Algorithms for a hashed file with variable-length records

Variable-length records are very frequent in database environments. This work gives a set of algorithms for maintaining a new model of static hashed files designed to accommodate variable-length records. Two special techniques for variable-length records are used. First, records are ordered by length and not by key value, e.g., the shortest (longest) records are stored in the main (overflow) blocks. Second, at the cost of a small directory for the overflow records stored in the main file blocks, the successful and unsuccessful searches are faster. The directory consists of the triplet key value, record length, and pointer to the overfiow block. In addition, this work reports a simulation of the file structure. Numerical results show the benefits of the model. These two techniques may be applied in any primary key file organization that uses overflow chaining, including the recent versions of dynamic hashing.