Implementation of the fast table grid user interface element for working with large database tables

Table grid user interface element with a vertical scrollbar is a standard way of working with database table records. There are two basic operations each grid should support: scrolling records with vertical scrollbar and positioning to a record with a given primary key. This paper addresses the case when the number of records is so large that it is not feasible to load them all into memory, and database functions like "select..offset" work insufficiently fast and put undue load on RDBMS. Our main idea is to use only queries that involve index lookup (a O(Log(N)-fast operation) and to use statistic properties of hypergeometric distribution to "guess" primary keys of records given their ordinal numbers. The proposed method allows us to implement a grid with O(Log(N))-fast scrolling and positioning performance.