Summary of Normal Forms

The following are not formal definitions of normal form, but it is hoped they will serve as reminders of the important reasons that these normal forms were originally suggested. 

First normal form Each table should contain only data about a unique entity. Each row should have a unique identifier. If data tables violate first normal form, a careful reconsideration of which information belongs in which table should be undertaken. A structure table should be about structures, a logP table about logP values, and a method table about methods. 

Second normal form Each row should contain only one value for each column. If multiple values of a data item are needed, a related table should be created for those values. Do not encode data fields into a data column. Create separate columns for each field and in a separate table, if necessary. 

Third normal form Do not repeat data values needlessly. Be wary of using codes, such as exp for experimental data values. But do not be afraid to violate third normal form at first and correct it when necessary. The Appendix shows a possible method for correcting violations of third normal form that might be encountered when importing data from another source, or for correcting violations that have crept into data tables over the lifetime of the database. 

A relational model of data for large shared data banks. Communications of the ACM, 13(6) 377—387. 

---