Getting started with cppally

Below we explore some of the capabilities of cppally, from its custom C++ scalar and vectors, to using templates and concepts.

Note: The classes r_vec and r_vector are aliases of one another and thus can be used interchangeably.

Setup

Let’s start by loading cppally

library(cppally)

Registering C++ functions (to R)

To make a C++ function available to R we use the [[cppally::register]] tag.

#include <cppally.hpp>
using namespace cppally;

[[cppally::register]]
void hello_world(){
  print("Hello World!");
}

After tagging our functions we want to make them available to R. To do that we have a few routes.

Registering C++ functions outside of a package context

After writing our hello world program in foo.cpp we can use cpp_source() to compile and register the function to R.

cpp_source(file = "src/foo.cpp")

Now the function is available in R

hello_world()
#> Hello World!

Similarly we can use the helper cpp_eval to run simple expressions and return the result without needing to include cppally.hpp and register the function.

cpp_eval('print("Hello World Again!")')
#> Hello World Again!

Note - For the rest of the examples it is assumed that the following code is always included beforehand.

#include <cppally.hpp>
using namespace cppally;

Registering C++ functions inside a cppally-linked package

Since cppally is header-only, we can include the headers directly into our own package.

General steps to using cppally in a package

Create package (if you haven’t already done so) using usethis::create_tidy_package()
Run cppally::use_cppally()
Run cppally::document()

This will automatically add the necessary package content needed to start working with cppally. For continuous development, use cppally::load_all() to compile and register cppally tagged functions, including our hello world function.

Note: We aim to integrate cppally registration into the devtools framework for ease-of-use.

C++ types

cppally offers a rich set of R types in C++ that are NA-aware. This means that common arithmetic and logical operations will account for NA in a similar fashion to R.

Scalars

logical scalar - `r_lgl`

cppally’s scalar version of logical, r_lgl can represent true, false or NA.

r_true
r_false
r_na

#> [1] TRUE
#> [1] FALSE
#> [1] NA

Logical operators work just like in R


[[cppally::register]]
r_vector<r_lgl> lgl_ops(){
  return make_vec<r_lgl>(
    r_true || r_false, // true
    r_true && r_false, // false
    r_na || r_true,    // true
    r_na && r_true,    // NA
    r_na && r_false,   // false
    r_na || r_na,      // NA
    r_na && r_na      // NA
  );
}

lgl_ops()
#> [1]  TRUE FALSE  TRUE    NA FALSE    NA    NA

Using r_lgl in if-statements

For type-safety reasons r_lgl cannot be implicitly converted to bool except in if-statements where an error is thrown if the value is NA.

DON’T do this:


[[cppally::register]]
void bad_lgl_print(r_lgl condition){
  if (condition){
    print("true");
  } else {
    print("false");
  }
}

bad_lgl_print(TRUE)
#> true
bad_lgl_print(FALSE)
#> false
bad_lgl_print(NA) # Can't implicitly convert NA to bool
#> Error:
#> ! Cannot implicitly convert r_lgl NA to bool, please check

DO this:


[[cppally::register]]
void good_lgl_print(r_lgl condition){
  if (is_na(condition)){
    print("NA");
  } else if (condition){
    print("true");
  } else {
    print("false");
  }
}

good_lgl_print(TRUE)
#> true
good_lgl_print(FALSE)
#> false
good_lgl_print(NA) # NA is handled explicitly so no issues
#> NA

We can also use r_lgl members is_true() and is_false() which return bool and are equivalent to R’s isTRUE() and isFALSE()


[[cppally::register]]
void also_good_lgl_print(r_lgl condition){
  if (condition.is_true()){
    print("true");
  } else {
    print("not true");
  }
}

also_good_lgl_print(TRUE)
#> true
also_good_lgl_print(FALSE)
#> not true
also_good_lgl_print(NA) # Falls into 'not true' branch here as expected
#> not true

Important: The && and || operators for r_lgl do NOT short-circuit like they do for bool. Both sides of the expression are always evaluated. If you specifically require short-circuiting behaviour, use is_true() and is_false() before using && and ||.

