如何优化嵌套选择？

Question

My use-case is to retrieve invoices for a customer and a format code for each invoice which is found by a join to a customer_list table which a join to a generalised code table to work out the format code.

A simplified version of my query follows, which is working fine, yet the subquery would return around 5k of rows, I have to wonder how this could be optimised so the sub-query did not contain as many rows.

select i.invoice_number,
       f.format_code
  from invoices i
  left join (
    select gc.code_1 as format_code,
           ls.account
      from gen_codes gc,
           customer_list ls
     where gc.code_name = 'invoice-format'
       and gc.code_1 <> ''
       and ls.list_type = gc.code_value
  ) f on account = i.account
      or account = i.billing_account
where i.account = :accountNumber

NOTE : Not all invoices have a format code, which is fine and why I am performing a left join.

UPDATE : A similar query without using an inner select would be of the following form:

select i.invoice_number,
       f.code_1 as format_code
  from invoices i
  left join customer_list ls on
    (account = i.account or account = i.billing_account) and ls.list_type in (
           select code_value
             from gen_codes
            where code_name = 'invoice-format'
              and code_1 <> ''
        )
  left join gen_codes f on code_value = ls.list_type
                       and code_name = 'invoice-format'
                       and code_1 <> ''

I have a feeling this latter version may be more efficient as there are only two dozen records in gen_codes with the matched query while customer_list contains ~ half a million records with ~5k matching in the first version.

Answer 1

I would write it as follows:

Get all the invoices corresponding to an account or billing account (using JOIN and ON clauses) and get only those gen_codes associated with the list_types of the account matched (using LEFT JOIN) whose code_name is 'invoice-format' and gc.code_1 is not empty.

SELECT 
    i.invoice_number,
    gc.code_1 as format_code
FROM 
    invoices i
    JOIN customer_list ls
    ON 
        (i.billing_account = ls.account OR i.account = ls.account) AND 
        i.account = :accountNumber
    LEFT JOIN gen_codes gc
    ON 
        ls.list_type = gc.code_value AND 
        gc.code_name = 'invoice-format' AND 
        gc.code_1 != ''

This will prune the customer_list records that doesn't match the accountNumber input upfront, compared to the subquery in your question.