Fast rolling functions through Rcpp
...or something similar.
# the outputs of these functions calls are identical (n doesn't alter the result)
roll_meanr(1:10, n = 0, c(0, 1))
roll_meanr(1:10, n = 2, c(0, 1))
roll_meanr(1:10, n = 100, c(0, 1))
# however, you have to give it an arbitrary value as the following throws an error
roll_meanr(1:10, weights = c(0, 1))hi. thanks very much for RcppRoll. but... :)
i find the following behavior odd
`
require(RcppRoll)
Loading required package: RcppRoll
roll_sum(1:9, n=2)
[1] 3 5 7 9 11 13 15 17
roll_sum(1:9, weights=1, n=2)
[1] 1 2 3 4 5 6 7 8 9
roll_sum(1:9, weights=c(1,1), n=2)
[1] 3 5 7 9 11 13 15 17
`
i would have expected weights to have been recycled, but, if i understand the code correctly, n is forced to the length of weights (if they are specified). i wonder if you might think of generating a warning message in this case? i certainly spent time scratching my head puzzling over this.
cheers!
I'm confused by what na.rm = TRUE is supposed to do in roll_median:
library(RcppRoll)
x <- 1:20
x[3:4] <- NA
x
roll_medianr(x = x,n = 5,na.rm = TRUE)
[1] NA NA NA NA 2 2 NA NA 7 8 9 10 11 12 13 14 15 16 17 18
roll_maxr(x = x,n = 5,na.rm = TRUE)
[1] NA NA NA NA 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Why is the median in position 6 equal to 2 rather than 3.5? Why are there NAs positions 7 and 8?
Am I missing something obvious or is this a bug?
A lot of the internals should be rewritten (as this package was generated when I was less proficient with C++ / Rcpp).
In addition, it would be awesome if we could take advantage of RcppParallel for easily parallelizing operations both:
The "align" argument has no effect on the behavior of function roll_mean(). The function roll_mean() always calculates the rolling mean using a centered window, regardless of the "align" argument.
da_ta <- cumsum(rnorm(500))
foo <- RcppRoll::roll_mean(da_ta, align="left", n=51)
bar <- RcppRoll::roll_mean(da_ta, align="right", n=51)
all.equal(foo, bar)
[1] TRUE
For roll_mean(), the documentation specifies align can be {"left", "middle", "right"} but the function specifies {"left", "center", "right"}.
Lines 36 and 23 of this file: https://github.com/kevinushey/RcppRoll/blob/master/R/RcppRoll.R
I have a sequence
0 1 0 1 1 1 1 0 1 0 1 0
I want the rolling sum of 3 periods with a right align.
1, 2, 2, 3, 3, 2, 2 , 1, 2, 1, 1, 0.
Instead I am getting
1, 2, 2, 3, 3, 2, 2 , 1, 2, 1, NA, NA
Is this what partial is supposed to solve? How can I get the last n-1 periods to sum correctly?
roll_max(c(NA, NA, 3, 2, 1), n=2L, fill=NA, align="right", na.rm=TRUE) now gives c(NA, -Inf, 3, 3, 2).
If I remember correctly, previous package versions gave the "desired" result c(NA, NA, 3, 3, 2).
Similarly roll_min(c(NA, NA, 3, 2, 1), 2L, fill=NA, align="right", na.rm=TRUE) results to c(NA, Inf, 3, 2, 1).
I currently have package version 0.3.0 installed on R version 4.1.1
Thank you.
Hi there,
Thank you for developing this extremely useful package.
When using roll_median() in the presence of NA values the output is inconsistent with all other functions, e.g. roll_mean(), roll_sum(), etc.
According to the documentation, all of the parameters are the same by default across roll_...() functions.
# a vector with NAs
vec <- c(1:5, rep(NA,3), 6:10, NA)
# expected output of computed & NA values
roll_mean(vec, 5)
roll_var(vec, 5)
# output using roll_median()
roll_median(vec, 5)Also for this function, there is no difference whether na.rm = c(TRUE, FALSE).
Thank you.
Cheers,
Christian
In the README or in a separate vignette. I ended up going through https://www.mytinyshinys.com/2017/05/10/rcpproll/ in order to make sure that I am using it right. It would be helpful to add such examples to the package documentation.
With the by argument set to > 1 the left out points are returned as random/uninitialized values.
Example:
roll_mean(rep(1,20), n = 3, by = 3)
[1] 1.000000e+00 1.071419e+200 9.385277e+223 1.000000e+00 4.311955e-80 1.840099e-80 1.000000e+00 7.724200e-37 2.647809e-32
[10] 1.000000e+00 1.426530e-71 3.791529e+179 1.000000e+00 3.315553e-33 5.391965e+241 1.000000e+00 7.096584e-110 9.486060e-322
A workaround is:
foo = roll_mean(rep(1,20), n = 3, by = 3)
foo = foo[seq(1, length(foo), by = 3)]
but it would be nice if this was done inside the roll_ functions.
Hi,
I would like to use the functions defined in Performance Analytics package like Burke Ratio, MaxDrawdown and Sharpe Ratio and then roll them using your package as I believe your package will give me substantial performance benefits. My knowledge in C++ is limited and it would be very difficult for me to recreate those R functions available in Performance Analytics package in C++. I would like to know if I can use your package to roll those R functions or do I have to rewrite them in C++?
Regards
SD
It would be great if you could add a roll_mad (for median absolute deviation : median(abs(Xi - median(X))) ) in the functions of your package.
The CRAN version RcppRoll is not availible for the R version 4
rownames, colnames lost after applying roll_* function on matrix.
Randomly, from time to time, either my R crashes or I get an error message by roll_median_..._numeric about "NULL provided where numeric is expected" or "internal provided where numeric is expected" or something like that.
It's happening across many different computers. It may work for a good long time, but then it may crash in the first few minutes as well.
I'm using it exclusively with a window of 3. I have no idea how to provide more information.
Not often that useful, but nice to have for completeness.
Hello Kevin..
I am trying to install RcppRoll from source using install_github("kevinushey/RcppRoll") but i am getting the following error (which I couldn't decipher); and i was wondering if you knew the cause/reason?
Thanks
---------------------- error output below this line---------------
Installing github repo RcppRoll/master from kevinushey
Downloading master.zip from https://github.com/kevinushey/RcppRoll/archive/master.zip
Installing package from /var/folders/jc/md64bd2x7nq0_j5ppqp89k_00000gn/T//Rtmp6Mmi9i/master.zip
arguments 'minimized' and 'invisible' are for Windows only
Installing RcppRoll
'/Library/Frameworks/R.framework/Resources/bin/R' --vanilla CMD INSTALL
'/private/var/folders/jc/md64bd2x7nq0_j5ppqp89k_00000gn/T/Rtmp6Mmi9i/devtools4b25b9accdc/RcppRoll-master'
--library='/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages' --install-tests
* installing source package 'RcppRoll' ...
** libs
g++ -arch x86_64 -I/Library/Frameworks/R.framework/Resources/include -DNDEBUG -I/usr/local/include -I"/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include" -I"/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/RcppArmadillo/include" -fPIC -mtune=core2 -g -O2 -c RcppExports.cpp -o RcppExports.o
g++ -arch x86_64 -I/Library/Frameworks/R.framework/Resources/include -DNDEBUG -I/usr/local/include -I"/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include" -I"/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/RcppArmadillo/include" -fPIC -mtune=core2 -g -O2 -c rollit.cpp -o rollit.o
rollit.cpp:245:14: error: no member named 'isnan' in the global namespace; did you mean 'is_nan'?
