Notes

If you put grammar clauses which match something or nothing at the beginning of a rule, e.g.: 

ws = ([ ]+ | Nothing);
assignment = ws [a-z]+ ws '=' ws [0-9+];

Then given a string like

"  x = 5;"

will have a match at position 0, and then again at the 'x'. This is inefficient, because it doubles the number of memo entries per assignment.

=> solution: put optional whitespace consuming matches at the end of the rule 

...and actually that makes a lot of sense, since PEG parsing is about greedily consuming input from a fixed start position.

In fact any rule (X | Nothing) will always match in every position -- the question is just how much input is matched.

--

"Nothing" (or anything that can match 0 characters, e.g. (X | Nothing)) is inefficient if used in the first position, since it seeds a lot of parent clause evaluations.

It's more efficient to rewrite, e.g.:

    (X | Nothing) Y  =>  (X Y) | Y