我正在从运行了20多分钟的几个大表的查询开发数据集市。为了使代码更具可读性并避免使用幻数和单词，我使用hivevarMap，而不是常量：

select b.id1 as id1,
       -- near 30 fields with nested if, match case etc, like
       if( d7.recordid = '01'
               or d7.recordid = '02'
               or -- ... 10 expressions
           'expression1',
           'expression2'
           ) as trans_type,
       b.id2 as id2

此查询（没有配置单元Map）在13分钟内执行。
请告诉我如何改进这个查询，并避免里面的幻数，使它更快。

set hivevar:transaction_class_code=map('INTERNAL',     '01',
                                       'INTERNATIONAL',  '02',
                                       -- 10 elements
                                       );

select b.id1 as id1,
       -- near 30 fields with nested if, match case etc, like
       if( d7.recordid = ${transaction_class_code}['INTERNAL']
               or d7.recordid = ${transaction_class_code}['INTERNATIONAL']
               or -- ... 10 expressions
           'expression1',
           'expression2'
           ) as trans_type,
       b.id2 as id2
from banks b -- ~10_000_000_000 rows, ~100 columns
join trans t -- 10_000_000 rows
     on b.id1 = t.id1
    and b.id2 = t.id2
left join customer c -- 3_000_000_000 rows
     on b.customer_id = c.transaction_customer_id
    and b.id1 = c.bank_id
left join transactions_pa  tp -- 500_000_000 rows
     on b.id1 = tp.bank_id
    and b.id2 = tp.transaction_id
left join transaction_dict   td -- 100_000_000
     on b.id1 = td.bank_id
    and b.id2 = td.transaction_id
left join dict1 d1 --5_152
     on b.instance = d1.instance
    and b.transaction_dict1_id = d1.recid
-- 4 dictionaries with 5 - 10 columns and 500 - 3000 rows
left join dictionary6 d7 -- 2116
     on b.instance = d6.instance
    and b.transaction_dict6_id = d6.recid;