[Bioperl-l] Building Databases

[Mike Myers]

< This isn't a bioperl question per se, hope this is OK. >

In an effort to learn-by-doing, I am converting the LocusLink LL_tmpl file 
from NCBI into a mySQL database. I am almost done with the parser, 
pondering more about whether my schema is sound.

When dealing with relationships that are one-to-many, is it considered 
acceptable form to create child record IDs that serialize a parental ID? It 
seems that it would be convenient from a query standpoint. And for the 
child records, is it necessary to create a separate record ID that is 
unique for the table?

When the record ID is numeric, ie. a locus ID, is there any reason _not_ to 
use a numeric column type? I was planning to use a FLOAT to serialize a 
numeric ID (the whole number part being the parent ID). Given the 
relatively small size of LocusLink, perhaps worrying about numeric vs 
string IDs is overkill.

thanks,

-- mpm