Question

First, I am a SQL noob, second if there is a better solution that can be implement in code later, I’d love to hear it. Also, the data is currently stored in MySQL but will eventually have to be ported to MSSQL so a cross DB solution would be best (if one exists).

Now, the problem, my simplified data looks like this:

[STYLES]
ID   NAME
1    A Style
2    B Style
...
N    N Style

[EQUIPMENT]
ID   NAME
1    A Equipment
2    B Equipment
...
N    N Equipment

[AVAILABILITY]
STYLE  EQUIPMENT  TYPE
1      1          Standard
1      2          Optional
2      1          Optional
... #items will be missing and represent not available
2      2          Standard

Now I need a table that looks like this:

[DESIRED_VIEW]
EQUIPMENT_NAME   A_STYLE_TYPE  B_STYLE_TYPE ... N_STYLE_TYPE
A Equipment      Standard      Optional     ... NULL
B Equipment      Optional      NULL         ... Standard

I have seen a number of simple pivot examples and they all rely on having a set number of columns. Is there a way to set up the view with a variable number of columns based on how many rows are in the STYLES table?

As a note, I am using Visual Studio to create the Data xsd and letting it auto-generate the table fill methods then displaying the info in WPF DataGrids so being able to bind directly to a view with the correct data would be ideal.

Answer 1

No – there is no way to write a query that has a dynamic number of columns.

Your choices are (in order of preference):

1. write a query that has “a lot” of columns, some of which may be blank (easy to write and use, but there would be a hard maximum on the pivot width)
2. write app code that writes the dynamic query for you (not too bad: portable)
3. write a stored procedure to create the SQL (terrible choice: painful to write logic code in PL/SQL and totally non-portable)

If you are comfortable with a maximum number of styles (it can be arbitrarily large, but finite), code your pivot would look like this:

select
  e.name as equipment_name,
  a1.type as a_style_type,
  a2.type as b_style_type,
  a3.type as c_style_type,
  ...
  an.type as n_style_type
from equipment e
left join availability a1 on a1.equipment = e.id and a1.style = 1
left join availability a2 on a2.equipment = e.id and a2.style = 2
left join availability a3 on a3.equipment = e.id and a3.style = 3
...
left join availability an on an.equipment = e.id and an.style = 999;

When there is no availability type for a given style, you’ll get a NULL.

Also, you need to know the style ids and they need to be the same on all environemtns. If this is not the case (eg different in test/prod environments) you can change a1.style = 1 to a1.style = (select id from style where name = 'A Style') etc – it will still perform OK.