All cppally scalar types are implemented as structs that contain the underlying C/C++ types as well as other member functions.

cppally type	Description	Built on
`r_lgl`	Scalar logical	`int`
`r_int`	Scalar integer	`int`
`r_int64`	Scalar 64-bit integer	`int64_t`
`r_dbl`	Scalar double	`double`
`r_str`	Scalar string	`r_sexp`
`r_str_view`	Scalar string (view)	`SEXP`
`r_cplx`	Scalar double complex	`std::complex<double>`
`r_raw`	Scalar raw	`unsigned char`
`r_sym`	Symbol	`SEXP`
`r_date`	Scalar date	`r_dbl`
`r_psxct`	Scalar date-time	`r_dbl`
`r_sexp`	Generic R object (SEXP)¹	`SEXP`

NA values

Use is_na() to check that a value is NA and na<T>() to generate NA values.

is_na(na<r_int>())

#> [1] TRUE

C++ NA values and their R C API equivalents

Type	Value	R C API Value
`r_lgl`	`na<r_lgl>()`/`r_na`	`NA_LOGICAL`
`r_int`	`na<r_int>()`	`NA_INTEGER`
`r_int64`	`na<r_int64>()`	Not applicable
`r_dbl`	`na<r_dbl>()`	`NA_REAL`
`r_str`	`na<r_str>()`	`NA_STRING`
`r_cplx`	`na<r_cplx>()`	Not applicable
`r_sym`	Not applicable	Not applicable
`r_sexp`²	`na<r_sexp>()`/`r_null`	`R_NilValue`

Checking equality

There are two ways to check for exact equality of cppally scalars - with the == operator or with identical().

The cppally == operator always returns r_lgl and identical() always returns bool, which is a particularly important distinction when dealing with NA values as the former can represent NA while the latter cannot.


[[cppally::register]]
void cppally_equality(){

  r_int x = na<r_int>();
  r_int y = na<r_int>();
  
  r_lgl x_equal_to_y = x == y;
  bool x_identical_to_y = identical(x, y);
  
  // NA so not printed
  if ( x_equal_to_y.is_true() ){
    print("x is equal to y\n");
  }
  
  // NA so not printed
  if ( x_equal_to_y.is_false() ){
    print("x is not equal to y\n");
  }
  
  // NA so printed
  if (is_na(x_equal_to_y)){
    print("`x == y` produces `NA`\n");
  }
  
  // Both na<r_int>() therefore they are identical to each other
  if (x_identical_to_y){
    print("x is identical to y\n");
  }
}

cppally_equality()
#> `x == y` produces `NA`
#> x is identical to y

identical() can not only compare scalars, but also vectors, lists, factors, and data frames.


template <typename T, typename U>
[[cppally::register]]
bool cpp_identical(T x, U y){
  return identical(x, y);
}

cpp_identical(3L, 3L)
#> [1] TRUE
cpp_identical(NA, NA)
#> [1] TRUE
cpp_identical(3L, 3) # int != double
#> [1] FALSE
cpp_identical(1:10, 1:10)
#> [1] TRUE
cpp_identical(list(1, 2, 3), list(3, 2, 1))
#> [1] FALSE
cpp_identical(iris, iris)
#> [1] TRUE

Scalar operators

cppally also defines arithmetic and relational comparison operators for its scalar types. Like the logical and equality operators seen earlier, they are all NA-aware.

Scalar arithmetic operators

Addition

r_int(0) + r_dbl(2.5)

#> [1] 2.5

Subtraction

r_int(0) - r_int(1)

#> [1] -1

Multiplication

r_int(2) * r_int(3)

#> [1] 6

Division

r_dbl(9) / 3

#> [1] 3

Addition with NA

na<r_int>() + r_dbl(2.5)

#> [1] NA

Subtraction with NA

na<r_int>() - r_int(1)

#> [1] NA

Multiplication with NA

na<r_int>() * r_int(3)

#> [1] NA

Division with NA

na<r_dbl>() / 3

#> [1] NA

Scalar relational operators

Less than

r_int(1) < r_int(2)

