How to reduce eager_load query overhead (ActiveRecord)? - ruby-on-rails

Question

I have the following model:
# address.rb
class Address < ActiveRecord::Base
has_many :orders_for_cleaning, class_name: 'Order',
foreign_key: :cleaning_address_id
has_many :orders_for_billing, class_name: 'Order',
foreign_key: :billing_address_id
has_many :purchases_for_shipping, class_name: 'Purchase',
foreign_key: :shipping_address_id
has_many :purchases_for_billing, class_name: 'Purchase',
foreign_key: :billing_address_id
end
and I want to fetch all addresses of a certain Customer through his orders a purchases using eager_load:
# customer.rb
def addresses
Address.eager_load(:orders_for_cleaning,
:orders_for_billing,
:purchases_for_shipping,
:purchases_for_billing)
.where('orders.customer_id = ?
OR orders_for_billings_addresses.customer_id = ?
OR purchases.customer_id = ?
OR purchases_for_billings_addresses.customer_id = ?',
id, id, id, id)
end
This request generates the following query:
SELECT DISTINCT "addresses"."id" AS t0_r0,
"addresses"."latitude" AS t0_r1,
"addresses"."longitude" AS t0_r2,
"addresses"."house" AS t0_r3,
"addresses"."street" AS t0_r4,
"addresses"."city" AS t0_r5,
"addresses"."zip" AS t0_r6,
"addresses"."state" AS t0_r7,
"addresses"."country" AS t0_r8,
"addresses"."created_at" AS t0_r9,
"addresses"."updated_at" AS t0_r10,
"addresses"."street2" AS t0_r11,
"addresses"."custom_latitude" AS t0_r12,
"addresses"."custom_longitude" AS t0_r13,
"addresses"."name" AS t0_r14,
"addresses"."company" AS t0_r15,
"addresses"."archived" AS t0_r16,
"orders"."id" AS t1_r0,
"orders"."customer_id" AS t1_r1,
"orders"."cleaning_address_id" AS t1_r2,
"orders"."billing_address_id" AS t1_r3,
"orders"."start_at" AS t1_r4,
"orders"."finish_at" AS t1_r5,
"orders"."planned_at" AS t1_r6,
"orders"."started_at" AS t1_r7,
"orders"."finished_at" AS t1_r8,
"orders"."payed_price" AS t1_r9,
"orders"."status" AS t1_r10,
"orders"."pin" AS t1_r11,
"orders"."comment" AS t1_r12,
"orders"."created_at" AS t1_r13,
"orders"."updated_at" AS t1_r14,
"orders"."name" AS t1_r15,
"orders"."company" AS t1_r16,
"orders"."phone" AS t1_r17,
"orders"."coupon_code" AS t1_r18,
"orders"."coupon_discount" AS t1_r19,
"orders"."location_id" AS t1_r20,
"orders"."location_discount" AS t1_r21,
"orders"."job_count" AS t1_r22,
"orders"."bonus" AS t1_r23,
"orders"."organization_id" AS t1_r24,
"orders"."bill_id" AS t1_r25,
"orders"."cleaning_count_discount" AS t1_r26,
"orders"."created_by" AS t1_r27,
"orders"."price_group_id" AS t1_r28,
"orders"."travel_charge" AS t1_r29,
"orders"."departed_at" AS t1_r30,
"orders"."arrived_at" AS t1_r31,
"orders"."request_feedback" AS t1_r32,
"orders"."feedback_link" AS t1_r33,
"orders"."feedback_requested_at" AS t1_r34,
"orders"."rating" AS t1_r35,
"orders"."feedback" AS t1_r36,
"orders"."profit_center_code" AS t1_r37,
"orders"."feedback_received_at" AS t1_r38,
"orders"."reorder_bonus" AS t1_r39,
"orders"."paid_to" AS t1_r40,
"orders"."cleaner_id" AS t1_r41,
"orders"."station_id" AS t1_r42,
"orders"."cleaner_comment" AS t1_r43,
"orders_for_billings_addresses"."id" AS t2_r0,
"orders_for_billings_addresses"."customer_id" AS t2_r1,
"orders_for_billings_addresses"."cleaning_address_id" AS t2_r2,
"orders_for_billings_addresses"."billing_address_id" AS t2_r3,
"orders_for_billings_addresses"."start_at" AS t2_r4,
"orders_for_billings_addresses"."finish_at" AS t2_r5,
"orders_for_billings_addresses"."planned_at" AS t2_r6,
"orders_for_billings_addresses"."started_at" AS t2_r7,
"orders_for_billings_addresses"."finished_at" AS t2_r8,
"orders_for_billings_addresses"."payed_price" AS t2_r9,
"orders_for_billings_addresses"."status" AS t2_r10,
"orders_for_billings_addresses"."pin" AS t2_r11,
"orders_for_billings_addresses"."comment" AS t2_r12,
"orders_for_billings_addresses"."created_at" AS t2_r13,
"orders_for_billings_addresses"."updated_at" AS t2_r14,
"orders_for_billings_addresses"."name" AS t2_r15,
"orders_for_billings_addresses"."company" AS t2_r16,
"orders_for_billings_addresses"."phone" AS t2_r17,
"orders_for_billings_addresses"."coupon_code" AS t2_r18,
