This is the mail archive of the
kawa@sourceware.org
mailing list for the Kawa project.
new warnings for keywords
- From: Per Bothner <per at bothner dot com>
- To: kawa at sourceware dot org
- Date: Sun, 27 Apr 2014 17:03:30 -0700
- Subject: new warnings for keywords
- Authentication-results: sourceware.org; auth=none
The plan is for the syntax for using keywords in Kawa Scheme to be stricter.
I just checked in some warnings; these may switch to errors later.
First, if you want a literal keyword value, you should quote it;
otherwise you will get a warning.
Secondly, only literal unquoted keywords are recognized before keyword arguments
in function calls. The compiler warns if a literal non-quoted keyword is not
followed by an expression (i.e. something that is not a literal unquoted keyword).
Also, all the keyword-value pairs must be adjacent. The following are warned about:
(f x: y: 2)
(f x: 2 y:)
(f x: 2 3 y: 4)
For now, this is only partially enforced at rewrite time. The following is
allowed for now, but will be disallowed in the future:
(list 1 a: 2 b: 3) ;; bad - instead write: (list 1 'a: 2 'b: 3)
I'm working on a new calling convention (which will support pattern matching ...).
Keyword parameters will only match explicit literal keywords, not expressions that
*happen* to evaluate to keywords. This can allow better performance, and avoids
errors and accidental matches.
Consider Kawa's general object-allocation syntax:
http://www.gnu.org/software/kawa/Allocating-objects.html
Kawa really needs to know which sub-expressions are keywords, and
which are values. Currently Kawa decides that a sub-expression is a "keyword"
if it's a QuoteExp whose value is a KeyWord. This is fragile - if an expression
evaluates to a Keyword, it may behave differently if inlined by the compiler,
and otherwise.
To handle apply-style applications, there will be an "arglist" type, which is like a
list (or vector) but you
This change btw is similar to how Racket handles keywords.
--
--Per Bothner
per@bothner.com http://per.bothner.com/