FYI, in R devel (to become 3.5.0), there's isFALSE() which will cut some corners compared to identical():> microbenchmark::microbenchmark(identical(FALSE, FALSE), isFALSE(FALSE))Unit: nanoseconds expr min lq mean median uq max neval identical(FALSE, FALSE) 984 1138 1694.13 1218.0 1337.5 13584 100 isFALSE(FALSE) 713 761 1133.53 809.5 871.5 18619 100> microbenchmark::microbenchmark(identical(TRUE, FALSE), isFALSE(TRUE))Unit: nanoseconds expr min lq mean median uq max neval identical(TRUE, FALSE) 1009 1103.5 2228.20 1170.5 1357 14346 100 isFALSE(TRUE) 718 760.0 1298.98 798.0 898 17782 100> microbenchmark::microbenchmark(identical("array", FALSE), isFALSE("array"))Unit: nanoseconds expr min lq mean median uq max neval identical("array", FALSE) 975 1058.5 1257.95 1119.5 1250.0 9299 100 isFALSE("array") 409 433.5 658.76 446.0 476.5 9383 100 That could probably be used also is sapply(). The difference is that isFALSE() is a bit more liberal than identical(x, FALSE), e.g.> isFALSE(c(a = FALSE))[1] TRUE> identical(c(a = FALSE), FALSE)[1] FALSE Assuming the latter is not an issue, there are 69 places in base R where isFALSE() could be used: $ grep -E "identical[(][^,]+,[ ]*FALSE[)]" -r --include="*.R" | grep -F "/R/" | wc 69 326 5472 and another 59 where isTRUE() can be used: $ grep -E "identical[(][^,]+,[ ]*TRUE[)]" -r --include="*.R" | grep -F "/R/" | wc 59 307 5021 /Henrik On Tue, Mar 13, 2018 at 9:21 AM, Doran, Harold <HDoran at air.org> wrote:> Quite possibly, and I?ll look into that. Aside from the work I was doing, however, I wonder if there is a way such that sapply could avoid the overhead of having to call the identical function to determine the conditional path. > > > > From: William Dunlap [mailto:wdunlap at tibco.com] > Sent: Tuesday, March 13, 2018 12:14 PM > To: Doran, Harold <HDoran at air.org> > Cc: Martin Morgan <martin.morgan at roswellpark.org>; r-help at r-project.org > Subject: Re: [R] Possible Improvement to sapply > > Could your code use vapply instead of sapply? vapply forces you to declare the type and dimensions > of FUN's output and stops if any call to FUN does not match the declaration. It can use much less > memory and time than sapply because it fills in the output array as it goes instead of calling lapply() > and seeing how it could be simplified. > > Bill Dunlap > TIBCO Software > wdunlap tibco.com<http://tibco.com> > > On Tue, Mar 13, 2018 at 7:06 AM, Doran, Harold <HDoran at air.org<mailto:HDoran at air.org>> wrote: > Martin > > In terms of context of the actual problem, sapply is called millions of times because the work involves scoring individual students who took a test. A score for student A is generated and then student B and such and there are millions of students. The psychometric process of scoring students is complex and our code makes use of sapply many times for each student. > > The toy example used length just to illustrate, our actual code doesn't do that. But your point is well taken, there may be a very good counterexample why my proposal doesn't achieve the goal is a generalizable way. > > > > -----Original Message----- > From: Martin Morgan [mailto:martin.morgan at roswellpark.org<mailto:martin.morgan at roswellpark.org>] > Sent: Tuesday, March 13, 2018 9:43 AM > To: Doran, Harold <HDoran at air.org<mailto:HDoran at air.org>>; 'r-help at r-project.org<mailto:r-help at r-project.org>' <r-help at r-project.org<mailto:r-help at r-project.org>> > Subject: Re: [R] Possible Improvement to sapply > > > > On 03/13/2018 09:23 AM, Doran, Harold wrote: >> While working with sapply, the documentation states that the simplify >> argument will yield a vector, matrix etc "when possible". I was >> curious how the code actually defined "as possible" and see this >> within the function >> >> if (!identical(simplify, FALSE) && length(answer)) >> >> This seems superfluous to me, in particular this part: >> >> !identical(simplify, FALSE) >> >> The preceding code could be reduced to >> >> if (simplify && length(answer)) >> >> and it would not need to execute the call to identical in order to trigger the conditional execution, which is known from the user's simplify = TRUE or FALSE inputs. I *think* the extra call to identical is just unnecessary overhead in this instance. >> >> Take for example, the following toy example code and benchmark results and a small modification to sapply: >> >> myList <- list(a = rnorm(100), b = rnorm(100)) >> >> answer <- lapply(X = myList, FUN = length) simplify = TRUE >> >> library(microbenchmark) >> >> mySapply <- function (X, FUN, ..., simplify = TRUE, USE.NAMES = TRUE){ >> FUN <- match.fun(FUN) >> answer <- lapply(X = X, FUN = FUN, ...) >> if (USE.NAMES && is.character(X) && is.null(names(answer))) >> names(answer) <- X >> if (simplify && length(answer)) >> simplify2array(answer, higher = (simplify == "array")) >> else answer >> } >> >> >>> microbenchmark(sapply(myList, length), times = 10000L) >> Unit: microseconds >> expr min lq mean median uq max neval >> sapply(myList, length) 14.156 15.572 16.67603 15.926 16.634 650.46 >> 10000 >>> microbenchmark(mySapply(myList, length), times = 10000L) >> Unit: microseconds >> expr min lq mean median uq max neval >> mySapply(myList, length) 13.095 14.864 16.02964 15.218 15.573 >> 1671.804 10000 >> >> My benchmark timings show a timing improvement with only that small change made and it is seemingly nominal. In my actual work, the sapply function is called millions of times and this additional overhead propagates to some overall additional computing time. >> >> I have done some limited testing on various real data to verify that the objects produced under both variants of the sapply (base R and my modified) yield identical objects when simply is both TRUE or FALSE. >> >> Perhaps someone else sees a counterexample where my proposed fix does not cause for sapply to behave as expected. >> > > Check out ?sapply for possible values of `simplify=` to see why your proposal is not adequate. > > For your example, lengths() is an order of magnitude faster than sapply(., length). This is a example of the advantages of vectorization (single call to an R function implemented in C) versus iteration (`for` loops but also the *apply family calling an R function many times). > vapply() might also be relevant. > > Often performance improvements come from looking one layer up from where the problem occurs and re-thinking the algorithm. Why would one need to call sapply() millions of times, in a situation where this becomes rate-limiting? Can the algorithm be re-implemented to avoid this step? > > Martin Morgan > >> Harold >> >> ______________________________________________ >> R-help at r-project.org<mailto:R-help at r-project.org> mailing list -- To UNSUBSCRIBE and more, see >> https://stat.ethz.ch/mailman/listinfo/r-help >> PLEASE do read the posting guide >> http://www.R-project.org/posting-guide.html >> and provide commented, minimal, self-contained, reproducible code. >> > > This email message may contain legally privileged and/or confidential information. If you are not the intended recipient(s), or the employee or agent responsible for the delivery of this message to the intended recipient(s), you are hereby notified that any disclosure, copying, distribution, or use of this email message is prohibited. If you have received this message in error, please notify the sender immediately by e-mail and delete this email message from your computer. Thank you. > > ______________________________________________ > R-help at r-project.org<mailto:R-help at r-project.org> mailing list -- To UNSUBSCRIBE and more, see > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html > and provide commented, minimal, self-contained, reproducible code. > > > [[alternative HTML version deleted]] > > ______________________________________________ > R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html > and provide commented, minimal, self-contained, reproducible code.
>>>>> Henrik Bengtsson <henrik.bengtsson at gmail.com> >>>>> on Tue, 13 Mar 2018 10:12:55 -0700 writes:> FYI, in R devel (to become 3.5.0), there's isFALSE() which will cut > some corners compared to identical():> > microbenchmark::microbenchmark(identical(FALSE, FALSE), isFALSE(FALSE)) > Unit: nanoseconds > expr min lq mean median uq max neval > identical(FALSE, FALSE) 984 1138 1694.13 1218.0 1337.5 13584 100 > isFALSE(FALSE) 713 761 1133.53 809.5 871.5 18619 100> > microbenchmark::microbenchmark(identical(TRUE, FALSE), isFALSE(TRUE)) > Unit: nanoseconds > expr min lq mean median uq max neval > identical(TRUE, FALSE) 1009 1103.5 2228.20 1170.5 1357 14346 100 > isFALSE(TRUE) 718 760.0 1298.98 798.0 898 17782 100> > microbenchmark::microbenchmark(identical("array", FALSE), isFALSE("array")) > Unit: nanoseconds > expr min lq mean median uq max neval > identical("array", FALSE) 975 1058.5 1257.95 1119.5 1250.0 9299 100 > isFALSE("array") 409 433.5 658.76 446.0 476.5 9383 100Thank you Henrik! The speed of the new isTRUE() and isFALSE() is indeed amazing compared to identical() which