Direct Addressing Techniques of Table Look-up

Table look-up is the most time-consuming part of many SAS programs. Base SAS offers a rich collection of built-in searching techniques. Merging, SQL joins, formats, SAS indexes all serve the purpose of locating relevant data. For custom programming, SAS offers arrays, whose direct addressability lends itself to implementing just about any searching algorithm. Array-based lookup is not a ready-to-go food; it has to be cooked at home. However, it may result in dishes digested by the computer more easily, more programmatically nutritious, and with fewer computer resources ending up in the garbage disposer. This paper shows how arrays can be used to organize the fastest class of in-memory table look-up -direct-address searching. Three such techniques -key-indexing, bitmapping, hashing -are considered in a logical sequence using a real-life example of matching two data files by a common key. The results of benchmarking presented in the paper show that home-cooked direct addressing methods beat even the quickest readyto-go tools like the "large formats" by a wide margin. As such, they can be indispensable in any massive data processing setting, where speed and efficiency considerations are paramount.