PR Review

Bug: when Nil branch is unreachable — all-nil column causes bad_request error

File: src/lavinmq/http/controller.cr:79-82

sample_value = sorted_items.each do |item|
  v = dig(item, sort_by)
  break v unless v.nil?
end

In Crystal, Array#each returns self (the array) when the block completes without break. So when all values are nil, sample_value is assigned the array itself, not nil. This means the when Nil branch on line 90 is dead code — instead, the else branch fires and returns a spurious bad_request error like "Can't sort on type Array(...)".

Fix: Use a separate variable so the nil case is handled correctly:

sample_value = nil
sorted_items.each do |item|
  v = dig(item, sort_by)
  if !v.nil?
    sample_value = v
    break
  end
end

Or more idiomatically, something like:

sample_value = sorted_items.find { |item| !dig(item, sort_by).nil? }.try { |item| dig(item, sort_by) }

The spec doesn't cover the all-nil case (it always creates at least one exchange with a policy), so this bug would not be caught by the current test.