I might be mistaken, but it looks like Scribble depends very much on
source location info for formatting this -- it looks like even a form
read as "(a b)" will format as "(ab)" in some cases if the source
location info is stripped.
If this is true, is my best bet to feed the generated syntax through the
pretty-printer (with a sensible max. width), and read the output of that
so that the location info is Scribble-friendly?
Neil Van Dyke wrote at 05/01/2012 02:57 AM:
Some more info...
I see that I am indeed stomping on the syntax location info for the
"(or/c soundex-ordinal? #f)" syntax object, which I assume is a factor
in it being formatted badly. I still have questions.
First, I made a procedure to do a detailed dump of the syntax object:
(define (debug-dump-syntax-object stx-or-pair)
(cond ((pair? stx-or-pair)
`(*NONSYNTAX-PAIR* ,(debug-dump-syntax-object (car stx-or-pair))
,(debug-dump-syntax-object (cdr
stx-or-pair))))
((null? stx-or-pair)
`(*NONSYNTAX-NULL*))
(else
`(*SYNTAX* :SRC ,(syntax-source stx-or-pair)
:L ,(syntax-line stx-or-pair)
:C ,(syntax-column stx-or-pair)
:P ,(syntax-position stx-or-pair)
:S ,(syntax-span stx-or-pair)
:O ,(syntax-original? stx-or-pair)
,(let ((stx-e (syntax-e stx-or-pair)))
(cond ((pair? stx-e)
`(*SYNTAX-PAIR*
,(debug-dump-syntax-object (car stx-e))
,(debug-dump-syntax-object (cdr stx-e))))
((null? stx-e)
`(*SYNTAX-NULL*))
(else `(*ATOMIC* ,stx-e))))))))
So, the generated syntax object that looks pretty-printed like this:
(doc scribble (defproc (soundex-ordinal (chr char?))
(or/c soundex-ordinal? #f)
"Yields the Soundex ordinal value of character "
(racket chr)
"!!!"))
...has a debugging dump like this:
(*SYNTAX* :SRC "myfile" :L 56 :C 0 :P 2065 :S 648 :O #f
(*SYNTAX-PAIR* (*SYNTAX* :SRC "myfile" :L 56 :C 1 :P 2066 :S
3 :O #t
(*ATOMIC* doc))
(*NONSYNTAX-PAIR*
(*SYNTAX* :SRC #f :L #f :C #f :P #f :S 0 :O #f
(*ATOMIC* scribble))
(*NONSYNTAX-PAIR*
(*SYNTAX* :SRC #f :L #f :C #f :P #f :S 0 :O #f
(*SYNTAX-PAIR*
(*SYNTAX* :SRC #f :L #f :C #f :P
#f :S 0 :O #f
(*ATOMIC* defproc))
(*NONSYNTAX-PAIR*
(*SYNTAX* :SRC #f :L #f :C #f
:P #f :S 0 :O #f
(*SYNTAX-PAIR*
(*SYNTAX* :SRC
"myfile" :L 75 :C 9 :P 2798 :S 15 :O #t
(*ATOMIC*
soundex-ordinal))
(*SYNTAX* :SRC #f :L
#f :C #f :P #f :S 0 :O #f
(*SYNTAX-PAIR*
(*SYNTAX*
:SRC #f :L #f :C #f :P #f :S 0 :O #f
(*SYNTAX-PAIR*
(*SYNTAX*
:SRC "myfile" :L 75 :C 25 :P 2814 :S 3 :O #t
(*ATOMIC* chr))
(*NONSYNTAX-PAIR*
(*SYNTAX* :SRC "myfile" :L 74 :C 39 :P 2753 :S 5 :O #t
(*ATOMIC* char?))
