Unification of bottom and top.

At the end of a derivation the bottom- and top-parts of a node are to be unified. The relation $\mbox{\it unify\_nodes}$ is true of trees of which the nodes have identical bottom- and top values. Note that this predicate essentially implements a non-monotonic device, and hence cannot be straightforwardly defined in a grammar of $\cal {R}$($\cal {L}$); on the other hand, it is quite easy to augment the meta-interpreter to implement this device, as part of the definition of $\mbox{\it start\_parse}$. Thus, to ensure that the nodes of the trees built by the parser have identical bottom- and top parts, the relation `unify_nodes' could be added to the definition of $\mbox{\it start\_parse}$.

However, once it is known that at some node, no further adjunctions are possible, then we might just as well unify the bottom- and top parts of that node immediately. This may be useful, in cases where this unification fails -- in that case we may abandon a search path without a solution much earlier. Thus, it is possible to implement the $\mbox{\it unify\_nodes}$ predicate in an incremental fashion.

The incremental unification of top and bottom features is easily implemented. Firstly, the predicate $\mbox{\it unify\_nodes}$ is defined in such a way that completed nodes are not penetrated: these nodes are already unified in a previous cycle. Furthermore, this predicate is now called when partial trees are completed; that is, after a substitution, and after an adjunction.

The predicate $\mbox{\it unify\_nodes}$ is defined as follows.

The predicate $\mbox{\it head\_corner}$ is changed as follows: