[R-SIG-Finance] Follow up on limit orderbook

[Daniel Cegiełka]

Hi,

It's from R-devel list:

http://www.opensubscriber.com/message/r-devel@r-project.org/14086352.html

"A general goal for the next version of R is to make S4 and S3 play
better together."

regards,
daniel



2010/6/14 Jeff Ryan <jeff.a.ryan at gmail.com>:
> A quick and painful example.  I will craft a better one if I have more
> time...
>
>> bigData <- 1:1e7L  # obviously this isn't "big"
>
>> setClass("bigS4", representation(d="integer", comment="character"))
> [1] "bigS4"
>> s4 <- new("bigS4")
>> system.time(s4 at d <- bigData)
>   user  system elapsed
>  0.046   0.028   0.078
>> system.time(s4 at comment <- "comment")
>   user  system elapsed
>  0.045   0.028   0.074
>>
>
>>
>> s3 <- list(d=NULL, comment="")
>> system.time(s3$d <- bigData)
>   user  system elapsed
>  0.028   0.000   0.028
>> system.time(s3$comment <- "comment")
>   user  system elapsed
>      0       0       0
>
> Point is, the entire S4 object gets copied once you call @, whereas the list
> structure (which can be a class) doesn't.
>
> If you really want to optimize, look at using environments.  Then your
> objects _won't_ ever be copied accidentally.
>
> In the case of S4, I haven't yet found a solid use case where I can justify
> the very real costs (performance and opacity).
>
> A bigger issue is really in the Methods used for S4.  That is constantly
> moving target/mess IMO.  Just 2 days ago John Chambers committed another
> change that requires S3 generics for S4 classes... try keeping up with THAT
> ;-)
>
> w.r.t BioC, I have no idea why or how they employ S4, all I know is that I
> can't afford more copies in R than I already have to endure.
>
> Jeff
>
> On Mon, Jun 14, 2010 at 1:25 PM, Khanh Nguyen <nguyen.h.khanh at gmail.com>wrote:
>
>> Hi,
>>
>> Thank you for your suggestions! We are hard at work :D but wanted to
>> follow up on a few points. In particular S3 vs S4:
>>
>> On 09/06/2010 21:00, Jeff Ryan wrote:
>>
>> > Is there a particular reason for using S3 over S4?  I realize this isn't
>> > on-topic per se, but it will affect other packages ability to use (and
>> use
>> > in general, as interoperability is the name of the game these days)
>> >
>> > S4 really isn't excellent for much at present, in my biased though rather
>> > lengthy experience in this area.
>> >
>> > Aside from general obfuscation and bugginess to S4 itself, the only other
>> > thing it can do that S3 can't do directly is dispatch on multiple args
>> and
>> > provide some sort of protection with respect to types per slot.
>> >
>> > If we restrict to a paradigm where objects only are altered with accessor
>> > functions (recommended), then all the type checking can be handled there.
>> >  Multiple args as far as I can tell don't/won't come into play here
>> > though.
>> >  At least not in a way that would warrant all the "bad" of S4.
>> >
>> > Additionally S4 is not overly friendly to large objects, as any slot
>> > manipulation does a full object copy.  Order book is large almost by
>> > definition.
>>
>> On Thu, Jun 10, 2010 at 4:43 AM, Patrick Burns <patrick at burns-stat.com>
>> wrote:
>>
>> > But I'm confused.  The Bioconductor code is all
>> > S4 and the genomics data they deal with is
>> > usually huge.
>>
>> We too are confused! Our datasets will typically be ~1-2 gigabytes,
>> and we would greatly appreciate any comments to help us decide between
>> S3 and S4 for analyzing datasets of that size. Additionally,
>> suggestions as to the best way to read in large amounts of data would
>> be appreciated.
>>
>> Thanks,
>>
>> Khanh & Andrew.
>>
>> _______________________________________________
>> R-SIG-Finance at stat.math.ethz.ch mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
>> -- Subscriber-posting only. If you want to post, subscribe first.
>> -- Also note that this is not the r-help list where general R questions
>> should go.
>>
>
>
>
> --
> Jeffrey Ryan
> jeffrey.ryan at insightalgo.com
>
> ia: insight algorithmics
> www.insightalgo.com
>
>        [[alternative HTML version deleted]]
>
> _______________________________________________
> R-SIG-Finance at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
> -- Subscriber-posting only. If you want to post, subscribe first.
> -- Also note that this is not the r-help list where general R questions should go.
>