PyPltRedex is a subset of Racket-based PLTRedex that compiles to RPython.

First, install Ply - pip3 install ply. I also recommend using venv.

Then, from project's root directory:

python3 -m src your-plt-redex-spec.rkt

and then

python3 rpyout/out.py

make test

• Built-in patterns: number, integer, natural, real, boolean, string, and any patterns and handling of exact matches of terms of mentioned kinds.
• define-language and redex-match forms.
• Arbitrary patterns involving non-terminals and built-in patterns mentioned above.
• Constraint checking (e.g in pattern (e_1 ... e_1 ...) both e_1 bindings are checked for equality).
• match-equals? form (not part of Redex) used for comparing results produced by redex-match against expected matches.
• Matching hole pattern.
• Matching in-hole pattern.
• Plugging terms into terms, including (in-hole term term) and python function calls (',' and ',@'). See plugtest.rkt for testcases.
• define-reduction-relation and apply-reduction-relation forms
• define-metafunction and metafunction application. Metafunction applications are detected statically when processing terms.
• read-from-stdin-and-apply-reduction-relation* form.

From most to least important.

• Term level non-terminal / built-in pattern membership caching. For example, when asking if a given term is e, store e symbol in the set. Thus, next time we try to determine if the term is e, we just look it up in the set.
• More testing.
• Sample languages as examples (and evaluation).
• Ability to trace application of reduction relation and output series of reductions as graphviz digraph.
• Feature freeze.
• Saner error reporting.
• More reasonable copying in in-hole pattern. Edit 25.07.2020 : actually I think current implementation is pretty sane.
• Fix bug in EllipsisMatchModeRewriter. See here. Aside from general PatSequence structural checking logic being incorrect, I also suspect NtClosure computation is not sufficient.
• Compile reduction-relation in smarter way - instead of pattern matching rule-by-rule, find common subpatterns and run for a set of rules with said subpattern matcher only once. Also want to merge multiple in-hole into one. (in-hole V (+ n_1 n_2) and (in-hole V (- n_1 n_2)) --> `(in-hole V [(+ n_1 n_2) (- n_1 n_2)]) to traverse term only once while looking for redexes.
• Add assert-term-throws to test term plugging that is supposed to fail due to wrong ellipsis match counts.

• Glynn Winskell The Formal Semantics of Programming Languages, Operational semantics and principles of induction chapters.
• The Revised Reporton theSyntactic Theories of Sequential Control and State, Matthias Felleisen, Robert Hieb https://www2.ccs.neu.edu/racket/pubs/tcs92-fh.pdf
• Term Rewriting with Traversal Functions MARK G.J.VANDENBRAND and PAUL KLINT and JURGEN J. VINJU (need to look into this one)
• http://www.meta-environment.org/doc/books/extraction-transformation/term-rewriting/term-rewriting.html#section.tr-substitution
• DAG Representation and Optimization of Rewriting, Ke Li https://pdfs.semanticscholar.org/68d5/945e1ac0f5a09397612c9ab774d41e41f0f4.pdf