deep-matching on syntax objects

[gelisam]

pmatch is very convenient for deep-matching on datatypes, here is a version for syntax objects. That should make writing macros more pleasant!

[gelisam]

The following implementation is way more complicated than it needs to be. I wrote it first, and then I wrote syntax-pmatch. I then tried to rewrite define-syntax-rule in terms of syntax-pmatch, but I encountered #80 . So I'll leave it as-is for now.

[gelisam]

One feature I would like to implement next is the foo ... pattern to match on a list of items. But in order to make that useful, we'd need a corrsponding ,(foo ...) pattern in quasiquote, and I've never quite managed to understand how that function works.

[gelisam]

hmm, did I do something dumb performance-wise? A few tests are running much slower now:

define-syntax-rule: was 1.13s, now 7.87s
which-problem: was 2.15s, now 17.42s
double-define: was 1.48s, now 10.63s

[david-christiansen]

Did you ever get some insight into the slowdowns? I suspect that it's a combination of the deep matching taking time to be expanded into large code, combined with said large code running slowly, but it would be interesting to know how much of it is our expander and how much is our interpreter.