Encuentra un número único de días

Deseo escribir una consulta SQL para encontrar el número de días hábiles únicos para cada empleado de la tabla times.

*---------------------------------------*
|emp_id  task_id  start_day   end_day   |
*---------------------------------------*
|  1        1     'monday'  'wednesday' |
|  1        2     'monday'  'tuesday'   |
|  1        3     'friday'  'friday'    |
|  2        1     'monday'  'friday'    |
|  2        1     'tuesday' 'wednesday' |
*---------------------------------------*

Rendimiento esperado:

*-------------------*
|emp_id  no_of_days |
*-------------------*
|  1        4       |
|  2        5       |
*-------------------*

He escrito la consulta sqlfiddle que me está dando el expectedresultado, pero por curiosidad, ¿hay una mejor manera de escribir esta consulta? ¿Puedo usar el calendario o la tabla de conteo?

with days_num as  
(
  select
    *,
    case 
      when start_day = 'monday' then 1
      when start_day = 'tuesday' then 2
      when start_day = 'wednesday' then 3
      when start_day = 'thursday' then 4
      when start_day = 'friday' then 5
    end as start_day_num,

    case 
      when end_day = 'monday' then 1
      when end_day = 'tuesday' then 2
      when end_day = 'wednesday' then 3
      when end_day = 'thursday' then 4
      when end_day = 'friday' then 5
    end as end_day_num

  from times
),
day_diff as
(
  select
    emp_id,
    case
      when  
        (end_day_num - start_day_num) = 0
      then
        1
      else
        (end_day_num - start_day_num)
    end as total_diff
  from days_num  
)

select emp_id,
  sum(total_diff) as uniq_working_days
from day_diff
group by
  emp_id

Cualquier sugerencia seria genial.

Básicamente, necesita encontrar la intersección de los días trabajados por cada uno emp_iden cada uno taskcon todos los días de la semana, y luego contar los días distintos:

with days_num as (
  SELECT *
  FROM (
    VALUES ('monday', 1), ('tuesday', 2), ('wednesday', 3), ('thursday', 4), ('friday', 5)
  ) AS d (day, day_no)
),
emp_day_nums as (
  select emp_id, d1.day_no AS start_day_no, d2.day_no AS end_day_no
  from times t
  join days_num d1 on d1.day = t.start_day
  join days_num d2 on d2.day = t.end_day
)
select emp_id, count(distinct d.day_no) AS distinct_days
from emp_day_nums e
join days_num d on d.day_no between e.start_day_no and e.end_day_no
group by emp_id

Salida:

emp_id  distinct_days
1       4
2       5

Demostración en SQLFiddle

Un posible enfoque para simplificar el enunciado en la pregunta (violín) es utilizar el VALUESconstructor de valores de tabla y las uniones apropiadas:

SELECT 
   t.emp_id,
   SUM(CASE 
      WHEN d1.day_no = d2.day_no THEN 1
      ELSE d2.day_no - d1.day_no
   END) AS no_of_days
FROM times t
JOIN (VALUES ('monday', 1), ('tuesday', 2), ('wednesday', 3), ('thursday', 4), ('friday', 5)) d1 (day, day_no) 
   ON t.start_day = d1.day
JOIN (VALUES ('monday', 1), ('tuesday', 2), ('wednesday', 3), ('thursday', 4), ('friday', 5)) d2 (day, day_no) 
   ON t.end_day = d2.day
GROUP BY t.emp_id

Pero si desea contar los días distintos , la declaración es diferente. Es necesario encontrar todos los días entre el start_dayy el end_dayrango y contar los días distintos:

;WITH daysCTE (day, day_no) AS (
   SELECT 'monday', 1 UNION ALL
   SELECT 'tuesday', 2 UNION ALL
   SELECT 'wednesday', 3 UNION ALL
   SELECT 'thursday', 4 UNION ALL
   SELECT 'friday', 5 
)
SELECT t.emp_id, COUNT(DISTINCT d3.day_no)
FROM times t
JOIN daysCTE d1 ON t.start_day = d1.day
JOIN daysCTE d2 ON t.end_day = d2.day
JOIN daysCTE d3 ON d3.day_no BETWEEN d1.day_no AND d2.day_no
GROUP BY t.emp_id

Su consulta no es correcta Intente de lunes a martes con miércoles a jueves. Esto debería resultar en 4 días, pero su consulta devuelve 2 días. Su consulta ni siquiera detecta si dos rangos son adyacentes o se superponen o tampoco.

Una forma de resolver esto es escribir un CTE recursivo para obtener todos los días de un rango y luego contar días distintos.

with weekdays (day_name, day_number) as
(
  select * from (values ('monday', 1), ('tuesday', 2), ('wednesday', 3),
                        ('thursday', 4), ('friday', 5)) as t(x,y)
)
, emp_days(emp_id, day, last_day)
as
(
  select emp_id, wds.day_number, wde.day_number
  from times t
  join weekdays wds on wds.day_name = t.start_day
  join weekdays wde on wde.day_name = t.end_day
  union all
  select emp_id, day + 1, last_day
  from emp_days
  where day < last_day
)
select emp_id, count(distinct day)
from emp_days
group by emp_id
order by emp_id;

Demostración: http://sqlfiddle.com/#!18/4a5ac/16

_{(Como se puede ver, no pude aplicar el constructor de valores directamente como en with weekdays (day_name, day_number) as (values ('monday', 1), ...). No sé por qué. ¿Ese es SQL Server o yo? Bueno, con la selección adicional funciona :-)}

with cte as 
(Select id, start_day as day
   group by id, start_day
 union 
 Select id, end_day as day
   group by id, end_day
)

select id, count(day)
from cte
group by id

declare @times table
(
  emp_id int,
  task_id int,
  start_day varchar(50),
  end_day varchar(50)
);

insert into @times(emp_id, task_id, start_day, end_day)
values
(1, 1, 'monday', 'wednesday'),
(1, 2, 'monday', 'tuesday'),
(1, 3, 'friday', 'friday'),
--
(2, 1, 'monday', 'friday'),
(2, 2, 'tuesday', 'wednesday'),
--
(3, 1, 'monday', 'wednesday'),
(3, 2, 'monday', 'tuesday'),
(3, 3, 'monday', 'tuesday');

--for sql 2019, APPROX_COUNT_DISTINCT() eliminates distinct sort (!!)...
-- ...with a clustered index on emp_id (to eliminate the hashed aggregation) the query cost gets 5 times cheaper ("overlooking" the increase in memory) !!??!!
/*
select t.emp_id, APPROX_COUNT_DISTINCT(v.val) as distinctweekdays
from
(
select *, .........
*/


select t.emp_id, count(distinct v.val) as distinctweekdays
from
(
select *, 
case start_day when 'monday' then 1
      when 'tuesday' then 2
      when 'wednesday' then 3
      when 'thursday' then 4
      when 'friday' then 5
    end as start_day_num,
case end_day when 'monday' then 1
      when 'tuesday' then 2
      when 'wednesday' then 3
      when 'thursday' then 4
      when 'friday' then 5
    end as end_day_num
from @times
) as t
join (values(1),(2), (3), (4), (5)) v(val) on v.val between t.start_day_num and t.end_day_num
group by t.emp_id;