On Tue, May 5, 2015 at 4:36 PM, Dirk Eddelbuettel <e...@debian.org> wrote:

>
> On 5 May 2015 at 15:18, Tim Keitt wrote:
> | On Tue, May 5, 2015 at 9:11 AM, Dirk Eddelbuettel <e...@debian.org>
> wrote:
> |     On 4 May 2015 at 14:42, Tim Keitt wrote:
> |     | Copying is where all the costs accumulate for sure. I find it
> mostly
> |     guesswork
> |     | whether something copies or not with Rcpp data structures.
> |
> |     By now we have said about 1e4 times that Rcpp data structures are
> ("no
> |     copies") whereas other (ie Boost, STL, ...) data structures
> generally need
> |     copies.
> |
> | I have no idea what that means. I just know its very easy for me to
> figure out
> | when things are copied using STL. Not so much with Rcpp.
>
> In the editing process I dropped a 'proxies' -- sorry. Let's try again:
>
>    By now we have said about 1e4 times that Rcpp data structures are proxy
>    objects ("no copies") whereas other (ie Boost, STL, ...) data structures
>    generally need copies.
>

I understand that. Its during insertion, extraction, coercion and function
boundaries (wrap/as) where it gets a bit hard to know when you've done
shallow or deep copy.

THK


>
> Dirk
>
> --
> http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
>



-- 
http://www.keittlab.org/
_______________________________________________
Rcpp-devel mailing list
Rcpp-devel@lists.r-forge.r-project.org
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel

Reply via email to