Subject: Re: Vancouver Lisp Users Group meeting for November 2008 - 0x20 Years of Lisp Systems, a Personal Journey
From: (Rob Warnock)
Date: Sat, 22 Nov 2008 08:03:59 -0600
Newsgroups: comp.lang.lisp
Message-ID: <>
Kaz Kylheku  <> wrote:
| ["Followup-To:" header set to comp.lang.lisp.]
| On 2008-11-20, Rob Warnock <> wrote:
| >     (let* ((fmt1 "~~~:[~;:@~](~:[0x~;~]~~~:[8~;~:*~a~],'~:[0~;~:*~a~]x~~)")
| > 	   (fmt2 (format nil fmt1 colon-p at-sign-p mincol padchar)))
| >       (format stream fmt2 arg)))
| You might be interested in an S-exp to format string translator that I
| developed.
| Your string is a bit hard to understand; it took a while to reverse
| engineer it and render it to my syntax, but here goes: ...
| The output string is:
| "~~~:[~;@~](~:[0x~;~]~~~:[8~;~:*~A~],'~:[0~;~:*~A~]x~~)"
|        ^
|        \ here is a small difference from yours.
| You used a :; separator in conjunction with ~:[  which is useless.
| The :; separator is used to indicate a default case in enumerations.
| Booleans don't have a default case; it's just true and false.
| You should probaby avoid this construction in the interest of portability;
| the CLHS doesn't document the combination of ~:[ and ;:.

Please look again!! *VERY CLOSELY!!* Here, I'll extract the relevant bit:


Please notice that I didn't *use* a "~:;" separator!! I used a "~;"
separator, exactly what is supposed to be used for boolean conditionals.
The ":" that is confusing you is a literal ":" *after* the "~;" that
is emitted (along with an "@" character) when the conditional is true.

For others trying to follow along here, note that the alternate
(false branch) is empty. Only the consequent (true branch) outputs
anything here, the string ":@". The purpose of this bit is to pass
the COLON-P argument to |0X| ["capitalize the output"] on to the
generated "~(" directive, that is, emit "~:@(" if COLON-P is true
and plain "~(" if not. I suppose could have replaced the whole ugly
"~:[~;:@~]" conditional FORMAT directive above with simply "~a", by
replacing the COLON-P argument in the first following FORMAT with:

    (if colon-p "~:@(" "~(")

The ugly conditional FORMAT directive just pushes that IF down into
the FMT1 string.


p.s. Yes, I'll freely admit that the format string in |0X| is
"hard to understand" (Kaz was being kind, actually!) -- it was
certainly hard for me to write (clearly "write once/read never"
code!!). But it only had to be written once, since the |0X| function
is buried deep in my UTILS package. In fact, I haven't ever needed
to read it again in the year and a half since it was first written,
until now.

Rob Warnock			<>
627 26th Avenue			<URL:>
San Mateo, CA 94403		(650)572-2607