was written to be fast itself. Note that the new code goes back to a proposal by Herv? Pag?s (of Bioconductor fame) in a thread with R core in April 2017. The goal of the new code actually *was* to allow call like isTRUE(c(a = TRUE)) to become TRUE rather than improving speed. The new source code is at the end of R/src/library/base/R/identical.R ## NB: is.logical(.) will never dispatch: ## -- base::is.logical(x) <==> typeof(x) == "logical" isTRUE <- function(x) is.logical(x) && length(x) == 1L && !is.na(x) && x isFALSE <- function(x) is.logical(x) && length(x) == 1L && !is.na(x) && !x and one *reason* this is so fast is that all 6 functions which are called are primitives :> sapply(codetools::findGlobals(isTRUE), function(fn) is.primitive(get(fn)))! && == is.logical is.na length TRUE TRUE TRUE TRUE TRUE TRUE and a 2nd reason is probably with the many recent improvements of the byte compiler.> That could probably be used also is sapply(). The difference is that > isFALSE() is a bit more liberal than identical(x, FALSE), e.g.> > isFALSE(c(a = FALSE)) > [1] TRUE > > identical(c(a = FALSE), FALSE) > [1] FALSE> Assuming the latter is not an issue, there are 69 places in base R > where isFALSE() could be used:> $ grep -E "identical[(][^,]+,[ ]*FALSE[)]" -r --include="*.R" | grep -F "/R/" | wc > 69 326 5472> and another 59 where isTRUE() can be used:> $ grep -E "identical[(][^,]+,[ ]*TRUE[)]" -r --include="*.R" | grep -F "/R/" | wc > 59 307 5021Beautiful use of 'grep' -- thank you for those above, as well. It does need a quick manual check, but if I use the above grep from Emacs (via 'M-x grep') or even better via a TAGS table and M-x tags-query-replace I should be able to do the changes pretty quickly... and will start looking into that later today. Interestingly and to my great pleasure, the first part of the 'Subject' of this mailing list thread, "Possible Improvement", *has* become true after all -- -- thanks to Henrik ! Martin Maechler ETH Zurich> On Tue, Mar 13, 2018 at 9:21 AM, Doran, Harold <HDoran at air.org> wrote: > > Quite possibly, and I?ll look into that. Aside from the work I was doing, however, I wonder if there is a way such that sapply could avoid the overhead of having to call the identical function to determine the conditional path. > > > > > > > > From: William Dunlap [mailto:wdunlap at tibco.com] > > Sent: Tuesday, March 13, 2018 12:14 PM > > To: Doran, Harold <HDoran at air.org> > > Cc: Martin Morgan <martin.morgan at roswellpark.org>; r-help at r-project.org > > Subject: Re: [R] Possible Improvement to sapply > > > > Could your code use vapply instead of sapply? vapply forces you to declare the type and dimensions > > of FUN's output and stops if any call to FUN does not match the declaration. It can use much less > > memory and time than sapply because it fills in the output array as it goes instead of calling lapply() > > and seeing how it could be simplified. > > > > Bill Dunlap > > TIBCO Software > > wdunlap tibco.com<http://tibco.com> > > > > On Tue, Mar 13, 2018 at 7:06 AM, Doran, Harold <HDoran at air.org<mailto:HDoran at air.org>> wrote: > > Martin > > > > In terms of context of the actual problem, sapply is called millions of times because the work involves scoring individual students who took a test. A score for student A is generated and then student B and such and there are millions of students. The psychometric process of scoring students is complex and our code makes use of sapply many times for each student. > > > > The toy example used length just to illustrate, our actual code doesn't do that. But your point is well taken, there may be a very good counterexample why my proposal doesn't achieve the goal is a generalizable way. > >[.................]
Well thanks, Martin, and glad to see there is some potential here. This wasn?t reported as a bug, but as you note really as a question originally and with an invitation to critique my code. On 3/14/18, 5:11 AM, "Martin Maechler" <maechler at stat.math.ethz.ch> wrote:>>>>>> Henrik Bengtsson <henrik.bengtsson at gmail.com> >>>>>> on Tue, 13 Mar 2018 10:12:55 -0700 writes: > >> FYI, in R devel (to become 3.5.0), there's isFALSE() which will cut >> some corners compared to identical(): > >> > microbenchmark::microbenchmark(identical(FALSE, FALSE), >>isFALSE(FALSE)) >> Unit: nanoseconds >> expr min lq mean median uq max neval >> identical(FALSE, FALSE) 984 1138 1694.13 1218.0 1337.5 13584 100 >> isFALSE(FALSE) 713 761 1133.53 809.5 871.5 18619 100 > >> > microbenchmark::microbenchmark(identical(TRUE, FALSE), isFALSE(TRUE)) >> Unit: nanoseconds >> expr min lq mean median uq max neval >> identical(TRUE, FALSE) 1009 1103.5 2228.20 1170.5 1357 14346 100 >> isFALSE(TRUE) 718 760.0 1298.98 798.0 898 17782 100 > >> > microbenchmark::microbenchmark(identical("array", FALSE), >>isFALSE("array")) >> Unit: nanoseconds >> expr min lq mean median uq max neval >> identical("array", FALSE) 975 1058.5 1257.95 1119.5 1250.0 9299 100 >> isFALSE("array") 409 433.5 658.76 446.0 476.5 9383 100 > >Thank you Henrik! > >The speed of the new isTRUE() and isFALSE() is indeed amazing >compared to identical() which was written to be fast itself. > >Note that the new code goes back to a proposal by Herv? Pag?s >(of Bioconductor fame) in a thread with R core in April 2017. >The goal of the new code actually *was* to allow call like > > isTRUE(c(a = TRUE)) > >to become TRUE rather than improving speed. >The new source code is at the end of R/src/library/base/R/identical.R > >## NB: is.logical(.) will never dispatch: >## -- base::is.logical(x) <==> typeof(x) == "logical" >isTRUE <- function(x) is.logical(x) && length(x) == 1L && !is.na(x) && x >isFALSE <- function(x) is.logical(x) && length(x) == 1L && !is.na(x) && !x > >and one *reason* this is so fast is that all 6 functions which >are called are primitives : > >> sapply(codetools::findGlobals(isTRUE), function(fn) >>is.primitive(get(fn))) > ! && == is.logical is.na length > TRUE TRUE TRUE TRUE TRUE TRUE > >and a 2nd reason is probably with the many recent improvements of the >byte compiler. > > >> That could probably be used also is sapply(). The difference is that >> isFALSE() is a bit more liberal than identical(x, FALSE), e.g. > >> > isFALSE(c(a = FALSE)) >> [1] TRUE >> > identical(c(a = FALSE), FALSE) >> [1] FALSE > >> Assuming the latter is not an issue, there are 69 places in base R >> where isFALSE() could be used: > >> $ grep -E "identical[(][^,]+,[ ]*FALSE[)]" -r --include="*.R" | grep -F >>"/R/" | wc >> 69 326 5472 > >> and another 59 where isTRUE() can be used: > >> $ grep -E "identical[(][^,]+,[ ]*TRUE[)]" -r --include="*.R" | grep -F >>"/R/" | wc >> 59 307 5021 > >Beautiful use of 'grep' -- thank you for those above, as well. >It does need a quick manual check, but if I use the above grep >from Emacs (via 'M-x grep') or even better via a TAGS table >and M-x tags-query-replace I should be able to do the changes >pretty quickly... and will start looking into that later today. > >Interestingly and to my great pleasure, the first part of the >'Subject' of this mailing list thread, "Possible Improvement", >*has* become true after all -- > >-- thanks to Henrik ! > >Martin Maechler >ETH Zurich > > > >> On Tue, Mar 13, 2018 at 9:21 AM, Doran, Harold <HDoran at air.org> wrote: >> > Quite possibly, and I?ll look into that. Aside from the work I was >>doing, however, I wonder if there is a way such that sapply could avoid >>the overhead of having to call the identical function to determine the >>conditional path. >> > >> > >> > >> > From: William Dunlap [mailto:wdunlap at tibco.com] >> > Sent: Tuesday, March 13, 2018 12:14 PM >> > To: Doran, Harold <HDoran at air.org> >> > Cc: Martin Morgan <martin.morgan at roswellpark.org>; >>r-help at r-project.org >> > Subject: Re: [R] Possible Improvement to sapply >> > >> > Could your code use vapply instead of sapply? vapply forces you to >>declare the type and dimensions >> > of FUN's output and stops if any call to FUN does not match the >>declaration. It can use much less >> > memory and time than sapply because it fills in the output array as >>it goes instead of calling lapply() >> > and seeing how it could be simplified. >> > >> > Bill Dunlap >> > TIBCO Software >> > wdunlap tibco.com<http://tibco.com> >> > >> > On Tue, Mar 13, 2018 at 7:06 AM, Doran, Harold >><HDoran at air.org<mailto:HDoran at air.org>> wrote: >> > Martin >> > >> > In terms of context of the actual problem, sapply is called millions >>of times because the work involves scoring individual students who took >>a test. A score for student A is generated and then student B and such >>and there are millions of students. The psychometric process of scoring >>students is complex and our code makes use of sapply many times for each >>student. >> > >> > The toy example used length just to illustrate, our actual code >>doesn't do that. But your point is well taken, there may be a very good >>counterexample why my proposal doesn't achieve the goal is a >>generalizable way. >> > > > >[.................] >