Aspects of the invention provide methods for rewriting and computing database queries that include outer join conditions, and for determining columns that satisfy a criteria. According to one aspect, a component relation key is assigned to each unique column in a first table and an aggregate relation key is generated for each record in a second table, wherein the aggregate relation key is based on one or more of the component relation keys corresponding to columns with non-null values from the second table. Distinct aggregate relation keys are determined, and an original query with an outer join is rewritten into at least a number of sub-queries equal to the number of distinct aggregate keys, wherein each sub-query includes a join condition based on each column represented by the respective distinct aggregate keys. Significantly, the combination of sub-queries is logically equivalent to the original query, but does not include outer join conditions. Therefore, a more efficient query is provided.