在AS400数据库上的SQL中使用“ order by”并在联合内部获取

Question

Let's say I have this table

Table name: Traffic
Seq.     Type     Amount
1        in       10
2        out      30
3        in       50
4        out      70

What I need is to get the previous smaller and next larger amount of a value. So, if I have 40 as a value, I will get...

Table name: Traffic
Seq.     Type     Amount
2        out      30
3        in       50

I already tried doing it with MYSQL and quite satisfied with the results

(select * from Traffic where
Amount < 40 order by Amount desc limit 1)
union
(select * from Traffic where
Amount > 40 order by Amount desc limit 1)

The problem lies when I try to convert it to a SQL statement acceptable by AS400. It appears that the order by and fetch function (AS400 doesn't have a limit function so we use fetch, or does it?) is not allowed inside the select statement when I use it with a union. I always get a keyword not expected error. Here is my statement;

(select seq as sequence, type as status, amount as price from Traffic where
Amount < 40 order by price asc fetch first 1 rows only)
union
(select seq as sequence, type as status, amount as price  from Traffic where
Amount > 40 order by price asc fetch first 1 rows only)

Can anyone please tell me what's wrong and how it should be? Also, please share if you know other ways to achieve my desired result.

Answer 1

How about a CTE? From memory (no machine to test with):

with 
  less as (select * from traffic where amount < 40),
  more as (select * from traffic where amount > 40)
select * from traffic
  where id = (select id from less where amount = (select max(amount from less))) 
     or id = (select id from more where amount = (select min(amount from more))) 

Answer 2

I looked at this question from possibly another point of view. I have seen other questions about date-time ranges between rows, and I thought perhaps what you might be trying to do is establish what range some value might fall in.

If working with these ranges will be a recurring theme, then you might want to create a view for it.

create or replace view traffic_ranges as
  with sorted as
  ( select t.*
         , smallint(row_number() over (order by amount)) as pos
     from traffic t
  )
  select b.pos    range_seq
       , b.id     beg_id
       , e.id     end_id
       , b.typ    beg_type
       , e.typ    end_type
       , b.amount beg_amt
       , e.amount end_amt
    from sorted b
    join sorted e  on e.pos = b.pos+1
;

Once you have this view, it becomes very simple to get your answer:

select * 
  from traffic_ranges
  where 40 is between beg_amt and end_amt

Or to get only one range where the search amount happens to be an amount in your base table, you would want to pick whether to include the beginning value or ending value as part of the range, and exclude the other:

  where beg_amt < 40 and end_amt >= 40

One advantage of this approach is performance. If you are finding the range for multiple values, such as a column in a table or query, then having the range view should give you significantly better performance than a query where you must aggregate all the records that are more or less than each search value.

Answer 3

Here's my query using CTE and union inspired by Buck Calabro's answer. Credits go to him and WarrenT for being SQL geniuses!

I won't be accepting my own answer. That will be unfair. hehe

 with 
apple(seq, type, amount) as (select seq, type, amount from traffic where amount < 40 
order by amount desc fetch first 1 rows only),
banana(seq, type, amount) as (select seq, type, amount from traffic where
amount > 40 fetch first 1 rows only)
 select * from apple
union
 select * from banana

It's a bit slow but I can accept that since I'll only use it once in the progam.

This is just a sample. The actual query is a bit different.