[racket] Randomized bisimulation tests and Metropolis-Hastings random walk
On 01/06/2013 08:38 AM, Sam Tobin-Hochstadt wrote:
> On Sat, Jan 5, 2013 at 7:31 PM, Neil Toronto <neil.toronto at gmail.com> wrote:
>> Last time I checked Hari's RAList implementation, it was broken, so I
>> started my own from scratch (which is what I worked from just now). He's
>> fixed it, though. I could run some benchmarks, but I'm sure mine's a little
>> faster. (I use index types for static guarantees and to get TR to optimize
>> fixnum ops.) His has more functions, though...
>
> Is the use of index types a fundamental refactoring, or something that
> could be added to Hari's code?
No, not fundamental. Their use isn't always straightforward, though.
(The tricky changes are things like (<= i n) to (let ([new-i (- i n)])
(<= new-i 0)).) I'll send a patch or a pull request.
> I ask because having 3+ different functional random-access data
> structures seems like a loss for Racket code interoperability over
> all.
Duly noted.
Neil ⊥