email_agent/internal/store/unsubscribe.go

// Unsubscribe tracking for Phase 6 (FR-7).
//
// Two things are persisted here:
//
//   - the raw unsubscribe links detected on a message (messages.list_unsubscribe,
//     stored by the mail fetch and parsed by internal/unsubscribe); and
//   - the domain-level ignore list and the record of what was marked, in the
//     `unsubscribe` table.
//
// The app never follows a link. Marking a domain records Sam's decision; the
// actual unsubscribe stays a manual, explicitly approved action (AUD-72).
package store

import (
	"context"
	"database/sql"
	"errors"
	"fmt"
	"strings"
	"time"
)

// Unsubscribe statuses. `ignored` is the ignore list; `done` records a domain
// Sam has decided to unsubscribe from.
const (
	UnsubscribeStatusDone    = "done"
	UnsubscribeStatusIgnored = "ignored"
)

// ErrUnsubscribeStatus is returned for an unknown unsubscribe status.
var ErrUnsubscribeStatus = errors.New("unsubscribe status must be done or ignored")

// UnsubscribeRecord is one row of the domain-level tracker.
type UnsubscribeRecord struct {
	ID        int64
	Domain    string
	Status    string
	URL       string
	CreatedAt string
	UpdatedAt string
}

// ListUnsubscribe returns every domain record, newest first.
func (s *Store) ListUnsubscribe(ctx context.Context) ([]UnsubscribeRecord, error) {
	rows, err := s.db.QueryContext(ctx, `
SELECT id, domain, status, COALESCE(url, ''), created_at, COALESCE(updated_at, '')
FROM unsubscribe ORDER BY COALESCE(updated_at, created_at) DESC, id DESC`)
	if err != nil {
		return nil, fmt.Errorf("list unsubscribe records: %w", err)
	}
	defer rows.Close()
	var out []UnsubscribeRecord
	for rows.Next() {
		var r UnsubscribeRecord
		if err := rows.Scan(&r.ID, &r.Domain, &r.Status, &r.URL, &r.CreatedAt, &r.UpdatedAt); err != nil {
			return nil, fmt.Errorf("scan unsubscribe record: %w", err)
		}
		out = append(out, r)
	}
	if err := rows.Err(); err != nil {
		return nil, fmt.Errorf("list unsubscribe records: %w", err)
	}
	return out, nil
}

// UnsubscribeByDomain loads one domain record. ok is false when the domain is
// not tracked.
func (s *Store) UnsubscribeByDomain(ctx context.Context, domain string) (UnsubscribeRecord, bool, error) {
	domain = strings.TrimSpace(domain)
	if domain == "" {
		return UnsubscribeRecord{}, false, nil
	}
	var r UnsubscribeRecord
	err := s.db.QueryRowContext(ctx, `
SELECT id, domain, status, COALESCE(url, ''), created_at, COALESCE(updated_at, '')
FROM unsubscribe WHERE domain = ?`, domain).
		Scan(&r.ID, &r.Domain, &r.Status, &r.URL, &r.CreatedAt, &r.UpdatedAt)
	if errors.Is(err, sql.ErrNoRows) {
		return UnsubscribeRecord{}, false, nil
	}
	if err != nil {
		return UnsubscribeRecord{}, false, fmt.Errorf("read unsubscribe record %q: %w", domain, err)
	}
	return r, true, nil
}

// SetUnsubscribeStatus upserts the domain record. An empty url keeps the stored
// one, so an "ignore" does not wipe a link recorded earlier.
func (s *Store) SetUnsubscribeStatus(ctx context.Context, domain, status, url string) error {
	domain = strings.TrimSpace(domain)
	if domain == "" {
		return errors.New("unsubscribe domain must not be empty")
	}
	if status != UnsubscribeStatusDone && status != UnsubscribeStatusIgnored {
		return ErrUnsubscribeStatus
	}
	const q = `
INSERT INTO unsubscribe (domain, status, url, updated_at)
VALUES (?, ?, ?, datetime('now'))
ON CONFLICT (domain) DO UPDATE SET
    status     = excluded.status,
    url        = CASE WHEN excluded.url <> '' THEN excluded.url ELSE unsubscribe.url END,
    updated_at = datetime('now')`
	if _, err := s.db.ExecContext(ctx, q, domain, status, strings.TrimSpace(url)); err != nil {
		return fmt.Errorf("set unsubscribe status for %q: %w", domain, err)
	}
	return nil
}

// ClearUnsubscribe removes a domain record, taking it off the ignore list.
func (s *Store) ClearUnsubscribe(ctx context.Context, domain string) error {
	if _, err := s.db.ExecContext(ctx,
		`DELETE FROM unsubscribe WHERE domain = ?`, strings.TrimSpace(domain)); err != nil {
		return fmt.Errorf("clear unsubscribe record %q: %w", domain, err)
	}
	return nil
}

// UnsubscribeCandidate is a message that may carry an unsubscribe link.
type UnsubscribeCandidate struct {
	MessageID           int64
	FromAddr            string
	Subject             string
	SourceAddress       string
	BodyText            string
	ListUnsubscribe     string
	ListUnsubscribePost string
	Category            string
	ReceivedAt          time.Time
}

// UnsubscribeCandidates returns messages with a List-Unsubscribe header or a
// body that mentions unsubscribing, newest first. The caller runs the detector
// over them, because body parsing lives in internal/unsubscribe.
func (s *Store) UnsubscribeCandidates(ctx context.Context, limit int) ([]UnsubscribeCandidate, error) {
	if limit < 1 {
		limit = 200
	}
	const q = `
SELECT m.id, COALESCE(m.from_addr, ''), COALESCE(m.subject, ''),
       COALESCE(m.source_address, ''), COALESCE(m.body_text, ''),
       COALESCE(m.list_unsubscribe, ''), COALESCE(m.list_unsubscribe_post, ''),
       COALESCE(c.category, ''), m.received_at
FROM messages m
LEFT JOIN classifications c ON c.id = (
    SELECT id FROM classifications WHERE message_id = m.id ORDER BY id DESC LIMIT 1
)
WHERE (m.list_unsubscribe IS NOT NULL AND m.list_unsubscribe <> '')
   OR instr(lower(COALESCE(m.body_text, '')), 'unsubscribe') > 0
ORDER BY COALESCE(NULLIF(m.date_sent, ''), m.received_at) DESC, m.id DESC
LIMIT ?`
	rows, err := s.db.QueryContext(ctx, q, limit)
	if err != nil {
		return nil, fmt.Errorf("list unsubscribe candidates: %w", err)
	}
	defer rows.Close()
	var out []UnsubscribeCandidate
	for rows.Next() {
		var (
			c        UnsubscribeCandidate
			received string
		)
		if err := rows.Scan(
			&c.MessageID, &c.FromAddr, &c.Subject, &c.SourceAddress,
			&c.BodyText, &c.ListUnsubscribe, &c.ListUnsubscribePost,
			&c.Category, &received,
		); err != nil {
			return nil, fmt.Errorf("scan unsubscribe candidate: %w", err)
		}
		c.ReceivedAt = parseSQLTime(received)
		out = append(out, c)
	}
	if err := rows.Err(); err != nil {
		return nil, fmt.Errorf("list unsubscribe candidates: %w", err)
	}
	return out, nil
}