On Thu, 29 Mar 2007 11:14:04 -0700, "Mac Chan" <emailmac@gmail.com> wrote:

The following code

(defparameter *checked* nil)

(with-html-output (#:out nil :prologue nil :indent nil) (:input :type "checkbox" :checked *checked*))

will get expanded to

=>

(let ((#:out #:out)) (progn nil (write-string "<input type=\"checkbox\"" #:out) (let ((#:g2546 *checked*)) (cond ((null #:g2546)) ((and (eq #:g2546 t) (eq cl-who::*html-mode* :xml)) (progn (write-string " checked=\"checked\"" #:out))) ((and (eq #:g2546 t) (eq cl-who::*html-mode* :sgml)) (progn (write-string " checked" #:out))) (t (progn (write-string " checked=\"" #:out) (princ #:g2546 #:out) (write-string "\"" #:out))))) (write-string " />" #:out)))

My question is, should the macro generate code based on the value of *html-mode* at compile/load time instead of runtime?

If there are a lot of attributes the generated code will be bloated and I can't imagine a situation where you want to apply the same cl-who syntax tree with different *html-mode* at runtime.

If there's a rational behind this I'd like to learn about it.

I don't remember the details, but I guess this was just an oversight of Stefan when he sent the HTML-MODE patch in 2005. Feel free to send a patch to change this behaviour if you think it's worth it. Thanks, Edi.