granular realtime
-- consumers replace a single comment in place and would otherwise keep
-- displaying the stale resolution.
WITH RECURSIVE root_of AS (
    -- Walk up from the target to its thread root.
    SELECT c.id, c.parent_id
    FROM comment c
    WHERE c.id = @target_id AND c.issue_id = @issue_id AND c.workspace_id = @workspace_id
    UNION ALL
    SELECT p.id, p.parent_id
    FROM comment p
    JOIN root_of r ON p.id = r.parent_id
),
thread_root AS (
    SELECT id FROM root_of WHERE parent_id IS NULL LIMIT 1
),
descendants AS (
    -- Expand back down from the root over the whole subtree. Cycle-safe under
    -- the PK constraint (a comment cannot be its own ancestor).
    SELECT c.id
    FROM comment c
    JOIN thread_root tr ON c.id = tr.id
    UNION
    SELECT c.id
    FROM comment c
    JOIN descendants d ON c.parent_id = d.id
    WHERE c.issue_id = @issue_id AND c.workspace_id = @workspace_id
)
UPDATE comment SET
    resolved_at = NULL,
    resolved_by_type = NULL,
    resolved_by_id = NULL,
    revision = revision + 1,
    updated_at = now()
WHERE comment.id IN (SELECT id FROM descendants)
  AND comment.id <> @target_id
  AND comment.resolved_at IS NOT NULL
RETURNING *;

-- name: UnresolveComment :one
-- Idempotent: a no-op clear (already unresolved) just returns the row.
UPDATE comment SET
    resolved_at = NULL,
    resolved_by_type = NULL,
    resolved_by_id = NULL,
    revision = revision + CASE WHEN resolved_at IS NOT NULL THEN 1 ELSE 0 END,
    updated_at = CASE WHEN resolved_at IS NOT NULL THEN now() ELSE updated_at END
WHERE id = $1
RETURNING *;
-- name: ListCommentAncestorPath :many
WITH RECURSIVE ancestor_path AS (
  SELECT c.*, ARRAY[c.id]::uuid[] AS visited_ids, 1::integer AS depth, false AS cycle
  FROM comment c
  WHERE c.id = sqlc.arg(comment_id)
    AND c.workspace_id = sqlc.arg(workspace_id)
    AND c.issue_id = sqlc.arg(issue_id)

  UNION ALL

  SELECT parent.*,
         path.visited_ids || parent.id,
         path.depth + 1,
         parent.id = ANY(path.visited_ids)
  FROM ancestor_path path
  JOIN comment parent ON parent.id = path.parent_id
  WHERE parent.workspace_id = sqlc.arg(workspace_id)
    AND parent.issue_id = sqlc.arg(issue_id)
    AND path.depth <= 256
    AND NOT path.cycle
)
SELECT id, issue_id, author_type, author_id, content, type, created_at,
       updated_at, workspace_id, parent_id, resolved_at, resolved_by_type,
       resolved_by_id, source_task_id, revision, quick_action_id, via_plugin_id,
       deleted_at, depth, cycle
FROM ancestor_path
ORDER BY depth DESC;

-- name: ListCommentThreadHistory :many
-- Capture the anchor comment's complete chronological thread through that
-- comment. Later replies are outside the immutable context boundary.
-- UUID is the stable tiebreaker used by the issue timeline when timestamps tie.
WITH RECURSIVE thread_history AS (
  SELECT root.*
  FROM comment root
  WHERE root.id = sqlc.arg(root_id)
    AND root.workspace_id = sqlc.arg(workspace_id)
    AND root.issue_id = sqlc.arg(issue_id)
    AND root.parent_id IS NULL
    AND (root.created_at, root.id) <= (
      sqlc.arg(anchor_created_at)::timestamptz,
      sqlc.arg(anchor_id)::uuid
    )

  UNION ALL

  SELECT child.*
  FROM comment child
  JOIN thread_history parent ON child.parent_id = parent.id
  WHERE child.workspace_id = sqlc.arg(workspace_id)
    AND child.issue_id = sqlc.arg(issue_id)
    AND (child.created_at, child.id) <= (
      sqlc.arg(anchor_created_at)::timestamptz,
      sqlc.arg(anchor_id)::uuid
    )
)
SELECT *
FROM thread_history
ORDER BY created_at, id
LIMIT sqlc.arg(row_limit);

-- name: LockCommentAncestorPath :many
WITH RECURSIVE ancestor_ids AS (
  SELECT c.id, c.parent_id, 1::integer AS depth, ARRAY[c.id]::uuid[] AS visited_ids
  FROM comment c
  WHERE c.id = sqlc.arg(comment_id)
    AND c.workspace_id = sqlc.arg(workspace_id)
    AND c.issue_id = sqlc.arg(issue_id)

  UNION ALL

  SELECT parent.id, parent.parent_id, path.depth + 1, path.visited_ids || parent.id
  FROM ancestor_ids path
  JOIN comment parent ON parent.id = path.parent_id
  WHERE parent.workspace_id = sqlc.arg(workspace_id)
    AND parent.issue_id = sqlc.arg(issue_id)
    AND path.depth <= 256
    AND NOT parent.id = ANY(path.visited_ids)
)
SELECT c.id
FROM comment c
JOIN ancestor_ids path ON path.id = c.id
ORDER BY c.id
FOR UPDATE OF c;