T-SQL function loop [closed] - sql

Question

This question is unlikely to help any future visitors; it is only relevant to a small geographic area, a specific moment in time, or an extraordinarily narrow situation that is not generally applicable to the worldwide audience of the internet. For help making this question more broadly applicable, visit the help center.
Closed 9 years ago.
I was wondering if I could get some help on a T-SQL function I am trying to create:
Here is some sample data that needs to be queried:
Simplified table:
ID|PersonID|ValueTypeID|ValueTypeDescription|Value
1|ZZZZZ000L6|ZZZZZ00071|Start Prison Date|3/28/2012
2|ZZZZZ000L6|ZZZZZ00071|Start Prison Date|10/10/2012
3|ZZZZZ000L6|ZZZZZ00072|End Prison Date |3/29/2012
4|ZZZZZ000MD|ZZZZZ00071|Start Prison Date|1/15/2012
5|ZZZZZ000MD|ZZZZZ00072|End Prison Date |2/15/2012
6|ZZZZZ000MD|ZZZZZ00071|Start Prison Date|4/1/2012
7|ZZZZZ000MD|ZZZZZ00072|End Prison Date |4/5/2012
8|ZZZZZ000MD|ZZZZZ00071|Start Prison Date|9/3/2012
9|ZZZZZ000MD|ZZZZZ00072|End Prison Date |12/1/2012
What I need is a T-SQL function that accepts the PersonID and the Year (#PID, #YR) and returns the number of days that person has been in prison for that year.
dbo.NumDaysInPrison(#PID, #YR) as int
Example:
dbo.NumDaysInPrison('ZZZZZ000L6', 2012) returns 84
dbo.NumDaysInPrison('ZZZZZ000MD', 2012) returns 124
So far, I have come up with this query that gives me the answer sometimes.
DECLARE #Year int
DECLARE #PersonID nvarchar(50)
SET #Year = 2012
SET #PersonID = 'ZZZZZ000AA'
;WITH StartDates AS
(
SELECT
Value,
ROW_NUMBER() OVER(ORDER BY Value) AS RowNumber
FROM Prisoners
WHERE ValueTypeDescription = 'Start Prison Date' AND PersonID = #PersonID AND YEAR(Value) = #Year
), EndDates AS
(
SELECT
Value,
ROW_NUMBER() OVER(ORDER BY Value) AS RowNumber
FROM Prisoners
WHERE ValueTypeDescription = 'End Prison Date' AND PersonID = #PersonID AND YEAR(Value) = #Year
)
SELECT
SUM(DATEDIFF(d, s.Value, ISNULL(e.Value, cast(str(#Year*10000+12*100+31) as date)))) AS NumDays
FROM StartDates s
LEFT OUTER JOIN EndDates e ON s.RowNumber = e.RowNumber
This fails to capture if a record earlier in the year was left without an end date:
for example if a person has only two records:
ID|PersonID|ValueTypeID|ValueTypeDescription|Value
1|ZZZZZ000AA|ZZZZZ00071|Start Prison Date|3/28/2012
2|ZZZZZ000AA|ZZZZZ00071|Start Prison Date|10/10/2012
(3/28/2012 -> End of Year)
(10/10/2012 -> End of Year)
will returns 360, not 278.

Answer 1

So it seems that you have the data that you need to split out your 'start date' values and your 'end date' values. You don't really need to loop through anything, you can just pull out your start values then your end values based on your person and compare them.
The important thing is to pull out all you need to begin with and then compare the appropriate values.
Here's an example based on your data above. It would need some heavy tweaking to work with production data; it makes assumptions about the Value data. It's also a bad idea to hard-code valuetypeid as I have here; if you're making a function, you'd want to handle that, I think.
DECLARE #pid INT, #yr INT;
WITH startdatecalc AS
(
SELECT personid, CAST([value] AS date) AS startdate, DATEPART(YEAR, CAST([value] AS date)) AS startyear
FROM incarctbl
WHERE valuetypeid = 'ZZZZZ00071'
),
enddatecalc AS
(
SELECT personid, CAST([value] AS date) AS enddate, DATEPART(YEAR, CAST([value] AS date)) AS endyear
FROM incarctbl
WHERE valuetypeid = 'ZZZZZ00072'
)
SELECT CASE WHEN startyear < #yr THEN DATEDIFF(day, CAST(CAST(#yr AS VARCHAR(4)) + '-01-01' AS date), ISNULL(enddatecalc.enddate, CURRENT_TIMESTAMP))
ELSE DATEDIFF(DAY, startdate, ISNULL(enddatecalc.enddate, CURRENT_TIMESTAMP)) END AS NumDaysInPrison
FROM startdatecalc
LEFT JOIN enddatecalc
ON startdatecalc.personid = enddatecalc.personid
AND enddatecalc.enddate >= startdatecalc.startdate
AND NOT EXISTS
(SELECT 1 FROM enddatecalc xref
WHERE xref.personid = enddatecalc.personid
AND xref.enddate < enddatecalc.enddate
AND xref.enddate >= startdatecalc.startdate
AND xref.endyear < #yr)
WHERE startdatecalc.personid = #pid
AND startdatecalc.startyear <= #yr
AND (enddatecalc.personid IS NULL OR endyear >= #yr);
EDIT: Added existence check to attempt to handle if the same personid was used multiple times in the same year.

