Access SQL - Count consecutive weeks with a fail - sql

Question

I have a table in Access that has a record of test results for each person for each day. Some people may have taken more than one test on the same day. I have shown a simplified version of this table below:
| ID | testDate | Person | Pass? | ConsecFailDays |
----------------------------------------------------------
| 01 | 01/08/18 | John | Fail | |
| 02 | 01/08/18 | John | Pass | |
| 03 | 03/08/18 | John | Fail | |
| 04 | 01/08/18 | Mark | Fail | |
| 05 | 02/08/18 | Mark | Pass | |
I have tried to write an SQL UPDATE query that will update the last column with the number of consecutive days that that person has failed at least one test (not necessarily consecutive calender days, just the days that they actually did a test). The final result should look like this...
| ID | testDate | Person | Pass? | ConsecFailDays |
----------------------------------------------------------
| 01 | 01/08/18 | John | Fail | 1 |
| 02 | 01/08/18 | John | Pass | 1 |
| 03 | 03/08/18 | John | Fail | 2 |
| 04 | 01/08/18 | Mark | Fail | 1 |
| 05 | 02/08/18 | Mark | Pass | 0 |
I was really struggling to get this to work and in the end I resorted to using VBA to create a recordset of each unique person and then loop through each day for that person to check if they had a fail on that day. My dataset is quite large and it is taking hours to run.
I expect that a query that operates on the entire set of data would be much quicker. Does anyone know if there is an SQL solution to this after all?

Answer 1

You need several subqueries to get the last date with a passed test and no fail, then count the days between this date and testDate:
SELECT t1.ID
,t1.testDate
,t1.Person
,t1.[Pass?]
,(
SELECT count(testDate)
FROM (
SELECT DISTINCT testDate
,Person
,[Pass?]
FROM yourTable
) AS t2
WHERE t2.[Pass?] = false
AND t2.Person = t1.Person
AND t2.testDate >= Nz((
SELECT max(t3.testDate)
FROM (
SELECT t4.testDate
,t4.Person
FROM yourTable AS t4
WHERE (
((t4.[Pass?]) = True)
AND (
(
(
SELECT count(*) AS failed
FROM yourTable AS t5
WHERE t5.testDate = t4.testDate
AND t5.Person = t4.Person
AND t5.[Pass?] = false
)
) = 0
)
)
) AS t3
WHERE t1.Person = t3.Person
AND t3.testDate <= [t1].[testDate]
GROUP BY Person
))
AND t2.testDate <= t1.testDate
) AS ConsecFailDays
FROM yourTable AS t1;
Where t2 counts the distinct days (you can drop DISTINCT if just one fail per day is posible to speed up).
t3 are the days with passed tests and no fail.
t4 are the days with passed tests (and maybe failed tests).
t5 counts the failed tests on a day with a passed test.
As you wanted an Update-Query you can use:
UPDATE yourTable SET ConsecFailDays = DLookUp("ConsecFailDays","newQuery","ID = " & yourTable.ID)
but you should try the Select-Query with indices first. If performance is too poor you can use the update, but you have to run it every time your data changes.
Suggestions:
Don't use specialchars like ? in Pass? for column or table names to avoid being forced to use square brackets.
Person should be a foreign-key to table persons as persons can have equal names (e.g John Smith).
[Pass?] should be boolean (true/false). If you want to stay on string you have to replace [Pass?] = false with [Pass?] = 'Fail' and [Pass?] = true with
[Pass?] = 'Pass'
there should be an index for testDate, Person and a combined index for testDate, Person to increase query performance.