"orders_for_billings_addresses"."coupon_discount" AS t2_r19,
"orders_for_billings_addresses"."location_id" AS t2_r20,
"orders_for_billings_addresses"."location_discount" AS t2_r21,
"orders_for_billings_addresses"."job_count" AS t2_r22,
"orders_for_billings_addresses"."bonus" AS t2_r23,
"orders_for_billings_addresses"."organization_id" AS t2_r24,
"orders_for_billings_addresses"."bill_id" AS t2_r25,
"orders_for_billings_addresses"."cleaning_count_discount" AS t2_r26,
"orders_for_billings_addresses"."created_by" AS t2_r27,
"orders_for_billings_addresses"."price_group_id" AS t2_r28,
"orders_for_billings_addresses"."travel_charge" AS t2_r29,
"orders_for_billings_addresses"."departed_at" AS t2_r30,
"orders_for_billings_addresses"."arrived_at" AS t2_r31,
"orders_for_billings_addresses"."request_feedback" AS t2_r32,
"orders_for_billings_addresses"."feedback_link" AS t2_r33,
"orders_for_billings_addresses"."feedback_requested_at" AS t2_r34,
"orders_for_billings_addresses"."rating" AS t2_r35,
"orders_for_billings_addresses"."feedback" AS t2_r36,
"orders_for_billings_addresses"."profit_center_code" AS t2_r37,
"orders_for_billings_addresses"."feedback_received_at" AS t2_r38,
"orders_for_billings_addresses"."reorder_bonus" AS t2_r39,
"orders_for_billings_addresses"."paid_to" AS t2_r40,
"orders_for_billings_addresses"."cleaner_id" AS t2_r41,
"orders_for_billings_addresses"."station_id" AS t2_r42,
"orders_for_billings_addresses"."cleaner_comment" AS t2_r43,
"purchases"."id" AS t3_r0,
"purchases"."product_price_group_id" AS t3_r1,
"purchases"."customer_id" AS t3_r2,
"purchases"."shipping_address_id" AS t3_r3,
"purchases"."billing_address_id" AS t3_r4,
"purchases"."created_by" AS t3_r5,
"purchases"."token" AS t3_r6,
"purchases"."shipping" AS t3_r7,
"purchases"."coupon_code" AS t3_r8,
"purchases"."coupon_discount" AS t3_r9,
"purchases"."comment" AS t3_r10,
"purchases"."status" AS t3_r11,
"purchases"."purchased_at" AS t3_r12,
"purchases"."created_at" AS t3_r13,
"purchases"."updated_at" AS t3_r14,
"purchases"."purchase_invoice_id" AS t3_r15,
"purchases"."payment" AS t3_r16,
"purchases"."pickup" AS t3_r17,
"purchases"."tracking_number" AS t3_r18,
"purchases"."paper_invoice_sent_at" AS t3_r19,
"purchases_for_billings_addresses"."id" AS t4_r0,
"purchases_for_billings_addresses"."product_price_group_id" AS t4_r1,
"purchases_for_billings_addresses"."customer_id" AS t4_r2,
"purchases_for_billings_addresses"."shipping_address_id" AS t4_r3,
"purchases_for_billings_addresses"."billing_address_id" AS t4_r4,
"purchases_for_billings_addresses"."created_by" AS t4_r5,
"purchases_for_billings_addresses"."token" AS t4_r6,
"purchases_for_billings_addresses"."shipping" AS t4_r7,
"purchases_for_billings_addresses"."coupon_code" AS t4_r8,
"purchases_for_billings_addresses"."coupon_discount" AS t4_r9,
"purchases_for_billings_addresses"."comment" AS t4_r10,
"purchases_for_billings_addresses"."status" AS t4_r11,
"purchases_for_billings_addresses"."purchased_at" AS t4_r12,
"purchases_for_billings_addresses"."created_at" AS t4_r13,
"purchases_for_billings_addresses"."updated_at" AS t4_r14,
"purchases_for_billings_addresses"."purchase_invoice_id" AS t4_r15,
"purchases_for_billings_addresses"."payment" AS t4_r16,
"purchases_for_billings_addresses"."pickup" AS t4_r17,
"purchases_for_billings_addresses"."tracking_number" AS t4_r18,
"purchases_for_billings_addresses"."paper_invoice_sent_at" AS t4_r19
FROM "addresses"
LEFT OUTER JOIN "orders"
ON "orders"."cleaning_address_id" = "addresses"."id"
LEFT OUTER JOIN "orders" "orders_for_billings_addresses"
ON "orders_for_billings_addresses"."billing_address_id" = "addresses"."id"
LEFT OUTER JOIN "purchases"
ON "purchases"."shipping_address_id" = "addresses"."id"
LEFT OUTER JOIN "purchases" "purchases_for_billings_addresses"
ON "purchases_for_billings_addresses"."billing_address_id" = "addresses"."id"
WHERE (orders.customer_id = 3282 OR orders_for_billings_addresses.customer_id = 3282
OR purchases.customer_id = 3282 OR purchases_for_billings_addresses.customer_id = 3282)
this query takes about 100ms on my dataset.