if (!::isnan(x[offset + i])) {
~~^~~~~
is_nan
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: 'is_nan' declared here
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:245:12: error: no matching function for call to 'is_nan'
if (!::isnan(x[offset + i])) {
^~~~~~~
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: candidate template ignored: could not match 'VectorBase<RTYPE, na, type-parameter-0-2>' against 'const double'
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:260:14: error: no member named 'isnan' in the global namespace; did you mean 'is_nan'?
if (!::isnan(x[offset + i])) {
~~^~~~~
is_nan
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: 'is_nan' declared here
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:260:12: error: no matching function for call to 'is_nan'
if (!::isnan(x[offset + i])) {
^~~~~~~
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: candidate template ignored: could not match 'VectorBase<RTYPE, na, type-parameter-0-2>' against 'const double'
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:324:14: error: no member named 'isnan' in the global namespace; did you mean 'is_nan'?
if (!::isnan(x[offset + i])) {
~~^~~~~
is_nan
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: 'is_nan' declared here
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:324:12: error: no matching function for call to 'is_nan'
if (!::isnan(x[offset + i])) {
^~~~~~~
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: candidate template ignored: could not match 'VectorBase<RTYPE, na, type-parameter-0-2>' against 'const double'
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:337:14: error: no member named 'isnan' in the global namespace; did you mean 'is_nan'?
if (!::isnan(x[offset + i])) {
~~^~~~~
is_nan
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: 'is_nan' declared here
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:337:12: error: no matching function for call to 'is_nan'
if (!::isnan(x[offset + i])) {
^~~~~~~
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: candidate template ignored: could not match 'VectorBase<RTYPE, na, type-parameter-0-2>' against 'const double'
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:357:13: error: no member named 'isnan' in the global namespace; did you mean 'is_nan'?
if (::isnan(x[offset + i])) {
~~^~~~~
is_nan
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: 'is_nan' declared here
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:357:11: error: no matching function for call to 'is_nan'
if (::isnan(x[offset + i])) {
^~~~~~~
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: candidate template ignored: could not match 'VectorBase<RTYPE, na, type-parameter-0-2>' against 'const double'
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:371:13: error: no member named 'isnan' in the global namespace; did you mean 'is_nan'?
if (::isnan(x[offset + i])) {
~~^~~~~
is_nan
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: 'is_nan' declared here
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:371:11: error: no matching function for call to 'is_nan'
if (::isnan(x[offset + i])) {
^~~~~~~
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: candidate template ignored: could not match 'VectorBase<RTYPE, na, type-parameter-0-2>' against 'const double'
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:422:13: error: no member named 'isnan' in the global namespace; did you mean 'is_nan'?
if (::isnan(x[offset + i])) {
~~^~~~~
is_nan
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: 'is_nan' declared here
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:422:11: error: no matching function for call to 'is_nan'
if (::isnan(x[offset + i])) {
^~~~~~~
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: candidate template ignored: could not match 'VectorBase<RTYPE, na, type-parameter-0-2>' against 'const double'
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:437:13: error: no member named 'isnan' in the global namespace; did you mean 'is_nan'?
if (::isnan(x[offset + i])) {
~~^~~~~
is_nan
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: 'is_nan' declared here
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:437:11: error: no matching function for call to 'is_nan'
if (::isnan(x[offset + i])) {
^~~~~~~
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: candidate template ignored: could not match 'VectorBase<RTYPE, na, type-parameter-0-2>' against 'const double'
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:455:13: error: no member named 'isnan' in the global namespace; did you mean 'is_nan'?
if (::isnan(x[offset + i])) continue;
~~^~~~~
is_nan
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: 'is_nan' declared here
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:455:11: error: no matching function for call to 'is_nan'
if (::isnan(x[offset + i])) continue;
^~~~~~~
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: candidate template ignored: could not match 'VectorBase<RTYPE, na, type-parameter-0-2>' against 'const double'
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
rollit.cpp:468:13: error: no member named 'isnan' in the global namespace; did you mean 'is_nan'?