Answer 2

Here's my implementation with test tables and data. You'll have to change where appropriate. NOTE: i take datediff + 1 for days in prison, so if you go in on monday and leave on tuesday, that counts as two days. if you want it to count as one day, remove the "+ 1"
create table PrisonRegistry
(
id int not null identity(1,1) primary key
, PersonId int not null
, ValueTypeId int not null
, Value date
)
-- ValueTypeIDs: 1 = start prison date, 2 = end prison date
insert PrisonRegistry( PersonId, ValueTypeId, Value ) values ( 1, 1, '2012-03-28' )
insert PrisonRegistry( PersonId, ValueTypeId, Value ) values ( 1, 1, '2012-10-12' )
insert PrisonRegistry( PersonId, ValueTypeId, Value ) values ( 1, 2, '2012-03-29' )
insert PrisonRegistry( PersonId, ValueTypeId, Value ) values ( 2, 1, '2012-01-15' )
insert PrisonRegistry( PersonId, ValueTypeId, Value ) values ( 2, 2, '2012-02-15' )
insert PrisonRegistry( PersonId, ValueTypeId, Value ) values ( 2, 1, '2012-04-01' )
insert PrisonRegistry( PersonId, ValueTypeId, Value ) values ( 2, 2, '2012-04-05' )
insert PrisonRegistry( PersonId, ValueTypeId, Value ) values ( 2, 1, '2012-09-03' )
insert PrisonRegistry( PersonId, ValueTypeId, Value ) values ( 2, 2, '2012-12-1' )
go
create function dbo.NumDaysInPrison(
#personId int
, #year int
)
returns int
as
begin
declare #retVal int
set #retVal = 0
declare #valueTypeId int
declare #value date
declare #startDate date
declare #noDates bit
set #noDates = 1
set #startDate = DATEFROMPARTS( #year, 1, 1 )
declare prisonCursor cursor for
select
pr.ValueTypeId
, pr.Value
from
PrisonRegistry pr
where
DATEPART( yyyy, pr.Value ) = #year
and pr.ValueTypeId in (1,2)
and PersonId = #personId
order by
pr.Value
open prisonCursor
fetch next from prisonCursor
into #valueTypeId, #value
while ##FETCH_STATUS = 0
begin
set #noDates = 0
-- if end date, add date diff to retVal
if 2 = #valueTypeId
begin
--if #startDate is null
--begin
-- -- error: two end dates in a row
-- -- handle
--end
set #retVal = #retVal + DATEDIFF( dd, #startDate, #value ) + 1
set #startDate = null
end
else if 1 = #valueTypeId
begin
set #startDate = #value
end
fetch next from prisonCursor
into #valueTypeId, #value
end
close prisonCursor
deallocate prisonCursor
if #startDate is not null and 0 = #noDates
begin
set #retVal = #retVal + DATEDIFF( dd, #startDate, DATEFROMPARTS( #year, 12, 31 ) ) + 1
end
return #retVal
end
go
select dbo.NumDaysInPrison( 1, 2012 )
select dbo.NumDaysInPrison( 2, 2012 )
select dbo.NumDaysInPrison( 2, 2011 )

Answer 3

This is a complicated question. It is not so much "asking for a function" as it is dealing with two competing problems. The first is organizing the data, which is transaction-based, into records with start and stop dates for the prison period. The second is summarizing this for time spent within another given span of time (a year).
I think you need to spend some time investigating the data to understand the anomalies in it, before progressing to writing a function. The following query should help you. It does the calculate for all prisoners for a given year (which is the year in the first CTE):
with vals as (
select 2012 as yr
),
const as (
select cast(CAST(yr as varchar(255))+'-01-01' as DATE) as periodstart,
cast(CAST(yr as varchar(255))+'-12-31' as DATE) as periodend
from vals
)
select t.personId, SUM(datediff(d, (case when StartDate < const.periodStart then const.periodStart else StartDate end),
(case when EndDate > const.PeriodEnd or EndDate is NULL then const.periodEnd, else EndDate end)
)
) as daysInYear
from (select t.*, t.value as StartDate,
(select top 1 value
from t t2
where t.personId = t2.personId and t2.Value >= t.Value and t2.ValueTypeDescription = 'End Prison Date'
order by value desc
) as EndDate
from t
where valueTypeDescription = 'Start Prison Date'
) t cross join
const
where StartDate <= const.periodend and (EndDate >= const.periodstart or EndDate is NULL)
group by t.PersonId;
This query can be adapted as a function. But, I would encourage you to investigate the data before going there. Once you wrap things up in a function, it will be much more difficult to find and understand anomalies -- why did someone go in and out on the same day? How has the longest periods in prison? And so on.