[racket] regexp-replace* allows no sub-expressions?

The change was not intentional, and I'll fix it soon.

Thanks for the report!

At Sat, 26 Jun 2010 15:38:49 -0700, synx wrote:
> 
> Okay what. I've long relied on the behavior of regexp-replace* to single
> out a number of substrings within a big string, and let me mangle those
> substrings in a programmatic fashion. Example:
> 
> (regexp-replace*
>   #rx"([a-z]+) ([a-z]+)" "red fox, blue seal. red trout, blue trout!"
>   (lambda (total color what)
>     (cond
>       ((equal? color "red") (string-append what " in socks"))
>       ((equal? what "trout") (string-append color " fish"))
>       (else (string-append color " " what)))))
> 
> => "fox in socks, blue seal. trout in socks, blue fish!"
> 
> Recently every program of mine that did that, they all started erroring
> out on contract errors. I took a look at the code and was shocked to
> find this in racket/private/string.rkt:
> (check
>   replacement
>   (sub buf mstart mend))
> 
> which effectively becomes (replacement (sub buf mstart mend)).
> 
> It gives no access to subexpressions, and only matches on the entire
> pattern. For patterns like #rx"(someheader*)([0-9a-zA-Z]+)(somefooter*)"
> that's especially nasty, since I need those three fields separate, but
> there might not be a particular character sequence uniquely separating them.
> 
> Is there no way to replace all instances of an expression in a string
> with some combination of its sub-expressions? I really can't use the
> "\\1 \\2" stuff since I'm replacing stuff that could either be a hex
> number, a decimal number, or a standard abbreviated name. Was this
> support removed for a reason, or was it just too much of a hassle to
> figure out? Is there some other procedure that does this now?