if (::isnan(x[offset + i])) continue;
~~^~~~~
is_nan
/Users/ahmednagi/Documents/R_WorkSpaces/Rpackages/Rcpp/include/Rcpp/sugar/functions/is_nan.h:49:33: note: 'is_nan' declared here
inline sugar::IsNaN<RTYPE,NA,T> is_nan( const Rcpp::VectorBase<RTYPE,NA,T>& t){
^
fatal error: too many errors emitted, stopping now [-ferror-limit=]
20 errors generated.
make: *** [rollit.o] Error 1
ERROR: compilation failed for package 'RcppRoll'
> library('RcppRoll')
> get_rollit_source(roll_max, edit=FALSE)
Error in get("outFile", envir = environment(fun)) :
object 'outFile' not found
> ls(environment(roll_max))
[1] "get_rollit_source" "roll_max" "roll_maxl" "roll_maxr" "roll_mean"
[6] "roll_meanl" "roll_meanr" "roll_median" "roll_medianl" "roll_medianr"
[11] "roll_min" "roll_minl" "roll_minr" "roll_prod" "roll_prodl"
[16] "roll_prodr" "roll_sd" "roll_sdl" "roll_sdr" "roll_sum"
[21] "roll_suml" "roll_sumr" "roll_var" "roll_varl" "roll_varr"
[26] "rollit" "rollit_example" "rollit_raw"
Perhaps your package organization scheme changed from when this function was written?
> sessionInfo()
R version 3.4.2 (2017-09-28)
Platform: x86_64-apple-darwin16.7.0 (64-bit)
Running under: macOS Sierra 10.12.6
Matrix products: default
BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libLAPACK.dylib
locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] RcppRoll_0.2.2
loaded via a namespace (and not attached):
[1] compiler_3.4.2 tools_3.4.2 yaml_2.1.14 Rcpp_0.12.13
Something like this:
> library(zoo)
> ( x <- c(2, 4, 6, NA, 8) )
[1] 2 4 6 NA 8
> rollapplyr(x, 2, mean, na.rm = F) # same as `roll_mean(x, 2)`
[1] 3 5 NA NA
> rollapplyr(x, 2, mean, na.rm = T) # `roll_mean(x, 2, na.rm = T)`
[1] 3 5 6 8Kevin, is there any replacement for the rollit?
Hi Kevin, just wondering if you might consider including an option to pad the returned vectors/matrices with NAs when n > 1?
I'd love to be able to do something like this
data.frame(a = 1:10, b = roll_sum(1:10, 3, padNA = TRUE))
instead of this
data.frame(a = 1:10, b = c(rep(NA,2),roll_sum(1:10, 3))).
Of course data.frame(a = 1:x, b = roll_sum(1:x, n = 3)) throws an error because length of roll_sum(x,n) for n > 1 is shorter than length(1:x).
populate will fill a vector with values from the actual function applied for each method, while fill will fill that vector with e.g. NA.
These functions could potentially run in parallel since they do not read or write the same memory units.
For unit tests.
Hi Kevin. I am using your RcppRoll package from github, and will use it as a dependency in a package I am writing for identifying anomalous time series. In order to upload my package to CRAN, it would be useful if the CRAN version of RcppRoll was updated to match the version on github. Would that be possible?
Thanks.
Rob
Needed for skewed distributions.
Seems like it should be an easy extension of roll_median.
> tmp = c(1,1,1,1,NA,NA,NA,NA,1,1)
> roll_mean(tmp, 4, c(1,3,3,1), na.rm=T)
[1] 1.000000 1.166667 1.000000 0.500000 NaN 0.500000 1.000000
1.166667 and 0.5 are unexpected values. There should be only 1s, and one NaN.
This seems to work correctly when a default weighting is used:
> roll_mean(tmp, 4, na.rm=T)
[1] 1 1 1 1 NaN 1 1
any idea about rolling with multivarible for a matrix or data fram (data table)?
