[racket-dev] Falling through cond clauses

No, you recall correctly. We conducted this experiment a long
time ago (w/o inserting else clauses though as far as I recall). 
That is, when the code base was small compared to what we have now. 

I forgot why the issue came up but there was an additional, external 
impetus. 

The cond issue also shows up in TR. We should and will have to tackle it. 

;; --- 

Racket2: we start from TR and move backwards :-) 



On Oct 2, 2012, at 10:43 AM, Robby Findler wrote:

> On Tue, Oct 2, 2012 at 9:39 AM, Carl Eastlund <cce at ccs.neu.edu> wrote:
>> I'm not surprised, I pretty much expected this response, but I'm curious
>> what the difficulty was.  Is cond-as-void relied on that much more
>> pervasively than mutable pairs?
> 
> I'm not sure I can comment on the relative reliance, but in the case
> of mutable pairs, we had a lot at stack (ie many abstractions are
> broken in their presence). That's not the case with the cond/void
> situation.
> 
>> Is it exceedingly hard to add else-clauses
>> to all the necessary conds?  You two, Robby and Matthew, probably have a
>> better idea than I do about the prevailing trends in the Racket code base.
> 
> I don't think we considered writing a script to stick else clauses in.
> We just were surprised at how often it was relied on and decided to
> give up. (But memory is a fragile thing, so I could be getting this
> completely wrong.)
> 
> Robby
> 
>> Carl Eastlund
>> 
>> 
>> On Tue, Oct 2, 2012 at 10:36 AM, Robby Findler <robby at eecs.northwestern.edu>
>> wrote:
>>> 
>>> IIRC, we even experimented with this one briefly and quickly gave up
>>> as we got overwhelmed.
>>> 
>>> Robby
>>> 
>>> On Tue, Oct 2, 2012 at 9:22 AM, Matthew Flatt <mflatt at cs.utah.edu> wrote:
>>>> I think this a great change for a future `racket2' (someone should be
>>>> keeping a list of these), but it's too incompatible a change for `cond'
>>>> in `racket'.
>>>> 
>>>> At Tue, 2 Oct 2012 09:52:38 -0400, Carl Eastlund wrote:
>>>>> I often wish cond would raise an exception if all the tests failed and
>>>>> there were no else clause.  I have taken to writing a macro to enforce
>>>>> this; I usually call it cond!.  The void default for cond seems like an
>>>>> un-Racketish holdover from primarily-imperative programming.  With some
>>>>> of
>>>>> the other changes we've made in Racket, are we willing to consider
>>>>> changing
>>>>> the fall-through behavior of cond?  It seems like an experiment worth
>>>>> running to me.
>>>>> 
>>>>> If not, I would at least like to add an erroring version of cond
>>>>> somewhere
>>>>> in the language.  It's a shame to have to keep writing such a primitive
>>>>> feature.  Right now in my dracula github repo I have cond! implemented
>>>>> in
>>>>> racket/cond and re-exported from racket, but I'm not thrilled about
>>>>> either
>>>>> the location or the name.  I kept it out of racket/base so I could
>>>>> depend
>>>>> on the syntax collection for good source location reporting in the
>>>>> error
>>>>> message.
>>>>> 
>>>>> Carl Eastlund
>>>>> _________________________
>>>>>  Racket Developers list:
>>>>>  http://lists.racket-lang.org/dev
>>>> _________________________
>>>>  Racket Developers list:
>>>>  http://lists.racket-lang.org/dev
>>> 
>> 
> _________________________
>  Racket Developers list:
>  http://lists.racket-lang.org/dev