If I load the reader into SWI-Prolog, and if I run also the fuzzer, I can
compute some defect rate. I get that SWI-Prolog has a half as high
error rate than the pPEG parser from 14.04.2022:
?- aggregate_all(count, (between(1,10000,_), random_expr(0, _, A),
ppeg_parse(A, X), simp_parse(A, Y), X \== Y), C).
C = 583.
?- aggregate_all(count, (between(1,10000,_), random_expr(0, _, A),
swi_parse(A, X), simp_parse(A, Y), X \== Y), C).
C = 273.
The defect rates are ca. 6% for pPEG 14.04.2022 and ca. 3% for SWI-Prolog.