Would it be possible to implement backwards rolling. Thus, it would be a very convenient way to implement multi periods condition when handling panel data.
library(tidyverse)
#> + ggplot2 2.2.1 Date: 2017-05-29
#> + tibble 1.3.3 R: 3.3.2
#> + tidyr 0.6.3 OS: Windows 10 x64
#> + readr 1.1.1 GUI: RTerm
#> + purrr 0.2.2.2 Locale: Danish_Denmark.1252
#> + dplyr 0.5.0 TZ: Europe/Paris
#> + stringr 1.2.0
#> + forcats 0.2.0
#> Warning: package 'ggplot2' was built under R version 3.3.3
#> Warning: package 'tibble' was built under R version 3.3.3
#> Warning: package 'tidyr' was built under R version 3.3.3
#> Warning: package 'readr' was built under R version 3.3.3
#> Warning: package 'purrr' was built under R version 3.3.3
#> Warning: package 'dplyr' was built under R version 3.3.3
#> Warning: package 'forcats' was built under R version 3.3.3
#> -- Conflicts ----------------------------------------------------
#> * filter(), from dplyr, masks stats::filter()
#> * lag(), from dplyr, masks stats::lag()
library(RcppRoll)
#> Warning: package 'RcppRoll' was built under R version 3.3.3
df <- tibble(id = rep(1, each = 10), wage = c(980, rep(1000, each = 9)))
df %>%
mutate(wagecri = if_else(wage >= 1000, 1, 0)) %>%
mutate(crimean = roll_meanr(wagecri, n = 2))
#> # A tibble: 10 x 4
#> id wage wagecri crimean
#> <dbl> <dbl> <dbl> <dbl>
#> 1 1 980 0 NA
#> 2 1 1000 1 0.5
#> 3 1 1000 1 1.0
#> 4 1 1000 1 1.0
#> 5 1 1000 1 1.0
#> 6 1 1000 1 1.0
#> 7 1 1000 1 1.0
#> 8 1 1000 1 1.0
#> 9 1 1000 1 1.0
#> 10 1 1000 1 1.0
# I would like the result to be as follows
tibble(id = rep(1, each = 10), wage = c(980, rep(1000, each = 9))) %>%
mutate(wagecri = if_else(wage >= 1000, 1, 0)) %>%
mutate(crimean = c(0.5, rep(1, each = 8), NA_real_))
#> # A tibble: 10 x 4
#> id wage wagecri crimean
#> <dbl> <dbl> <dbl> <dbl>
#> 1 1 980 0 0.5
#> 2 1 1000 1 1.0
#> 3 1 1000 1 1.0
#> 4 1 1000 1 1.0
#> 5 1 1000 1 1.0
#> 6 1 1000 1 1.0
#> 7 1 1000 1 1.0
#> 8 1 1000 1 1.0
#> 9 1 1000 1 1.0
#> 10 1 1000 1 NABest regards,
Jakob
I am finding RcppRoll very convenient to use in conjunction with dplyr with one caveat: if I am doing rolling summaries over a numeric vector which is indexed by date (or a time period), then I may still want this to be used for calculating the rolling window (with value 0). It is analogous to using OLAP functions in SQL with range.
mutate_over() is my most recent attempt at implementing this functionality. @kevinushey @hadley I am wondering whether something like this will sit in RcppRoll or dplyr in the future?
Hi Kevin,
I hope you are well. Greetings from Lusaka.
I have installed RcppRoll through gitub
devtools::install_github("kevinushey/RcppRoll")
I have tried the standards function roll_mean, roll_sum, roll_sd. They work great. Now I want to create my own rolling function with rollit. But there is not such function in the package.
What am I missing?
Hello, I recently came across RcppRoll, sorry for the stupid question but is the rollit function allow the use of a user defined function (R function or cpp function).