#> [1] TRUE

Less than or equal to

r_dbl(2) <= r_dbl(2)

#> [1] TRUE

Greater than

r_int(3) > r_int(2)

#> [1] TRUE

Greater than or equal to

r_dbl(2) >= r_dbl(3)

#> [1] FALSE

Less than with NA

na<r_int>() < r_int(2)

#> [1] NA

Less than or equal to with NA

na<r_dbl>() <= r_dbl(2)

#> [1] NA

Greater than with NA

na<r_int>() > r_int(2)

#> [1] NA

Greater than or equal to with NA

na<r_dbl>() >= r_dbl(3)

#> [1] NA

Other defined operators not showcased: ++, --, +=, -=, *=, /=, %=, -, %, |, &, !

Vectors

cppally vectors are templated and can be thought of as containers of scalar elements like r_int, r_dbl, etc.

We can create vectors like so


// Integer vector of size n
[[cppally::register]]
r_vector<r_int> new_integer_vector(int n){
  r_vector<r_int> int_vctr(n, /*fill = */ r_int(0));
  return int_vctr;
}

new_integer_vector(3)
#> [1] 0 0 0

inline vectors

To create inline vectors, use make_vec<>

make_vec<r_dbl>(1, 1.5, 2, na<r_dbl>())

#> [1] 1.0 1.5 2.0  NA

We can add names on the fly with arg()


make_vec<r_dbl>(
    arg("first") = 1,
    arg("second") = 1.5,
    arg("third") = 2,
    arg("last") = na<r_dbl>()
  )

#>  first second  third   last 
#>    1.0    1.5    2.0     NA

In R a list is a generic vector, so cppally defines lists as r_vector<r_sexp>, a vector of the generic type r_sexp.

make_vec<r_sexp>(1, 2, 3)

#> [[1]]
#> [1] 1
#> 
#> [[2]]
#> [1] 2
#> 
#> [[3]]
#> [1] 3

A list of all cppally vectors of length 0


[[cppally::register]]
r_vector<r_sexp> all_vectors(){
  return make_vec<r_sexp>(
    arg("logical") = r_vector<r_lgl>(),
    arg("integer") = r_vector<r_int>(),
    arg("integer64") = r_vector<r_int64>(), // Requires bit64
    arg("double") = r_vector<r_dbl>(),
    arg("character") = r_vector<r_str>(),
    arg("character") = r_vector<r_str_view>(),
    arg("raw") = r_vector<r_raw>(),
    arg("date") = r_vector<r_date>(),
    arg("date-time") = r_vector<r_psxct>(),
    arg("list") = r_vector<r_sexp>()
  );
}

all_vectors()
#> $logical
#> logical(0)
#> 
#> $integer
#> integer(0)
#> 
#> $integer64
#> integer64(0)
#> 
#> $double
#> numeric(0)
#> 
#> $character
#> character(0)
#> 
#> $character
#> character(0)
#> 
#> $raw
#> raw(0)
#> 
#> $date
#> Date of length 0
#> 
#> $`date-time`
#> POSIXct of length 0
#> 
#> $list
#> list()

Scalar math

There is a rich suite of math functions that accept cppally types.


[[cppally::register]]
r_vector<r_dbl> cppally_math(r_dbl x){
    return make_vec<r_dbl>(
        arg("abs")          = abs(x),
        arg("floor")        = floor(x),
        arg("ceiling")      = ceiling(x),
        arg("trunc")        = trunc(x),
        arg("round")        = round(x),
        arg("signif")       = signif(x, 3),
        arg("sign")         = sign(x),
        arg("min")          = min(0, x),
        arg("max")          = max(0, x),
        arg("sqrt")         = sqrt(x),
        arg("pow")          = pow(x, 2),
        arg("exp")          = exp(x),
        arg("log")          = log(x),
        arg("log_base")     = log(x, 2),
        arg("log10")        = log10(x)
    );
}

cppally_math(2.5)
#>        abs      floor    ceiling      trunc      round     signif       sign 
#>  2.5000000  2.0000000  3.0000000  2.0000000  2.0000000  2.5000000  1.0000000 
#>        min        max       sqrt        pow        exp        log   log_base 
#>  0.0000000  2.5000000  1.5811388  6.2500000 12.1824940  0.9162907  1.3219281 
#>      log10 
#>  0.3979400
cppally_math(NA)
#>      abs    floor  ceiling    trunc    round   signif     sign      min 
#>       NA       NA       NA       NA       NA       NA       NA       NA 
#>      max     sqrt      pow      exp      log log_base    log10 
#>       NA       NA       NA       NA       NA       NA       NA

