search

clojure-emacs / parseedn

EDN parser for Emacs Lisp
62 stars 15 forks source link

parseedn-print tag readers #6

Open nivekuil opened 4 years ago

nivekuil commented 4 years ago

For parseedn-read, we can pass an alist of readers to convert e.g. #uuid into a string. Could this be done in the inverse operation, parseedn-print, as well? Currently I've just patched it like so:

   ((consp datum)
    (cond
     ((not (listp (cdr datum))) ; dotted pair
      (error "Don't know how to print: %s" datum))
     ((eq 'edn-set (car datum))
      (insert "#{") (parseedn-print-seq (cadr datum)) (insert "}"))
+     ((eq 'edn-uuid (car datum))
+      (insert "#uuid ") (parseedn-print-seq (cdr datum)))
     (t (insert "(") (parseedn-print-seq datum) (insert ")"))))
bbatsov commented 4 years ago

The suggestion seems reasonable to me. Let's see what @plexus thinks about it as well.

plexus commented 4 years ago

I think the reason we don't have custom print handlers is that elisp doesn't really have custom types. Even the common lisp style types are really just cons cells IIRC so it's hard to distinguish. But if we can offer custom print handlers in a way that makes sense then I'm all for it.

plexus commented 4 years ago

Having another look at this, I think it would be fair to provide an optional second argument to parseedn-print and parseedn-print-str which is an a-list from symbol to printing function, where we dispatch on the first item in a list if it is a symbol. This would only cover the case where "custom" types are represented as a cons with a symbol "tag", but this is the most common way to do so in Elisp and it's what cl-struct does (if I'm not mistaken).

A PR for that would be very welcome.

However for the issue that @nivekuil raised I think the actual problem is that while we parse inst and uuid (since they are part of the EDN spec) we don't print them correctly. I added printing for #inst and #uuid in #7 .