Data integrity

• The RDBMS checks data (insert, update, delete) to avoid inconsistencies.
  
  
• Several levels of data integrity:
  • NOT NULL: complain if no value given.
  • PRIMARY KEY: values must be unique in the table.
  • FOREIGN KEY: the row referred to must exist.
  • Transactions: a set of operations must all succeed, or all changes in the set are rolled back (COMMIT, ROLLBACK). This is extremely useful for complex databases.
  • Data locking, to prevent others from accessing data being changed.
  
  
• All SQL-compliant systems should make these checks, but not all have implemented this.
  
  
• Used in the right way, can stop mistaken operations from damaging the data.
  
  
• If wrong design, then data integrity constraints may complicate updates (circular foreign keys).
  
  
• These checks require computation: no free lunch!