Coercion

To coerce from one scalar to another we can use as<T>


[[cppally::register]]
r_int double_to_int(r_dbl x){
  return as<r_int>(x);
}

double_to_int(pi)
#> [1] 3
double_to_int(NA_real_)
#> [1] NA

We can also coerce from one vector type to another


[[cppally::register]]
r_vector<r_int> to_int_vec(r_vector<r_dbl> x){
  return as<r_vector<r_int>>(x);
}

to_int_vec(c(0, 1.5, NA))
#> [1]  0  1 NA

Since as<T> is extremely flexible, we can also coerce from a scalar to a vector or vice versa


[[cppally::register]]
r_vector<r_sexp> coercions(){
    r_dbl a(4.2);
    r_vector<r_dbl> b = make_vec<r_dbl>(2.5);
    return make_vec<r_sexp>(
        as<r_vector<r_int>>(a),
        as<r_int>(a),
        as<r_int>(b),
        as<r_dbl>(b)
    );
}

coercions()
#> [[1]]
#> [1] 4
#> 
#> [[2]]
#> [1] 4
#> 
#> [[3]]
#> [1] 2
#> 
#> [[4]]
#> [1] 2.5

We can even coerce to and from C++ vectors


[[cppally::register]]
r_vector<r_int> cpp_vectors_example(r_vector<r_int> x){
  std::vector x_cpp = as<std::vector<r_int>>(x);
  x_cpp.push_back(r_int(42));
  return as<r_vector<r_int>>(x_cpp);
}

cpp_vectors_example(41L)
#> [1] 41 42

While coercing to a std::vector just to push back an element before coercing back might not be the most efficient, it does showcase how easy it is to work with cppally vectors and C++ vectors.

Strings

cppally provides the useful string type r_str

We can create R strings easily

r_str("hello")

#> [1] "hello"

To get a C or C++ string, use the members c_str() and cpp_str() respectively

C string via c_str()

r_str("hello").c_str()

#> [1] "hello"

C++ string_view via cpp_str()

This can be converted into a std::string via its constructor


[[cppally::register]]
r_str str_concatenate(r_str x, r_str y, r_str sep){
  std::string left = std::string(x.cpp_str());
  std::string right = std::string(y.cpp_str());
  std::string middle = std::string(sep.cpp_str());
  std::string combined = left + middle + right;
  return r_str(combined.c_str());
}

str_concatenate("hello", "how are you?", sep = ", ")
#> [1] "hello, how are you?"

Dates and date-times

cppally offers two classes: r_date for dates and r_psxct for date-times (based on POSIX time). These classes offer a rich set of member functions to perform date arithmetic and manipulation.

To quickly re-familiarise ourselves - R Dates and Date-times are essentially double-valued offsets from Unix time. Unix time is defined as the number of seconds since 1 Jan 1970, and so r_date is effectively an offset representing the number of days since epoch, and similarly r_psxct is an offset representing the number of seconds since epoch.

To create a new r_date, use the r_date(year, month, day) constructor.


