Re: Functional record "setters", a different approach

[Mark H Weaver]

Hi Ludovic!

address@hidden (Ludovic Courtès) writes:
> At the time you were concerned about the “weight” of these macros.  Are
> you just more relaxed now, or do you have a psyntax optimization in the
> pipeline?  :-)

I did some experiments, and on my 64-bit system this patch increases the
bytecode associated with a record type definition by about 2755 bytes
plus 852 bytes per field.  I can live with that.

>> +      (and (equal? (set-fields p
>> +                     ((person-email) "address@hidden")
>> +                     ((person-address address-country) "Spain")
>> +                     ((person-address address-city) "Barcelona"))
>
> The choice of towns seems inaccurate.  ;-)

Heh, good point.  I changed "Spain" to "Catalonia" :)

>> +    (pass-if "set-fields with one path as a prefix of another"
>> +      (let ()
>> +        (define-immutable-record-type foo (make-foo x) foo?
>> +          (x foo-x)
>> +          (y foo-y set-foo-y)
>> +          (z foo-z set-foo-z))
>> +
>> +        (define-immutable-record-type :bar (make-bar i j) bar?
>> +          (i bar-i)
>> +          (j bar-j set-bar-j))
>> +
>> +        (catch 'syntax-error
>> +         (lambda ()
>> +           (compile '(let ((s (make-bar (make-foo 5) 2)))
>> +                       (set-fields s
>> +                         ((bar-i foo-x) 1)
>> +                         ((bar-i foo-z) 2)
>> +                         ((bar-i) 3)))
>> +                    #:env (current-module))
>> +           #f)
>> +         (lambda (key whom what src form subform)
>> +           (equal? (list key whom what form subform)
>> +                   '(syntax-error set-fields
>> +                                  "one field path is a prefix of another"
>> +                                  (set-fields s
>> +                                    ((bar-i foo-x) 1)
>> +                                    ((bar-i foo-z) 2)
>> +                                    ((bar-i) 3))
>> +                                  (bar-i)))))))))
>
> You might want to use ‘pass-if-equal’ here, for better reporting.

'pass-if-equal' seems inapplicable here, since I'm testing for an
exception.  'pass-if-exception' was closer to what I needed, but I
wanted to verify more than just the exception key.

> Please commit, I’ll take care of the doc.

Done, and thanks :)

     Mark