How to combine ActiveRecord Relations? Merge not working? - ruby-on-rails

Question

Initially when I was trying to build a histogram of all Items that have an Order start between a given set of dates based on exactly what the item was (:name_id) and the frequency of that :name_id, I was using the following code:
dates = ["May 27, 2016", "May 30, 2016"]
items = Item.joins(:order).where("orders.start >= ?", dates.first).where("orders.start <= ?", dates.last)
histogram = {}
items.pluck(:name_id).uniq.each do |name_id|
histogram[name_id] = items.where(name_id:name_id).count
end
This code worked FINE.
Now, however, I'm trying to build a histogram that's more expansive. I still want to capture frequency of :name_id over a period of time, but now I want to bound that time by Order start and end. I'm having trouble however, combining the ActiveRecord Relations that follow the queries. Specifically, if my queries are as follows:
items_a = Item.joins(:order).where("orders.start >= ?", dates.first).where("orders.start <= ?", dates.last)
items_b = Item.joins(:order).where("orders.end >= ?", dates.first).where("orders.end <= ?", dates.last)
How do I join the 2 queries so that my code below that acts on query objects still works?
items.pluck(:name_id).each do |name_id|
histogram[name_id] = items.where(name_id:name_id).count
end
What I've tried:
+, but of course that doesn't work because it turns the result into an Array where methods like pluck don't work:
(items_a + items_b).pluck(:name_id)
=> error
merge, this is what all the SO answers seem to say... but it doesn't work for me because, as the docs say, merge figures out the intersection, so my result is like this:
items_a.count
=> 100
items_b.count
=> 30
items_a.merge(items_b)
=> 15
FYI currently, I've monkey-patched this with the below, but it's not very ideal. Thanks for the help!
name_ids = (items_a.pluck(:name_id) + items_b.pluck(:name_id)).uniq
name_ids.each do |name_id|
# from each query object, return the ids of the item objects that meet the name_id criterion
item_object_ids = items_a.where(name_id:name_id).pluck(:id) + items_b.where(name_id:name_id).pluck(:id) + items_c.where(name_id:name_id).pluck(:id)
# then check the item objects for duplicates and then count up. btw I realize that with the uniq here I'm SOMEWHAT doing an intersection of the objects, but it's nowhere near as severe... the above example where merge yielded a count of 15 is not that off from the truth, when the count should be maybe -5 from the addition of the 2 queries
histogram[name_id] = item_object_ids.uniq.count
end

Answer 1

You can combine your two queries into one:
items = Item.joins(:order).where(
"(orders.start >= ? AND orders.start <= ?) OR (orders.end >= ? AND orders.end <= ?)",
dates.first, dates.last, dates.first, dates.last
)
This might be a little more readable:
items = Item.joins(:order).where(
"(orders.start >= :first AND orders.start <= :last) OR (orders.end >= :first AND orders.end <= :last)",
{ first: dates.first, last: dates.last }
)
Rails 5 will support an or method that might make this a little nicer:
items_a = Item.joins(:order).where(
"orders.start >= :first AND orders.start <= :last",
{ first: dates.first, last: dates.last }
).or(
"orders.end >= :first AND orders.end <= :last",
{ first: dates.first, last: dates.last }
)
Or maybe not any nicer in this case

Answer 2

Maybe this will be a bit cleaner:
date_range = "May 27, 2016".to_date.."May 30, 2016".to_date
items = Item.joins(:order).where('orders.start' => date_range).or('orders.end' => date_range)