[[cppally::register]]
r_date create_date(int year, int month, int day){
  return r_date(year, month, day);
}
[[cppally::register]]
r_date date_from_epoch_offset(double n){
  return r_date(n);
}

[[cppally::register]]
r_psxct create_datetime(int year, int month, int day, int hour, int minute, double second){
  return r_psxct(year, month, day, hour, minute, second);
}
[[cppally::register]]
r_psxct datetime_from_epoch_offset(double n){
  return r_psxct(n);
}

create_date(1970, 1, 1); # This builds the Unix epoch date
#> [1] "1970-01-01"

You can also build an r_date from an epoch offset.

date_from_epoch_offset(0) # Also builds the Unix epoch
#> [1] "1970-01-01"

Similarly, to create a new r_psxct, use the r_psxct(year, month, day, hour, minute, second) constructor.

create_datetime(1970, 1, 1, 0, 0, 0); # Unix epoch date-time
#> [1] "1970-01-01 UTC"
datetime_from_epoch_offset(0)         # Unix epoch from an offset
#> [1] "1970-01-01 UTC"

To get the current date and date-time, use r_date::today() and r_psxct::now() respectively.


[[cppally::register]]
r_date get_today(){
  return r_date::today();
}
[[cppally::register]]
r_psxct get_now(){
  return r_psxct::now();
}

get_today()
#> [1] "2026-10-25"
get_now()
#> [1] "2026-10-25 13:59:42 UTC"

Note: r_psxct currently only supports UTC and no other time-zones.

Adding days, months, and years

r_date and r_psxct both have the template function add<>, which allows us to add years, months, weeks, days, hours, minutes and seconds to our dates and date-times.


[[cppally::register]]
r_date add_days(r_date x, double n){
  return x.add<"days">(n);
}
[[cppally::register]]
r_date add_months(r_date x, double n){
  return x.add<"months">(n);
}

y2k <- create_date(2000, 1, 1)

# Day after y2k
y2k |>
  add_days(1)
#> [1] "2000-01-02"

# Week after y2k
y2k |>
  add_days(7)
#> [1] "2000-01-08"

# Month after y2k
y2k |>
  add_months(1)
#> [1] "2000-02-01"

# Year after y2k
y2k |>
  add_months(12)
#> [1] "2001-01-01"

Accessing date components

Those familiar with lubridate will instantly be familiar with the naming of these functions, something that was an intentional design choice to reduce confusion for those who are familiar with tidyverse.


[[cppally::register]]
r_int get_year(r_date x){
  return x.year();
}
[[cppally::register]]
r_int get_month(r_date x){
  return x.month();
}
[[cppally::register]]
r_int get_week(r_date x){
  return x.week();
}
[[cppally::register]]
r_int get_day(r_date x){
  return x.day();
}
[[cppally::register]]
r_int get_wday(r_date x, int week_start){
  return x.wday(week_start);
}
[[cppally::register]]
r_int get_yday(r_date x){
  return x.yday();
}
[[cppally::register]]
r_int get_iso_week(r_date x){
  return x.iso_week();
}
[[cppally::register]]
r_int get_iso_year(r_date x){
  return x.iso_year();
}

get_year(y2k)
#> [1] 2000
get_month(y2k)
#> [1] 1
get_week(y2k)
#> [1] 1
get_day(y2k) # Day of the month
#> [1] 1
get_wday(y2k, week_start = 1) # week_start = 1 = Mon
#> [1] 6
get_yday(y2k)
#> [1] 1

We can even retrieve the ISO week and ISO year.

get_iso_week(y2k) # Last (52nd) ISO week of the ISO year
#> [1] 52
get_iso_year(y2k) # ISO year is still 1999
#> [1] 1999

Date rounding

We can also round dates using the member functions floor(), ceiling(), and round().


[[cppally::register]]
r_date floor_month(r_date x){
  return x.floor<"months">();
}
[[cppally::register]]
r_date ceiling_year(r_date x){
  return x.ceiling<"years">();
}
[[cppally::register]]
r_date round_week(r_date x, int week_start){
  return x.round<"weeks">(week_start);
}

# Start of the month after advancing 100 days
y2k |> 
  add_days(100) |> 
  floor_month()
#> [1] "2000-04-01"

# ceiling() does not switch on boundary
y2k |> 
  ceiling_year()
#> [1] "2000-01-01"

y2k |> 
  add_days(1) |> 
  ceiling_year()
#> [1] "2001-01-01"

y2k |> 
  round_week(week_start = 1)
#> [1] "2000-01-03"

Difference between two dates

To calculate the time difference between two dates or date-times, use time_diff<>.


