Model & Job Fixes

Tasks

Task 1: Add missing scopes to ScrapeLog and ImportSession (M1/M3, M8)

Files: app/models/source_monitor/scrape_log.rb, app/models/source_monitor/import_session.rb, test/models/source_monitor/scrape_log_test.rb, test/models/source_monitor/import_session_test.rb Action: In ScrapeLog, add three scopes after the existing belongs_to declarations:

scope :by_source, ->(source) { where(source: source) }
scope :by_status, ->(success) { where(success: success) }
scope :by_item, ->(item) { where(item: item) }

In ImportSession, add a scope:

scope :in_step, ->(step) { where(current_step: step) }

Tests:

• ScrapeLog: test by_source returns only logs for given source, by_status(true) returns successful logs, by_item returns logs for given item.
• ImportSession: test in_step("upload") returns sessions in that step. Acceptance: All new scopes return correct results. Existing tests pass.

Task 2: Fix hardcoded table name and update_columns in Source (M5, M13)

Files: app/models/source_monitor/source.rb, test/models/source_monitor/source_test.rb Action: In avg_word_count, replace the hardcoded sourcemon_item_contents with ItemContent.table_name:

def avg_word_count
  items.joins(:item_content)
       .where.not(ItemContent.table_name => { scraped_word_count: nil })
       .average("#{ItemContent.table_name}.scraped_word_count")
       &.round
end

In reset_items_counter!, replace update_columns with update! so callbacks and validations run:

def reset_items_counter!
  actual_count = items.count
  update!(items_count: actual_count)
end

Tests:

• Test that avg_word_count returns correct average (may already be tested).
• Test that reset_items_counter! updates the count and runs validations (triggers update! not update_columns). Acceptance: No hardcoded table names in Source model. reset_items_counter! uses update!. All tests pass.

Task 3: Remove duplicate recent scope from LogEntry (M12)

Files: app/models/source_monitor/log_entry.rb, test/models/source_monitor/log_entry_test.rb Action: Remove line 13:

scope :recent, -> { order(started_at: :desc) }

LogEntry already includes this scope via Loggable concern... wait — checking the code, LogEntry does NOT include Loggable. It defines its own recent scope. However, LogEntry is a delegated_type that wraps FetchLog/ScrapeLog/HealthCheckLog which DO include Loggable. The recent scope on LogEntry is independent and intentional (it orders LogEntry records by started_at).

Actually, re-reading: LogEntry does NOT include Loggable. The scope is not a duplicate — it's the only recent scope on LogEntry. The scout finding may be incorrect. Skip this change — the scope is needed and is not a duplicate.

Update: Replace this task with: no action needed on M12. The recent scope on LogEntry is its own scope, not inherited.

Task 4: Add deadlock rescue and cleanup to jobs (S1, S9, S10)

Files: app/jobs/source_monitor/import_session_health_check_job.rb, app/jobs/source_monitor/source_health_check_job.rb, app/jobs/source_monitor/schedule_fetches_job.rb, test/jobs/source_monitor/import_session_health_check_job_test.rb, test/jobs/source_monitor/source_health_check_job_test.rb, test/jobs/source_monitor/schedule_fetches_job_test.rb Action:

S1 — ImportSessionHealthCheckJob: Add rescue_from ActiveRecord::Deadlocked after the existing discard_on line:

rescue_from ActiveRecord::Deadlocked do |error|
  Rails.logger&.warn("[SourceMonitor::ImportSessionHealthCheckJob] Deadlock: #{error.message}")
  retry_job(wait: 2.seconds + rand(3).seconds)
end

S9 — SourceHealthCheckJob: Remove the explicit result return value from the happy path in perform. Change:

result = SourceMonitor::Health::SourceHealthCheck.new(source: source).call
broadcast_outcome(source, result)
trigger_fetch_if_degraded(source, result)
result

to:

result = SourceMonitor::Health::SourceHealthCheck.new(source: source).call
broadcast_outcome(source, result)
trigger_fetch_if_degraded(source, result)

(Remove result on the last line, and remove nil from the rescue block.)

S10 — ScheduleFetchesJob: Add error rescue with logging:

def perform(options = nil)
  limit = extract_limit(options)
  SourceMonitor::Scheduler.run(limit:)
rescue StandardError => error
  Rails.logger&.error("[SourceMonitor::ScheduleFetchesJob] #{error.class}: #{error.message}")
  raise
end

The raise ensures the job still fails (for retry/visibility) but now has logging.

Tests:

• S1: Test that deadlocked error triggers retry (mock ActiveRecord::Deadlocked).
• S9: Verify perform doesn't return the result object (implementation detail, low priority).
• S10: Test that errors are logged and re-raised. Acceptance: All three jobs have improved error handling. All tests pass.