Re: [Chicken-users] unterminated string error

[Felix]

Peter Keller wrote:
> Hello,
>
> I was getting an unterminated string error in my code and the error was
> pretty bad in telling me where it happened. So I decided to fix it....
>
> It turned out to be a lot harder than I thought because so much stuff
> (like newlines, etc) can be inside a string.
>
> Example:
>
> (define str "this
> is
> a big
> (valid string).")
>
> The trouble is, what happens if you do this:
>
> (define str "this
> is
> a big
> (valid string).)
>
> (display (+ 2 3))
> (display "\n")
>
> You get an "Error: unterminated string" message. :) I tried to fix up
> this message by telling me the possible real start of the unterminated
> string, but I couldn't do it because basically EVERYTHING is allowed in
> a string until the next " is read all by itself. This means the second "
> in the last display line was being marked as the start of the unterminated
> string since technically, it is. In other languages that usually forbid
> newlines in literal string constants, you can do this pretty easily. But
> how would you do it here? Having to spend a long time debugging an
> unterminated string error in a huge body of code is mind numbing.

Hm. I have no real answer for this, I'm afraid. One could perhaps
optionally disallow multi-line strings... No - that's bad.

> Also, in library.scm: the r-string function, why isn't #\" also taken care
> of in the #\\ case when reading a string? This would leave me to believe
> that in certain situations you cannot read "\"these\"" kinds of strings.
         (define (r-string)
            (if (eq? (read-char port) #\")
                (let loop ((c (read-char port)) (lst '()))
                  (cond ((##core#inline "C_eofp" c)
                         (##sys#read-error "unterminated string") )
                        ((eq? #\\ c)
                         (set! c (read-char port))
                         (case c
                           ((#\t) (set! c #\tab))
                           ((#\r) (set! c #\return))
                           ((#\b) (set! c #\backspace))
                           ((#\n) (set! c #\newline)) )

                        ; Here the case falls through, c == '\"' and
                        ; it's handled as the general case: any char
                        ; that is escaped and not special (as the one
                        ; above) is simply put directly into the string.

                         (loop (read-char port) (cons c lst)) )
                        ((eq? #\" c) (list->string (reverse lst)))
                        ((eq? c #\newline)
                         (advance)
                         (loop (read-char port) (cons c lst)) )
                        (else (loop (read-char port) (cons c lst))) ) )
                (##sys#read-error "missing '\"'") ) )

(A typical example of my usual convoluted coding style ;-)


cheers,
felix