functions if possible. The following algorithm does that. (An example

which follows illustrates the algorithm.)

 I. Produce the FVIM. Flag all function indices whose IDM's

 do not precedence order.

 II. A. Break all assignments made by this algorithm.

 B. Choose a set of directions for incrementing the function

 indices not previously analyzed. If there are none,

 go to III.C.

 C. Untag all function and variable types and function

 indices.

 III. A. If the set of unassigned function and variable types does

 not fully precedence order, go to IV.A., otherwise,

 continue.

 B. A full precedence ordering has been discovered. Stop.

 C. Decoupling will not produce a full precedence ordering.

 Stop.

 IV. A. Choose an untagged, unassigned variable type. If there

 are none, go to II.A. Tag the variable type. Untag all

 function indices and unassigned function types.

 B. Choose an untagged, unassigned function type. If there

 are none, go to IV.A. Tag the function type.

 C. If the variable type chosen is not a legal output of the

 function type chosen, go to IV.B.

 V. A. Choose the first untagged function index. If there is

 none, go to IV.A. Otherwise, go to VI.A.

 B. Choose the next untagged function index. If there is

 none, go to IV.B.