[[cppally::register]]
r_dbl diff_days(r_date x, r_date y){
  return time_diff<"days">(x, y);
}
[[cppally::register]]
r_dbl diff_months(r_date x, r_date y){
  return time_diff<"months">(x, y);
}

x <- y2k
y <- x |> add_months(3)

diff_days(x, y)
#> [1] 91
diff_months(x, y)
#> [1] 3

Fractional months

Both add<"months"> and time_diff<"months"> can deal with fractional months.

The previous sections dealt with dates, so we need to define functions that work with r_psxct (date-times).


[[cppally::register]]
r_psxct as_dt(r_date x){
  return x.as_datetime();
}
[[cppally::register]]
r_psxct dt_add_months(r_psxct x, double n){
  return x.add<"months">(n);
}
[[cppally::register]]
r_dbl dt_diff_months(r_psxct x, r_psxct y){
  return time_diff<"months">(x, y);
}

y2k_datetime <- as_dt(y2k)

mid_month <- y2k_datetime |> 
  dt_add_months(0.5)

# Half-way through the month 
mid_month
#> [1] "2000-01-16 12:00:00 UTC"

The difference between the start of the month and halfway through the month should return 0.5, confirming that fractional month arithmetic is working.

dt_diff_months(y2k_datetime, mid_month)
#> [1] 0.5

Symbols

Symbols have class r_sym and can be created directly from a string literal

r_sym("new_symbol")

#> new_symbol

Or from a cppally string

r_sym(r_str("symbol_from_string"))

#> symbol_from_string

Cached strings & symbols

cppally provides an efficient caching strategy for constructing cppally strings/symbols from string literals

cached_str<>

cached_str<"cached_string">()

#> [1] "cached_string"

This initialises the string once, caches it (to R’s CHARSXP pool), and efficiently re-uses the cached string for each subsequent call.

We can cache symbols in a similar way

cached_sym<"cached_symbol">()

#> cached_symbol

Lists

r_sexp is generally interpreted as an “element of a list” since lists are defined as r_vector<r_sexp>, a vector that holds generic r_sexp elements.


using list = r_vector<r_sexp>;

[[cppally::register]]
list new_list(int n){
  return list(n);
}

new_list(0)
#> list()
new_list(3)
#> [[1]]
#> NULL
#> 
#> [[2]]
#> NULL
#> 
#> [[3]]
#> NULL

The problem with a class like r_sexp is that it is by design generic and therefore difficult to work with in C++. To disambiguate the actual type we can use r_sexp_visit() via a C++ lambda.

Example: using r_sexp_visit() to resize every vector to length n in-place


[[cppally::register]]
r_vector<r_sexp> resize_all(r_vector<r_sexp> x, r_size_t n){
    r_size_t list_length = x.length();
    for (r_size_t i = 0; i < list_length; ++i){
        r_sexp_visit(x.view(i), [&]<RVector T>(T vec) {
            x.set(i, vec.resize(n));
        });
    }
    return x;
}

Factors

We can create a factor via r_factors()


[[cppally::register]]
r_factors new_factor(r_vector<r_str> x){
    return r_factors(x);
}

new_factor(letters)
#>  [1] a b c d e f g h i j k l m n o p q r s t u v w x y z
#> Levels: a b c d e f g h i j k l m n o p q r s t u v w x y z

In cppally, like R, factors are not vectors and therefore do not satisfy the RVector concept. To access the underlying integer codes vector, use the public codes() member function


static_assert(!RVector<r_factors>);

[[cppally::register]]
r_vector<r_int> factor_codes(r_factors x){
    return x.codes();
}

letter_fct <- new_factor(letters)

letter_fct |>
    factor_codes()
#>  [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
#> [26] 26

Calling R functions

cppally provides the r_function class, which can be used to easily call any R function.


[[cppally::register]]
r_date r_get_today(){
  r_function sys_date("Sys.Date");
  return as<r_date>(sys_date());
}

r_get_today()
#> [1] "2026-10-25"

To get a function from a specific package, use pkg_env, a helper that returns the namespace of a package. The result of pkg_env can be passed to the env (or second) arg of r_function’s symbol and string constructors.


