On 7/28/23 5:52 PM, Bradley Lucier wrote:
> I have come to regret the choice that initially, to hypothetically speed
> array operations, the parameter specialized-array-default-safe? is not
> #t, which would guarantee run-time checks of array getter and setter
> arguments.  It's not a good look when by default, a library crashes the
> Scheme system when the programmer makes a mistake.
>
> The SRFI document implies, but does not explicitly say, those checks are
> not in place when an array is created and
> specialized-array-default-safe? is #f.
>
> I'm going to do some performance tests with Gambit's implementation of
> SRFI 231, which tracks the sample implementation.  If the performance
> hit is not large, I may set Gambit's implementation to start with safe
> arrays by default.
>
> In a Gambit PR Marc Feeley commented "I would have created 2 variants of
> the SRFI: (srfi 231) and (srfi 231 unsafe) so that ... by default you
> get the type checks ..."
>
> An interesting idea that would imply a compile-time choice of safe or
> unsafe arrays instead of a runtime choice at array-creation time.
>
> Brad

To maybe move this discussion in a slightly different direction ...

I am not aware of any SRFI or RNRS that talks about "safety" or about
faster code at the risk of crashing the system in the face of incorrect
code.

In fact, "safety" seems to be outside the realm of what SRFIs or Scheme
standards describe.

If SRFI 231 is the first and only SRFI or RNRS to add "safety" to a
specification, then perhaps that was a mistake.

Brad