A high cost in stack_step is not entirely unsurprising, but if it is taking
too long subjectively then there are some points that might be fixable.
Note that stack_step() implements the semantics of your parser, which are
only run *after* a successful parse. I don't think using events in the
http://ashishagarwal.org/2011/10/04/pdf-type-theory/
Basically, the author describes a DSL for describing statistical
distributions, and programming techniques for turning those
descriptions into random number generators.
Hmm, this is *interesting*. I look forward to a prototype of
In the last few days I've thrown together a prototype for Yet Another
DSL-frontend for Marpa. The design problems are solved quite differently
from the SLIF, as neither performance nor the quality of error reports were
deemed important for this prototype. The DSL is therefore happy to