I kind of see that calling a r function from Rcpp might be even worse than looping in R itself, but it would provide some kind of sugar is some cases. Typically using data.table by or equivalent dplyr behaviour one could call rolling functions by groups...
Do I understand correctly ?
If yes even if it seems trivial it might be worth adding a few words in the vignette so people won't expect this feature (and spend time trying :) ).
Thanks for providing this package to the community
Is this package still maintained?
I run into this problem. It seems 'na.rm' in roll_median doesn't work.
First, thanks for making RcppRoll available! It's very helpful for a new data science class that I'm teaching, especially because I can't expect my students to code Rcpp snippets.
Here's a little bug that I ran into. The normalize = TRUE flag will result in w itself being normalized. I can easily avoid this with weights = copy(w) but nonetheless I doubt this is the intended behavior.
Example:
x = 1:6
w = c(0.5, 0.5, 1.0)
roll_sum(x, n = 3L, weights = w, normalize = TRUE, align = "right", fill = NA)The result is that
> w
[1] 0.75 0.75 1.50
It would be nice if we can have a roll_ema
A bothering and quite dangerous bug is that the weights renormalization overwrites the given weights vector, instead of creating an inner temporary one :
a = 1:25
> d <- data.frame(w = c(0.2, 0.1, 0.1, 0.05, 0.05) ) # the total weight is 0.5 on purpose
> d
w
1 0.20
2 0.10
3 0.10
4 0.05
5 0.05
> roll_sum(a, n = length(d$w), weights = d$w)
[1] 11.5 16.5 21.5 26.5 31.5 36.5 41.5 46.5 51.5 56.5
61.5 66.5 71.5 76.5 81.5 86.5 91.5 96.5 101.5 106.5 111.5
> d
w
1 2.0
2 1.0
3 1.0
4 0.5
5 0.5
The fix should be quite simple, just creating a local weight vector inside the functions
More over, the weights renormalization seems to be enabled by default, which might be nice to mention in the popup help.
Thanks for all the work and this wonderful library.
If you repeatedly call a rolling function, e.g. roll_sumr and the vector length is smaller than the window length, then after some time it will segfault.
If the segfault bug is fixed, I think the current behavior of the dev version on github is better than v0.1.0 on CRAN, i.e. to just return a NA-vector (of appropriate class) that has the vector length (as opposed to the longer window length).
Because why would someone care how a window is aligned unless you planned on filling in the holes with something?
I think the weight is placed on the ordered values, not on the original values passed in.
E.g.
> roll_medianr(5:3, 3, weight = c(1,0,0))
gives NA NA 3 1xsmallest, 0x4, 0x5
Not NA NA 5 as expected (1x5,0x4,0x3)
My guess is this stems from sorting first without respect to the weights...
` NumericVector copy(x.begin() + offset, x.begin() + offset + n);
std::sort(copy.begin(), copy.end());
double weights_sum = sum(weights);
int k = 0;
double sum = weights_sum - weights[0];
while (sum > weights_sum / 2) {
++k;
sum -= weights[k];
}
return copy[k]; `
I think it would be nice to have the equivalent of zoo::na.locf in RcppRoll as well.
And it would be nice to limit how far the NAs are carried forward (or backward if fromLast=T). This is similar to the maxgap argument, but the difference is that maxgap just kills everything (which I think is not good, or deserves an additional argument):
x <- 1:8
x[4:5] <- NA
na.locf(x, maxgap = 1) # Current `zoo` behavior.
[1] 1 2 3 NA NA 6 7 8
na.locf(x, maxcarry = 1) # This is what I mean.
[1] 1 2 3 3 NA 6 7 8
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
Django
The Web framework for perfectionists with deadlines.
Laravel
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft
Open source projects and samples from Microsoft.
Google
Google ❤️ Open Source for everyone.
Alibaba
Alibaba Open Source for everyone
D3
Data-Driven Documents codes.
Tencent
China tencent open source team.