Using a subquery in the where clause to select 2nd highest date from a table

Question

I have a need to do (in psuedo code)

where yyyy_mm_dd >= '2019-02-01' 
and yyyy_mm_dd <= second highest date in a table

To achieve this, I've used this code:

where
    p.yyyy_mm_dd >= "2019-02-02"
    and p.yyyy_mm_dd <= (select max(yyyy_mm_dd) from schema.table1 where yyyy_mm_dd < (select max(yyyy_mm_dd) from schema.table1 where yyyy_mm_dd is not null))

The above works when it is wrapped in spark.sql() but when I run the query without Spark ie as raw HQL, I run into this error:

Error while compiling statement: FAILED: ParseException line 102:25 cannot recognize input near 'select' 'max' '(' in expression specification

I tried to fix it by aliasing all columns in the subquery like this:

where
    p.yyyy_mm_dd >= "2019-02-02"
    and p.yyyy_mm_dd <= (select max(t1.yyyy_mm_dd) from schema.table1 t1 where t1.yyyy_mm_dd < (select max(t2.yyyy_mm_dd) from schema.table2 t2 where t2.yyyy_mm_dd is not null))

Though, I still run into the same error.

---

Edit to include sample data and query:

table1:

| yyyy_mm_dd | company_id | account_manager |
|------------|------------|-----------------|
| 2020-11-10 | 321        | Peter           |
| 2020-11-09 | 632        | John            |
| 2020-11-08 | 598        | Doe             |
| 2020-11-07 | 104        | Bob             |
| ...        | ...        | ...             |
| ...        | ...        | ...             |

table2:

| yyyy_mm_dd        | company_id | tier   |
|-------------------|------------|--------|
| 2020-11-10        | 321        | Bronze |
| 2020-11-09        | 632        | Silver |
| 2020-11-08        | 598        | Gold   |
| 2020-11-07        | 104        | Bob    |
| ...               | ...        | ...    |
| ...               | ...        | ...    |
| 2019_12_13_backup | 321        | Bronze |
| 2019_12_13_backup | 632        | Silver |
| ...               |            |        |

Query:

select
    p.yyyy_mm_dd,
    p.company_id,
    p.account_manager,
 t.tier
from
    table1 p
left join(
    select
        yyyy_mm_dd,
        company_id,
        max(tier) as tier
    from 
        table2
    where
        yyyy_mm_dd >= "2019-02-02"
    group by
        1,2
) t on (t.company_id = p.company_id and t.yyyy_mm_dd = p.yyyy_mm_dd)

where
    p.yyyy_mm_dd >= "2019-02-02"
    and p.yyyy_mm_dd <= (select max(yyyy_mm_dd) from table2 where yyyy_mm_dd < (select max(yyyy_mm_dd) from table2 where yyyy_mm_dd is not null))

As table2 contains backup_2019_12_31 in the yyyy_mm_dd column, those rows will be returned when doing max() on the table. So I need to get the second highest value, which from the dataset here would be 2020-11-10 . There are multiple company_ids per yyyy_mm_dd .

In essence, I want to query table1 where yyyy_mm_dd is between table1 starting point (hardcoded as 2019-02-02 ) and the true max date from table2

Answer 1

To get the second highest date from table3 you can use dense_rank. All rows with second highest date will be assigned rn=2. Use LIMIT to get single row or use max() or distinct aggregation for the same, then cross join your table with max_date and filter.

with max_date as(
select yyyy_mm_dd
from
(
select yyyy_mm_dd, 
       dense_rank() over(order by yyyy_mm_dd desc) rn
 from table2
)s 
where rn=2 --second max date
limit 1    --need only one record
)

select t1.*   
   from table1 t1
        cross join max_date t2
 where t1.yyyy_mm_dd <= t2.yyyy_mm_dd