Update by LEAD/LAG or Recursive CTE? - mysql

Question

There is a table holding date and account balance.
However, the balance is not available for some dates.
Assuming the balance does not change when date is unavailable.
How to update the balance information for all dates?
Here is an example:
Table D contains all valid dates.
2000-01-01
2000-01-02
2000-01-03
2000-01-04
2000-01-05
2000-01-06
2000-01-07
2000-01-08
2000-01-09
Table A contains date and account balance.
2000-01-02 $100
2000-01-05 $200
2000-01-09 $700
Ultimately, I want to generate a table like this:
2000-01-01 null
2000-01-02 $100
2000-01-03 $100
2000-01-04 $100
2000-01-05 $200
2000-01-06 $200
2000-01-07 $200
2000-01-08 $200
2000-01-09 $700
I have thought about the following:
LEAD and LAG,
Recursive CTE
However, it seems that none of them is suitable for this scenario.

Answer 1

SQL Server does not support the IGNORE NULLS option for LAG() or LAST_VALUE(). That is actually the simplest method.
Instead, you can use APPLY:
select d.*, a.balance
from dates d outer apply
(select top (1) a.*
from a
where a.date <= d.date
order by a.date desc
) a;
Or the equivalent using a correlated subquery:
select d.*,
(select top (1) a.*
from a
where a.date <= d.date
order by a.date desc
fetch first 1 row only
)
from dates d;
This will work in both MySQL and SQL Server with the caveat that you need LIMIT in MySQL.
That said, if you had a large amount of data (which is unlikely at the granularity of "date"), then a two-steps of window functions are probably the better solution:
select ad.date,
max(ad.balance) over (partition by grp) as balance
from (select d.date, a.balance,
count(a.date) over (order by d.date) as grp
from dates d left join
a
on d.date = a.date
) ad;
The subquery assigns a "group" to each balance value and the following dates. This "group" is then used to assign the balance in the outer query.
This version will work in both MySQL or SQL Server.

Answer 2

One way you could achieve this is by using first value and creating some ranking functions. I am using SQL server
with cte as
(
select '2000-01-01' as Datenew union all
select '2000-01-02' as Datenew union all
select '2000-01-03' as Datenew union all
select '2000-01-04' as Datenew union all
select '2000-01-05' as Datenew union all
select '2000-01-06' as Datenew union all
select '2000-01-07' as Datenew union all
select '2000-01-08' as Datenew union all
select '2000-01-09' as Datenew ), cte2 as (
select '2000-01-02' as DateSal, '100' as Salary union all
select '2000-01-05' as DateSal, '200' as Salary union all
select '2000-01-09' as DateSal, '700' as Salary )
select datenew, Salary = FIRST_VALUE(salary) over (partition by ranking order by datenew) from (
select datenew, salary ,
sum(case when DateSal is not null then 1 end) over (order by datenew) ranking
from cte c
left join cte2 c2 on c.Datenew = c2.DateSal ) tst
order by datenew
--Sum creates running total to create a grouping and first value ensures that we are getting the same value for the given group.
this is the output

Answer 3

ANSI SQL.
Table_D
-------
dd(field name)
-------
2000-01-01
2000-01-02
2000-01-03
2000-01-04
2000-01-05
2000-01-06
2000-01-07
2000-01-08
2000-01-09
Table_A
-------
dd(field name) cost(field name)
-------
2000-01-02 $100
2000-01-05 $200
2000-01-09 $700
select a.dd
, (
case when a.cost is null then min(a.cost) OVER (partition by a.cost_group ORDER BY a.dd) else a.cost end
) as cost
from (
select a.dd, b.cost
, count(b.cost) over (order by a.dd) as cost_group
from Table_D a
left join Table_A b on (b.dd = a.dd)
) a

Answer 4

We can use count() over () window function to create different partitions and then min() over () function to spread minimum value to that partition.
First, I have created temporary variable table to hold OP data -
declare #xyz table (dt date, amount int)
insert into #xyz values
('2000-01-02','100'),
('2000-01-05','200'),
('2000-01-09','700');
Second, I will fetch max date from above table.
declare #maxDT date = (select cast(max(dt) as date) from #xyz);
Final, first CTE is recursive CTE to create data from 2000-01-01 to max date store in above variable. Second CTE is to create partitions.
;with cte as (
select cast('2000-01-01' as date) as dt
union all
select dateadd(day,1,cte.dt) from cte where cte.dt < #maxDT
), cte2 as (
select cte.dt, x.amount, x.dt as dt2, count(x.dt) over (order by cte.dt) as ranking
from cte
left join #xyz x on x.dt = cte.dt
)
select dt, min(amount) over (partition by ranking)
from cte2;