[racket] srcloc structures, lists, vectors

The srcloc struct is transparent.  And the syntax/srcloc library helps
with some of these issues -- it contains coercions to each type of
source location (from any of the others), generic accessors for source
location fields, and conversion to strings.

Carl Eastlund

On Wed, Sep 14, 2011 at 2:47 PM, Robby Findler
<robby at eecs.northwestern.edu> wrote:
> I think the other part is historical and an artifact of backwards compatibility.
>
> FWIW, I think the srcloc struct should be transparent or perhaps even
> a prefab, which would negate the benefit below.
>
> Robby
>
> On Wed, Sep 14, 2011 at 1:31 PM, Danny Yoo <dyoo at cs.wpi.edu> wrote:
>> I wanted to double check my understanding on why source locations have
>> several representations in Racket.
>>
>> One of them is the srcloc structure:
>> http://docs.racket-lang.org/reference/exns.html#(def._((lib._racket/private/base..rkt)._srcloc))
>>
>> But there's also the list and vector representations, all which are
>> handled by the syntax/srcloc library in
>> http://docs.racket-lang.org/syntax/Source_Locations.html#(def._((lib._syntax/srcloc..rkt)._source-location~3f))
>>
>> When I first saw the multitude of representations, I was surprised and
>> confused: I expected there to be one canonical representation.
>>
>>
>> I made sense of this to myself with the following reasoning: it's
>> easier to serialize a vector or list than it is a struct.  In many
>> cases, I want to embed source locations somewhere into the output of a
>> macro, and it's a lot easier to do something like this:
>>
>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
>> #lang racket
>> (define-syntax (where stx)
>>  (syntax-case stx ()
>>    [(_)
>>     (with-syntax ([loc (vector (syntax-source stx)
>>                                (syntax-line stx)
>>                                (syntax-column stx)
>>                                (syntax-position stx)
>>                                (syntax-span stx))])
>>       #'loc)]))
>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
>>
>> than it is to do the equivalent with a srcloc structure:
>>
>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
>> #lang racket
>> (define-syntax (where-2 stx)
>>  (syntax-case stx ()
>>    [(_)
>>     (with-syntax ([sloc #`(srcloc '#,(syntax-source stx)
>>                                   '#,(syntax-line stx)
>>                                   '#,(syntax-column stx)
>>                                   '#,(syntax-position stx)
>>                                   '#,(syntax-span stx))])
>>       #'sloc)]))
>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
>>
>> There's several more requirements that the struct approach needs, from
>> the use of the module and macro system to be able to bind "srcloc"
>> hygienically, to remembering to quote all the internal arguments.  So
>> it seems the vector approach is more lightweight.
>>
>>
>> That doesn't quite explain to me why there are both list and vector
>> representations for srcloc, though... In any event, it's something
>> that's been bothering me, because it's not obvious why there are so
>> many representations.