hi!
the (for var in list) clause silently skips non-lists. after a quick
look i saw this in the sources:
;;; This is so the advanced user can choose how the end of a list is checked
;;; for.
;;; There are three choices for termination predicate in FOR...ON and
;;; FOR...IN, differing in their behavior on lists with a non-nil cdr:
;;; NULL: If lucky, will get an error when taking the cdr. Bad choice.
;;; ATOM: Will terminate correctly with no error.
;;; ENDP: Will give an appropriate error message.
(defparameter *list-end-test* 'atom)
1) why is 'atom the default and not 'endp? it's not that hard to
accidentally pass in a vector instead of a list... and then it's just
silently skipped. imho 'endp as the default is much more reasonable,
unless i miss something.
2) thinking this further, (for in) could even be smart enough to work
on all kind sequences (possibly specializing for list at runtime to
avoid elt-ing lists) and an in-list/on-list could be introduced for
the current (for in) behaviour? this leads a bit far, though...
any opinions? i think if there's no huge performance penalty then we
should change 1 to use 'endp.
--
- attila
"- The truth is that I've been too considerate, and so became
unintentionally cruel...
- I understand.
- No, you don't understand! We don't speak the same language!"
Ingmar Bergman - Smultronstället (Wild Strawberries)
