Normalization of a table

There's a little aphorism that goes:

• The key (1NF)
• The whole key (2NF)
• And nothing but the key (3NF)

. . . So help me Codd

In your example we can assume 1NF to begin with as the relational structure doesn't imply any repeating groups within the row (i.e. no D₁, D₂, D₃ etc.).

R = {a, b, c, d, e, f, g} F = {AB --> C, A --> DE, B --> F, F --> GH, D --> IJ }

2NF deals with 'the whole key' - if you have a composite key and some members of the relation are dependent on a part of that key then they should be split off into their own relations. In your case:

• R1 = AB --> C is in 2NF as 'AB' is 'the whole key'

• R2 = A --> DEIJ is 2NF but not 3NF as A --> DE, but A --> IJ is not 3NF as the real dependency of I and J is D --> IJ. D is not a key of this relation (A --> DEIJ), so A --> IJ violates the 'nothing but the key' princple of 3NF.

• R3 = B --> FGH is in 2NF but not 3NF as B --> F is a correct relation and F is dependent on the entirety of the key 'B'. However, B --> GH is not in 3NF as the correct relation is F --> GH. B --> GH violates the 'nothing but the key' principle, as G ahd H are correctly dependent on the non-key attribute F.

Your decomposition to 2NF is correct. Decomposition to 3NF requires taking the non-key attributes that have their own dependencies into separate relations.

The relation in 3NF would look like:

• R1 = AB --> C

• R2 = A --> DE (I and J are dependent on the non-key attribute D)

• R3 = B --> F (G and H are dependent on the non-key attribute F)

• R4 = D --> IJ

• R5 = F --> GH

don't add more stat columns unless you are absolutely sure the required # of columns wont grow, tables tend to get unmanageable that way.

if you need to work with the individual stats in your comma separated text field, you probably should break them off into their own table where each row has 1 stat, make guid+statid your primary key and then you can make the stat column something more reasonable like a varchar (20). This might take up a little more diskspace but it should be faster working with individual stats...