Unsorted-array-based maps are sometimes used in the Java world, and for two or three elements will have much less overhead than hash tables. For instance, fastutil has http://fastutil.di.unimi.it/docs/it/unimi/dsi/fastutil/objec.... The map interface and encapsulation into a single “object” is the same.

It occurs to me that I don't know whether any of the major dynamic language implementations with maps/dicts/hashes as a central data structure use a similar approach for very small ones… huh.