[racket] constructing a boolean that's a filter for a given type? could be automatic?

On Mon, Jan 30, 2012 at 6:02 PM, Matthias Felleisen
<matthias at ccs.neu.edu> wrote:
>
> On Jan 30, 2012, at 6:00 PM, John Clements wrote:
>
>>
>> On Jan 30, 2012, at 1:14 PM, Matthias Felleisen wrote:
>>
>>>
>>> On Jan 30, 2012, at 4:11 PM, Sam Tobin-Hochstadt wrote:
>>>
>>>> Perhaps it should just have less information.
>>>
>>>
>>> I like that a lot.
>>
>> I'm not sure I do; if I were on a desert island or had just kicked Sam in the shin yesterday, I'd probably want to examine this type carefully to see whether I could interpret it and whether it actually looked like it should be equivalent to a filter for the type that I wanted.  At a minimum, I'd like to know that the information is available. I realize this is like Guillaume's syntax-errors-for-beginners discussion, but I don't think I reach the same conclusion
>
>
> This is precisely correct. Just as in the general types case, make information available on demand not immediately. Lazy lazy über alles.

Unfortunately, the user interface issues are harder here than in the
large-types-at-the-repl case.  In particular, the REPL *is* a user
interface, and one that TR gets to control.  Arbitrary other
interfaces that might produce this output (such as DrRacket or 'raco
make' or `dynamic-require') don't have an easy way to query for more
information.
-- 
sam th
samth at ccs.neu.edu