Browse files

Merge pull request #1647 from gilch/let-rebind

Fix let rebind bug.
  • Loading branch information...
gilch committed Jul 9, 2018
2 parents 71dfec9 + 8c79015 commit 0de8557fb5d7a2a18b90164d83db53ccf67faa57
Showing with 160 additions and 198 deletions.
  1. +8 −55 hy/contrib/walk.hy
  2. +152 −143 tests/native_tests/contrib/walk.hy
@@ -305,6 +305,7 @@ as can nested let forms.
(macro-error bindings "let bindings must be paired"))
(setv g!let (gensym 'let)
replacements (OrderedDict)
keys []
values [])
(defn expander [symbol]
(.get replacements symbol symbol))
@@ -313,63 +314,15 @@ as can nested let forms.
(macro-error k "bind targets must be symbols")
(if (in '. k)
(macro-error k "binding target may not contain a dot")))
(.append values (symbolexpand (macroexpand-all v &name) expander))
(assoc replacements k `(get ~g!let ~(name k))))
(.append values (symbolexpand (macroexpand-all v &name)
(.append keys `(get ~g!let ~(name k)))
(assoc replacements k (last keys)))
(setv ~g!let {}
~@(interleave (.values replacements) values))
~@(symbolexpand (macroexpand-all body &name) expander)))
~@(interleave keys values))
~@(symbolexpand (macroexpand-all body &name)
;; (defmacro macrolet [])
#_[special cases for let
;; Symbols containing a dot should be converted to this form.
;; attrs should not get expanded,
;; but [] lookups should.
;;; can shadow let bindings with Python locals
;; protect its bindings for the lexical scope of its body.
;; protect as bindings for the lexical scope of its body
;;; changes scope of named variables
;; protect the variables they name for the lexical scope of their container
;; should we provide a protect form?
;; it's an anaphor only valid in a `let` body.
;; this would make the named variables python-scoped in its body
;; expands to a do
;;; quoted variables must not be expanded.
;; but unprotected, unquoted variables must be.
;;;; deferred
;; should really only exist at toplevel. Ignore until someone complains?
;; raise an error? treat like fn?
;; should probably be implemented as macros in terms of fn/setv anyway.
;;; create Python-scoped variables. It's probably hard to avoid this.
;; Best just doc this behavior for now.
;; we can't avoid clobbering enclosing python scope, unless we use a gensym,
;; but that corrupts '__name__'.
;; It could be set later, but that could mess up metaclasses!
;; Should the class name update let variables too?
;; should this update let variables?
;; it could be done with gensym/setv.
;; I don't understand these. Ignore until someone complains?
'eval_and_compile', 'eval_when_compile', 'require',]
Oops, something went wrong.

0 comments on commit 0de8557

Please sign in to comment.