[racket-dev] [plt] Push #21079: master branch updated

From: Ryan Culpepper (ryanc at ccs.neu.edu)
Date: Mon Sep 13 17:53:23 EDT 2010

Robby Findler wrote:
> On Mon, Sep 13, 2010 at 4:18 PM,  <ryanc at racket-lang.org> wrote:
>> a3d1ff4 Ryan Culpepper <ryanc at racket-lang.org> 2010-09-09 17:29
>> :
>> | added contracts (->i) to racket/dict
>> :
>>  M collects/racket/dict.rkt   |  190 ++++++++++++++++++++++++++++++++---------
>>  M collects/unstable/dirs.rkt |   13 ++--
>>
> 
> Do you have a benchmark for splay trees or some app that uses them a
> lot that you can use to measure what kind of a performance hit you are
> taking from these contracts?

Not yet, but I should be able to adapt the test into a stress test.

Some hints for anyone else who'd like to try: There are two contract 
boundaries involved in dictionaries now. There's a boundary between the 
dictionary implementation and racket/dict, via prop:dict/contract and 
struct-type-property/c. There'a also a boundary between racket/dict and 
client modules. You can bypass the first boundary by using prop:dict 
instead of prop:dict/contract. You can bypass the second one by 
requiring dict-ref et al from racket/private/dict instead of racket/dict.

Ryan



Posted on the dev mailing list.