On 11/26/2013 11:52 AM, Kevin Wortman wrote:
>> The only use case for 'valid-sre?' mentioned so far is Peter's
>> interactive regular expression IDE. The C standard does not require the
>> -fsyntax-only option. These do not seem like compelling arguments for
>> including
>> 'valid-sre?'.
> There is a well-established Scheme idiom that every client-visible data
> type have a constructor, and a type predicate such as valid-sre? . In
> general user-created data types don't have explicit names in the source
> code, so the type predicate may stand for the type itself in
> documentation and comments. Speaking for myself, when learning about a
> new library the first thing I do is read about all the type predicates
> since they enumerate the relevant data structures and roughly how they
> interrelate.
>
> In terms of practicalities, sanity-checking predicates such as
> valid-sre? are helpful for error checking and interactive debugging.
> They are also necessary for specifying contracts on procedures. See e.g.
> http://docs.racket-lang.org/reference/contracts.html
> which depends upon every data type having a type predicate.
>
> Kevin Wortman
>

An <sre> is not a data type. It is a language. This is like having
'valid-syntax?' to check an expression before you call 'eval' with it.