Hi all,
this is a repetition of Romano's and my findings with a couple of notes. If
I want to define a Rebol function able to return any Rebol value, I can do
it as follows:
attempt-1: func [
value [any-type!]
] [
get/any 'value
]
The ATTEMPT-1 function uses RETURN-less method of returning a value. This
test reveals its weakness:
type? attempt-1 make error! ""
** User Error:
** Near: type? attempt-1 make error! ""
instead of getting the error! datatype as a result of the expression, the
error is being fired. My second attempt tries to use RETURN to achieve the
goal:
attempt-2: func [
value [any-type!]
] [
return get/any 'value
]
ATTEMPT-2 behaves better for errors, but it has another weakness:
type? attempt-2 first [a:] ; == word!
, while we want to get a SET-WORD! Here's another attempt:
attempt-3: func [
value [any-type!]
] [
either error? get/any 'value [
return value
] [get/any 'value]
]
This seems to work as it should (any testing appreciated). In my opinion all
three attempts should be equivalent and both ATTEMPT-1 and ATTEMPT-2 show a
Rebol design flaw.
WARNING! Alhough I tried to write e.g. my %highfun.r to be as general as
possible, It doesn't use the ATTEMPT-3 method! Therefore my %highfun.r is
undoubtedly flawed at least until I rewrite it or RT change the interpreter
behaviour (whatever happens first). Nevertheless, I hesitate to update it,
because the ATTEMPT-3 surely isn't nice.
Ciao
L
P.S. BTW, the last version of http://www.rebolforces.com/~ladislav/highfun.r
uses a changed version of the LOCALS? function (hope it doesn't cause any
inconveniences) and it sports a repaired version of the LFUNC function with
implicitly local set-words.
--
To unsubscribe from this list, please send an email to
[EMAIL PROTECTED] with "unsubscribe" in the
subject, without the quotes.
