Cardinality (data modeling)

The cardinality of a join between two tables is the numerical relationship between rows of one table and rows in the other. Common cardinalities include one-to-one, one-to-many, and many-to-many.

For example, consider a database of electronic health records. Such a database could contain tables like the following:

• A doctor table with information about physicians.
• A patient table for medical subjects undergoing treatment.
• An encounter table with an entry for each hospital visit.

Natural relationships exist between these entities, such as an encounter involving many doctors. There is a many-to-many relationship between records in doctor and records in patient because doctors have many patients and patients can see many doctors. There is a one-to-many relationship between records in patient and records in encounter because patients can have many encounters and each encounter involves only one patient.

A "one-to-one" relationship is mostly used to split a table in two in order to provide information concisely and make it more understandable. In the hospital example, such a relationship could be used to keep apart doctors' own unique professional information from administrative details.

In data modeling, collections of data elements are grouped into "data tables" which contain groups of data field names called "database attributes". Tables are linked by "key fields". A "primary key" assigns a field to its "special order table". For example, the "Doctor Last Name" field might be assigned as a primary key of the Doctor table with all people having same last name organized alphabetically according to the first three letters of their first name. A table can also have a foreign key which indicates that field is linked to the primary key of another table.

A complex data model can involve hundreds of related tables. Computer scientist Edgar F. Codd created a systematic method to decompose and organize relational databases. Codd's steps for organizing database tables and their keys is called database normalization, which avoids certain hidden database design errors (delete anomalies or update anomalies). In real life the process of database normalization ends up breaking tables into a larger number of smaller tables.

In the real world, data modeling is critical because as the data grows voluminous, tables linked by keys must be used to speed up programmed retrieval of data. If a data model is poorly crafted, even a computer applications system with just a million records will give the end-users unacceptable response time delays. For this reason, data modeling is a keystone in the skills needed by a modern software developer.