Question

Alright, first you’ll need to grab a barf bag. I’ve been tasked with optimizing several old stored procedures in our database. This SP does the following:

1) cursor loops through a series of “buildings”

2) cursor loops through a week, Sunday-Saturday

3) has a huge set of IF blocks that are responsible for counting how many Objects of what Types are present in a given building

Essentially what you’ll see in this code block is that, if there are 5 objects of type #2, it will increment @Type_2_Objects_5 by 1.

    IF @Number_Type_1_Objects = 0
    BEGIN
        SET @Type_1_Objects_0 = @Type_1_Objects_0 + 1
    END
    IF @Number_Type_1_Objects = 1
    BEGIN
        SET @Type_1_Objects_1 = @Type_1_Objects_1 + 1
    END
    IF @Number_Type_1_Objects = 2
    BEGIN
        SET @Type_1_Objects_2 = @Type_1_Objects_2 + 1
    END
    IF @Number_Type_1_Objects = 3
    BEGIN
        SET @Type_1_Objects_3 = @Type_1_Objects_3 + 1
    END
[... Objects_4 through Objects_20 for Type_1]

IF @Number_Type_2_Objects = 0
    BEGIN
        SET @Type_2_Objects_0 = @Type_2_Objects_0 + 1
    END
    IF @Number_Type_2_Objects = 1
    BEGIN
        SET @Type_2_Objects_1 = @Type_2_Objects_1 + 1
    END
    IF @Number_Type_2_Objects = 2
    BEGIN
        SET @Type_2_Objects_2 = @Type_2_Objects_2 + 1
    END
    IF @Number_Type_2_Objects = 3
    BEGIN
        SET @Type_2_Objects_3 = @Type_2_Objects_3 + 1
    END
[... Objects_4 through Objects_20 for Type_2]

In addition to being extremely hacky (and limited to a quantity of 20 objects), it seems like a terrible way of handling this. In a traditional language, this could easily be solved with a 2-dimensional array…

objects[type][quantity] += 1;

I’m a T-SQL novice, but since writing stored procedures often uses a lot of temporary tables (which could essentially be a 2-dimensional array) I was wondering if someone could illuminate a better way of handling a situation like this with two dynamic pieces of data to store.

Requested in comments

---

The columns are simply

• Number_Type_1_Objects
• Number_Type_2_Objects
• Number_Type_3_Objects
• Number_Type_4_Objects
• Number_Type_5_Objects
• CurrentDateTime.

Each row in the table represents 5 minutes.

The expected output is to figure out what percentage of time a given quantity of objects is present throughout each day.

Sunday - Object Type 1 
0 objects - 69 rows, 5:45, 34.85% 
1 object - 85 rows, 7:05, 42.93% 
2 objects - 44 rows, 3:40, 22.22%

On Sunday, there were 0 objects of type 1 for 34.85% of the day. There was 1 object for 42.93% of the day, and 2 objects for 22.22% of the day. Repeat for each object type.

Answer 1

a quick and easy solution that would get you close to where you want to be would be to do a union similar to

select 'type1objects' as objecttype
       datepart(dw, currentdatetime) as dayofweek, 
       number_type_1_objects as numberofobjects, 
       count(number_type_1_objects) as rows
from yourtable
group by convert(varchar(10), currentdatetime,101), datepart(dw, currentdatetime), number_type_1_objects
union
select 'type2objects' as objecttype
       datepart('d',currentdatetime) as dayofweek,
       number_type_2_objects as numberofobjects, 
       count(*) as rows
from yourtable
group by convert(varchar(10),currentdatetime,101), datepart(dw,currentdatetime), number_type_2_objects

and so on for each object column. that should get you an approximation of the data you are getting now without the use of those crazy variables and cursors, buildings not withstanding. you could probably join the buildings table in though, just adding it to the group by and be done with that as well. not sure though as i have no idea what the relationship would be drawn on…