[racket-dev] provide expand-clause in racket/private/for.rkt

On Fri, Sep 6, 2013 at 4:53 PM, Matthew Flatt <mflatt at cs.utah.edu> wrote:
> Now that you've pushed this, it occurs to me why it's probably unsafe:
> some `in-...' forms expand to unsafe operations, right?

Yes that is true. At least in-vector does and probably others.


> If I'm right about that, then we should probably move the function to
> `racket/unsafe/for-transform' and protect the export.
>
> (Or, since I see that you put the library in `syntax/for-transform',
> maybe `syntax/unsafe/for-transform'?)

I put it in syntax/ because I saw that's where split-for-body is.




>
> At Fri, 6 Sep 2013 14:44:20 -0400, Stephen Chang wrote:
>> On Fri, Sep 6, 2013 at 7:54 AM, Matthew Flatt <mflatt at cs.utah.edu> wrote:
>> > As far as I can see, exporting `expand-clause` is ok.
>> >
>> > Ideally, I think it should be exported from a new
>> > `racket/for-transform` library, instead of used directly from
>> > `racket/private/for`. Also, `expand-for-clause` might be a better name.
>>
>> Ok I will add it. Thanks.
>>
>>
>> >
>> > At Fri, 6 Sep 2013 00:45:40 -0400, Stephen Chang wrote:
>> >> Hi dev,
>> >>
>> >> I would like to provide (for-syntax) the "expand-clause" function in
>> >> racket/private/for.rkt. Would this cause any problems? Would anyone
>> >> object to this?
>> >>
>> >> I have an implementation of for/X in my generic-bind library that uses
>> >> expand-clause and with it, the generic-bind ~for forms are as fast,
>> >> sometimes a little faster, than racket's for/X in some preliminary
>> >> testing. (Without access to expand-clause, sequence traversal is slow
>> >> --- the current planet-available version of my library uses
>> >> sequence-generate and is twice as slow.)
>> >>
>> >> The "sometimes a little faster" is likely in part due to less
>> >> error-checking, but my implementation passes all the for/X unit tests
>> >> and so as is, it can reasonably be used in place of for/X.
>> >> _________________________
>> >>   Racket Developers list:
>> >>   http://lists.racket-lang.org/dev
>> >