(*NONSYNTAX-NULL*))))
(*NONSYNTAX-NULL*)))))
(*NONSYNTAX-PAIR*
(*SYNTAX* :SRC #f :L #f :C #f
:P #f :S 0 :O #f
(*SYNTAX-PAIR*
(*SYNTAX* :SRC
"myfile" :L 74 :C 46 :P 2760 :S 4 :O #t
(*ATOMIC*
or/c))
(*NONSYNTAX-PAIR*
(*SYNTAX* :SRC
"myfile" :L 74 :C 51 :P 2765 :S 16 :O #t
(*ATOMIC* soundex-ordinal?))
(*NONSYNTAX-PAIR*
(*SYNTAX* :SRC
"myfile" :L 74 :C 68 :P 2782 :S 2 :O #t
(*ATOMIC* #f))
(*NONSYNTAX-NULL*)))))
(*SYNTAX* :SRC #f :L #f :C #f
:P #f :S 0 :O #f
(*SYNTAX-PAIR*
(*SYNTAX* :SRC
"myfile" :L 57 :C 5 :P 2101 :S 48 :O #t
(*ATOMIC*
"Yields the Soundex ordinal value of character "))
(*NONSYNTAX-PAIR*
(*SYNTAX* :SRC
"myfile" :L 58 :C 5 :P 2155 :S 12 :O #t
(*SYNTAX-PAIR*
(*SYNTAX* :SRC "myfile" :L 58 :C 6 :P 2156 :S 6 :O #t
(*ATOMIC* racket))
(*NONSYNTAX-PAIR*
(*SYNTAX* :SRC "myfile" :L 58 :C 13 :P 2163 :S 3 :O #t
(*ATOMIC* chr))
(*NONSYNTAX-NULL*))))
(*NONSYNTAX-PAIR*
(*SYNTAX* :SRC
"myfile" :L 59 :C 5 :P 2173 :S 5 :O #t
(*ATOMIC* "!!!"))
(*NONSYNTAX-NULL*)))))))))
(*NONSYNTAX-NULL*)))))
If you search for "or/c" in the debug dump, you can see that its
parent "(*SYNTAX* (*SYNTAX-PAIR* ...))" has nulled/zero location info,
whereas "or/c" syntax object has valid info. This is presumably my
fault and fixable, since I see no reason to replace the location info
in this case. But even if I fix this one, I suspect I will have
similar situations in the code that cannot be fixed, and that I
imagine might mess up formatting similarly.
Also, I have both "*SYNTAX-PAIR*" and "*NONSYNTAX-PAIR*" in the dump.
I can change all the latter kind to the former kind easily enough, if
that actually matters.
Or perhaps I'm smoking crack, and Scribble is not as sensitive to
source positions as it seems to me at the moment.
Neil Van Dyke wrote at 05/01/2012 01:51 AM:
Can someone who knows Scribble internals point me in the right
direction on this?
I have a generated "defproc" syntax object that gets formatted badly
by Scribble, as shown in screenshot: http://i.imgur.com/FmdQr.png
The "defproc" syntax object is generated in this case by piecing
together information from three syntax objects read from a source file:
(doc procedure soundex-ordinal
"Yields the Soundex ordinal value of character "
(racket chr)
[...])
(provide/contract (soundex-ordinal (-> char? (or/c soundex-ordinal?
#f))))
(define (soundex-ordinal chr)
[...])
Note that the other "defproc" in the screenshot is formatted fine.
That "defproc" result of that one is an identifier, rather than a
parenthesized form.
The "syntax->datum" looks fine, and formats correctly if pasted into
a file and processed from there:
@(defproc (soundex-ordinal (chr char?))
(or/c soundex-ordinal? #f)
"Yields the Soundex ordinal value of character "
(racket chr)
[...])
I'm wondering whether the problem is the combinations of syntax
location info in the generated "defproc" syntax object. I try to
preserve the location info of the individual pieces, and often use
"quasisyntax/loc". Also, the "(or/c soundex-ordinal? #f)" syntax
object should be intact just as read originally, including its
parentheses.
Neil V.
____________________
Racket Users list:
http://lists.racket-lang.org/users
____________________
Racket Users list:
http://lists.racket-lang.org/users
