Logically, each MapSet node has a map of values to a subset. However, it is faster to represent that map as an array of key-value pairs rather than a D associative array, so we do that here.
See Implementation
Logically, each MapSet node has a map of values to a subset. However, it is faster to represent that map as an array of key-value pairs rather than a D associative array, so we do that here.