[Rcpp-devel] fun(Times) with STL

[Douglas Bates]

On Sat, Jun 4, 2011 at 11:26 AM, Douglas Bates <bates at stat.wisc.edu> wrote:
> On Sat, Jun 4, 2011 at 11:08 AM, Silkworth,David J.
> <SILKWODJ at airproducts.com> wrote:
>> I ran into this yesterday as I needed to get the position of the least time
>> value in a small Que of times that I had extracted from a more complex
>> structure.   I ended up with this problem during a separate work-around,
>> which is a story I’d love to tell also.
>>
>> But, to the point.  I explored using STL in areas I had seen no examples
>> previously.
>>
>> Nothing in the Rcpp sources suggested that min_element() or max_element()
>> functions were supported, but I tried anyway.
>>
>> Only by taking note of the many compiler errors I generated was I able to
>> come up with this stable line of code:
>>
>> double* myIterator = std::min_element (TimeQ.begin(), TimeQ.end());
>
> The usual idiom is
>
> double TQmin = *std::min_element(TimeQ.begin(), TimeQ.end());

Sorry.  I didn't read through to the end of your message.  If you want
the index of the minimum element you can use

int min_el_ind = *std::min_element(TimeQ.begin(), TimeQ.end()) - TimeQ.begin();

The beauty of iterators is that they have more flexibility than simple
pointers and they also carry more information, so that the difference
between two iterators is the index of an element in an array-like
structure.

Dirk and Romain recommended the freely-available book "C++ Annotaions"
to me (just google the title to find out where to download it) and
that contains several chapters explaining iterators, STL storage
classes (on which much of the Rcpp class structures are patterned) and
the STL algorithms.  Definitely worth reading if you have the time.

>> Now, in C++ training I flunked pointers and I skipped class for templates.
>> But now I need to know what the heck is myIterator.
>>
>> So, I managed to put it into a single element NumericVector which I could
>> return to R and examine.
>>
>> Okay, so this mysterious “iterator” thing is quite intuitively the expected
>> result of a min_element function (with some pointer witchcraft included).
>>
>> To get the position of this item in the TimeQ I ended up building a small
>> loop.
>>
>> for(int col=0; col<TimeQ.size(); col++)  {
>>
>> if(TimeQ[col] == *myIterator) {
>>
>> show_position[0]=col;
>>
>> break;  }   }
>>
>> My question is, “Is there a more elegant way to get the position value that
>> I need”?    This code will be traversed 10’s of thousands of times in the
>> function I am developing.
>>
>> Here is the full example as I have distilled it down:
>>
>> src <- '
>>
>> Rcpp::NumericVector TimeQ(arg1);
>>
>> Rcpp::NumericVector show_iterator(1);
>>
>> Rcpp::IntegerVector show_position(1);
>>
>>
>>
>> double* myIterator = std::min_element (TimeQ.begin(), TimeQ.end());
>>
>>
>> show_iterator[0] = *myIterator;
>>
>>
>>
>> for(int col=0; col<TimeQ.size(); col++)  {
>>
>> if(TimeQ[col] == *myIterator) {
>>
>> show_position[0]=col;
>>
>> break;  }   }
>>
>>
>>
>> return show_position;   // alternatively: return show_iterator;
>>
>> '
>>
>>  fun <- cxxfunction(signature(arg1="numeric"),src,plugin="Rcpp")
>>
>>
>>
>>
>>
>>
>> Times<-c(1944.285,2920.969,1720.230,1264.438,3607.507,1720.230,25176.020);
>>
>>
>> fun_test<- fun(Times)
>>
>> _______________________________________________
>> Rcpp-devel mailing list
>> Rcpp-devel at lists.r-forge.r-project.org
>> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
>>
>>
>