Normalization

Due to the requirement that the value of any column in any row in any table must be atomic, some relational databases suffer these problems:

• A data value becomes null
• A data value becomes repeated many times
• A logical update becomes multiple physical updates

What data values are likely to be null in our example database?
What data values are repeated many times in our example database?
What logical updates are manifested into multiple physical updates in our database?

To avoid such problems:

• either ``normalize'' database into 3rd normal form (3NF):
  • do not store the same fact twice
  • do not store unrelated facts in the same table
• or use ``nested'' relational database instead

Example normalized database:

genpepts: {(
   #accn:  string,
   #uid:   num,
   #title: string,
   #seq:   string)}

features: {(
   #uid:   num,
   #fid:   num,
   #start: num,
   #end:   num,
   #name:  string
   #anno:  string)}