But I don't need to load all join model columns just their foreign keys. If I execute stripped SQL query:
SELECT addresses.*,
orders.cleaning_address_id,
orders_for_billings_addresses.billing_address_id,
purchases.shipping_address_id,
purchases_for_billings_addresses.billing_address_id
FROM addresses
LEFT OUTER JOIN orders
ON orders.cleaning_address_id = addresses.id
LEFT OUTER JOIN orders orders_for_billings_addresses
ON orders_for_billings_addresses.billing_address_id = addresses.id
LEFT OUTER JOIN purchases
ON purchases.shipping_address_id = addresses.id
LEFT OUTER JOIN purchases purchases_for_billings_addresses
ON purchases_for_billings_addresses.billing_address_id = addresses.id
WHERE (orders.customer_id = 3282
OR orders_for_billings_addresses.customer_id = 3282
OR purchases.customer_id = 3282
OR purchases_for_billings_addresses.customer_id = 3282)
It returns the same results in only 50ms.
The question: is there a way to enforce ActiveRecord to select only necessary fields with eager_load? Adding select('addresses.*, <fk_fields...>) doesn't help.

Answer 1

eager_load is not what you are looking for based on your description. eager_load will load all the associations immediately. why not try using includes, references and select instead e.g.
def addresses
Address.includes(:orders_for_cleaning,
:orders_for_billing,
:purchases_for_shipping,
:purchases_for_billing).
references(:orders_for_cleaning,
:orders_for_billing,
:purchases_for_shipping,
:purchases_for_billing).
where("orders.customer_id = :id
OR orders_for_billings_addresses.customer_id = :id
OR purchases.customer_id = :id
OR purchases_for_billings_addresses.customer_id = :id",
{id: id}).
select("addresses.*,
orders.cleaning_address_id as cleaning_address_id,
orders_for_billings_addresses.billing_address_id as order_billing_address_id,
purchases.shipping_address_id as purchases_shipping_address_id,
purchases_for_billings_addresses.billing_address_id as purchases_billing_address_id")
end
I believe this will produce the query you are looking for and add methods for retrieval of cleaning_address_id,order_billing_address_id,purchases_shipping_address_id, and purchases_billing_address_id.
I have not tested this so I hope it helps if it does not let me know and I will remove it.
Update
Since the above code is still eager loading according to the OP. (of which I am not sure why and since the relationships are complex debugging personally seems difficult) I am suggesting building the query long hand.
relationship_joins = <<-SQL
LEFT OUTER JOIN orders
ON orders.cleaning_address_id = addresses.id
LEFT OUTER JOIN orders orders_for_billings_addresses
ON orders_for_billings_addresses.billing_address_id = addresses.id
LEFT OUTER JOIN purchases
ON purchases.shipping_address_id = addresses.id
LEFT OUTER JOIN purchases purchases_for_billings_addresses
ON purchases_for_billings_addresses.billing_address_id = addresses.id
SQL
Address.joins(relationship_joins).
where("orders.customer_id = :id
OR orders_for_billings_addresses.customer_id = :id
OR purchases.customer_id = :id
OR purchases_for_billings_addresses.customer_id = :id",
{id: id}).
select("addresses.*,
orders.cleaning_address_id as cleaning_address_id,
orders_for_billings_addresses.billing_address_id as order_billing_address_id,
purchases.shipping_address_id as purchases_shipping_address_id,
purchases_for_billings_addresses.billing_address_id as purchases_billing_address_id")
That should create the query you are requesting. It is not as pretty but sometimes you have to sacrifice beauty for functionality.