Using a Single Row configuration table in SQL Server database. Bad idea?

Question

In developing a shopping cart application I\'ve found that I needed to save settings and configurations based on the administrator\'s preferences and requirements. This info

Answer 1

A single row will work fine; it will even have strong types:

show_borders    bit
admin_name      varchar(50)
max_users       int

One disadvantage is that it requires a schema change (alter table) to add a new setting. One alternative is normalizing, where you end up with a table like:

pref_name       varchar(50) primary key
pref_value      varchar(50) 

This has weak types (everything is a varchar), but adding a new setting is just adding a row, something you can do with just database write access.

Answer 2

A Key and Value pair is similar to a .Net App.Config which can store configuration settings.

So when you want to retrieve the value you could do:

SELECT value FROM configurationTable
WHERE ApplicationGroup = 'myappgroup'
AND keyDescription = 'myKey';

Answer 3

Have a key column as varchar and a value column as JSON. 1 is numeric whereas "1" is a string. true and false are both boolean. You can have objects as well.

Answer 4

Sorry I come like, yeaars later. But anyways, what I do is simple and effective. I simply create a table with three () columns:

ID - int (11)

name - varchar (64)

value - text

What I do before creating a new config column, updating it or reading is to serialize the "value"! This way I am sure of the type (Well, php is :) )

For instance:

b:0; is for BOOLEAN (false)

b:1; is for BOOLEAN (true)

i:1988; is for INT

s:5:"Kader"; is for a STRING of 5 characters length

I hope this helps :)

Answer 5

You can do the Key/Value Pair without conversions by adding a column for each major type and one column telling you which column the data is in.

So your table would look something like:

id, column_num, property_name, intValue, floatValue, charValue, dateValue
1, 1, weeks, 51, , ,
2, 2, pi, , 3.14159, , 
3, 4, FiscYearEnd, , , , 1/31/2015
4, 3, CompanyName, , , ACME, 

It uses a little more room but at most you are using a few dozen attributes. You can use a case statement off the column_num value to pull / join the right field.

Answer 6

You should create a table with a column for the information type and information value (at least). This way you avoid having to create new columns every time a new information is added.