[racket] An error with dates and types
The way I understand the word "ought" comes with a moral connotation.
In this sense, I am saying
(1) require should not exist in TR
(2) require: should exist for importing from typed modules
(3) require-typed should exist for importing from untyped modules.
Why? The idea of explicit specifications is sound/safe *documentation*.
I consider this aspect particularly important at the boundary between
typed and untyped code.
What do we lose if we go this route? As far as I can tell, we lose
the ability to import macros from untyped modules. I consider this
*good* because macros cause problems with error messages. So what
we need is a way to require-typed macros (and I have an idea on that).
[[ I honestly didn't know that we could import stuff from untyped
modules and hope that require could figure out that it types. It
doesn't even work for
(module a racket/base (define x 5) (provide x))
so I am not sure what we could get to work. ]]
-- Matthias
On Aug 21, 2011, at 1:56 PM, Carl Eastlund wrote:
> I still don't follow. There's a shallow error -- 'ought' means the
> same as 'should', presumably you mean to remove 'be able to' instead.
> But there's also a deeper error, as far as I can tell. In general,
> one should use require rather than require/typed so you use a
> binding's originally assigned type rather than assigning a new one.
> One shouldn't use require for untyped bindings, but the problem here
> is that Typed Racket blew up internally before it could report that
> date->seconds is untyped.
>
> Carl Eastlund
>
> On Sun, Aug 21, 2011 at 1:47 PM, Matthias Felleisen
> <matthias at ccs.neu.edu> wrote:
>>
>> I should have used 'ought' instead of 'should'.
>>
>> On Aug 21, 2011, at 1:38 PM, Carl Eastlund wrote:
>>
>>> Plain require works fine in Typed Racket; there's only a problem if TR
>>> cannot typecheck the resulting code. You normally only need to use
>>> require/typed to assign types to otherwise-untyped imports. The error
>>> Norman reported looks like an internal error in TR related to the
>>> contract on date->seconds. Using require/typed is a workaround
>>> because it avoids typechecking contract-wrapped references.
>>>
>>> Carl Eastlund
>>>
>>> On Sun, Aug 21, 2011 at 1:16 PM, Matthias Felleisen
>>> <matthias at ccs.neu.edu> wrote:
>>>>
>>>> You shouldn't be able to use plain require in Typed Racket.
>>>>
>>>> Try this:
>>>>
>>>> #lang typed/racket
>>>> (require/typed racket/date (date->seconds (date -> Natural)))
>>>> (date->seconds (seconds->date (current-seconds)))
>>>>
>>>>
>>>> On Aug 21, 2011, at 11:06 AM, Norman Gray wrote:
>>>>
>>>>>
>>>>> Greetings.
>>>>>
>>>>> Another typed-racket problem, I'm afraid.
>>>>>
>>>>> With the following script, I get an error message from Typed Racket which is not, I think, really addressed to me.
>>>>>
>>>>> (Is the list generally OK for straight bug reports? There isn't a bug parade I should be posting this on, is there?)
>>>>>
>>>>> All the best,
>>>>>
>>>>> Norman
>>>>>
>>>>>
>>>>> % cat dates.rkt
>>>>> #lang typed/racket
>>>>> (require (only-in racket/date date->seconds))
>>>>> (date->seconds (seconds->date (current-seconds)))
>>>>> % "/Data/LocalApplications/Racket/5.1.3/bin/racket" dates.rkt
>>>>> for: expected a sequence for i, got something else: (tc-results (list (tc-result Variable-Reference (#0=Top | #0#) -)) #f)
>>>>>
>>>>> === context ===
>>>>> /Data/LocalApplications/Racket/5.1.3/collects/racket/private/for.rkt:445:2: make-sequence
>>>>> /Data/LocalApplications/Racket/5.1.3/collects/typed-scheme/types/utils.rkt:103:2: ret
>>>>> temp871
>>>>> /Data/LocalApplications/Racket/5.1.3/collects/typed-scheme/typecheck/tc-expr-unit.rkt:201:4: loop
>>>>> /Data/LocalApplications/Racket/5.1.3/collects/racket/private/map.rkt:18:11: map
>>>>> parse-loop1141
>>>>> /Data/LocalApplications/Racket/5.1.3/collects/typed-scheme/typecheck/tc-expr-unit.rkt:330:0: tc-expr
>>>>> /Data/LocalApplications/Racket/5.1.3/collects/racket/private/map.rkt:18:11: map
>>>>> temp971
>>>>> /Data/LocalApplications/Racket/5.1.3/collects/typed-scheme/typecheck/tc-expr-unit.rkt:330:0: tc-expr
>>>>> /Data/LocalApplications/Racket/5.1.3/collects/racket/private/map.rkt:18:11: map
>>>>> temp971
>>>>> /Data/LocalApplications/Racket/5.1.3/collects/typed-scheme/typecheck/tc-expr-unit.rkt:330:0: tc-expr
>>>>> parse-loop26
>>>>> dots-loop
>>>>> /Data/LocalApplications/Racket/5.1.3/collects/racket/private/map.rkt:18:11: map
>>>>> ...
>>>>>
>>>>> %
>>>>>
>>>>>
>>>>> --
>>>>> Norman Gray : http://nxg.me.uk
>>>>> School of Physics and Astronomy, University of Glasgow, UK