Nulls, three-valued logic, and ambiguity in SQL

Abstract

Date's popular critique of SQL's three-valued logic [4, 3] purports to demonstrate that SQL queries can produce erroneous results when nulls are present in the database. I argue that this critique is flawed in that Date misinterprets the meaning of his example query. In fact, SQL returns the correct answer to the query posed; Date, however, believes that he is asking a different question. Although his critique is flawed, I agree with Date's general conclusion: SQL's use of nulls and three-valued logic introduces a startling amount of complexity into seemingly straightforward queries.