[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Update available-- possibly last before finalization
Andre van Tonder wrote:
On Thu, 9 Dec 2004, Felix Winkelmann wrote:
Hm... Non-generative record definitions would be nice.
Yes, although the particular Chez Scheme specification would be
difficult to implement portably.
One thing I like about the current specification is that it can be
implemented as a macro layer on top of SRFI-9, as the reference
implementation indeed does, without having to manipulate the innards, in
particular the record type descriptors, of SRFI-9. Because of this
philosophy, it can be used on any Scheme implementation that has SRFI-9,
and it can also be easily adapted to Schemes that have their own
efficient native records. Including nongenerativity would make this
impossible, and for this reason is perhaps better left to a future SRFI.
From what my experiments show the current SRFI-57 reference implementation
does *not* work on systems that provide a non-generative SRFI-9, or non-
generative native records.
I haven't understood your code well enough to say whether it's too difficult
to handle generative and non-generatve records, but if you say so I
accept that. It would just be a handy thing to have. Having only generative
records makes it just impossible to use the constructor macros in a
separate compilation model, AFAICT.