[[cppally::register]]
r_function base_sum_fn(){
  return r_function(cached_sym<"sum">(), pkg_env<"base">());
}

base_sum_fn()
#> function (..., na.rm = FALSE)  .Primitive("sum")

r_function also allows named arguments to be passed.


[[cppally::register]]
r_dbl base_sum(r_vector<r_dbl> x, bool na_rm){
  return as<r_dbl>(base_sum_fn()(x, arg("na.rm") = na_rm));
}

base_sum(c(NA, 1:3, NA), na_rm = TRUE)
#> [1] 6

Value matching

Use r_vector member find() to find the 0-indexed locations of a scalar value.


[[cppally::register]]
r_vector<r_int> find_empty_string(r_vector<r_str> x){
    return x.find(r_str(""));
}

x <- c("zero", "one", "two", "three", "four")

find_empty_string(x)
#> integer(0)

# Add empty strings
x[c(1, 3)] <- ""

find_empty_string(x)
#> [1] 0 2

To find the locations of multiple values (first match), use match(). It works like R’s base::match() but is 0-indexed.


[[cppally::register]]
r_vector<r_int> match_strs(r_vector<r_str> x, r_vector<r_str> table){
    return match(x, table);
}

letters
#>  [1] "a" "b" "c" "d" "e" "f" "g" "h" "i" "j" "k" "l" "m" "n" "o" "p" "q" "r" "s"
#> [20] "t" "u" "v" "w" "x" "y" "z"

vowels <- c("a", "e", "i", "o", "u")

# cppally::match is 0-indexed
match_strs(vowels, letters)
#> [1]  0  4  8 14 20

cppally provides the IS_IN infix operator, identical to R’s %in%


[[cppally::register]]
r_vector<r_lgl> cpp_in(r_vector<r_str> x, r_vector<r_str> table){
    return x IS_IN table;
}

cpp_in(c("a", "A", NA), letters)
#> [1]  TRUE FALSE FALSE

To mimic R’s new %notin% operator, simply use the logical negation operator.


[[cppally::register]]
r_vector<r_lgl> cpp_not_in(r_vector<r_str> x, r_vector<r_str> table){
    return !(x IS_IN table);
}

Technical note: cppally internally negates the intermediate result of x IS_IN table in-place in this particular case because it satisfies specific properties of exclusivity which while not covered here, may be covered in a later vignette. This in-place negation is naturally efficient as it avoids allocating a new vector.

cpp_not_in(c("a", "A", NA), letters)
#> [1] FALSE  TRUE  TRUE

Subsetting

Subsetting is 0-indexed in cppally, as is all other indexing.


template <RVector T, typename subscript_t>
requires any<subscript_t, r_lgl, r_int, r_str>
[[cppally::register]]
T cpp_subset(T x, r_vector<subscript_t> y){
    return subset(x, y);
}

Subsetting with integer indices

x <- 1:10
cpp_subset(x, 0L) # index 0 is 1st value
#> [1] 1
cpp_subset(x, 9L) # index 9 is last value here
#> [1] 10
cpp_subset(x, 10L) # index 10 is out-of-bounds so NA is returned
#> [1] NA
cpp_subset(x, NA_integer_) # NA is returned with integer NA index
#> [1] NA

Subsetting with logical indices

cpp_subset(x, x > 5)
#> [1]  6  7  8  9 10
cpp_subset(x, x > 100)
#> integer(0)

# It differs to base subsetting (via `[`)
cpp_subset(x, rep(NA, 10)) # cppally only returns values associated with TRUE
#> integer(0)
x[rep(NA, 10)] # base R returns NA values when subsetting with NA logicals
#>  [1] NA NA NA NA NA NA NA NA NA NA

base R performs negative subsetting by using negative numbers. This is not possible with our 0-indexed subset as we would have to also represent negative zero to subset everything except the first element at index 0. Instead cppally has an invert argument.


template <RVector T, typename subscript_t>
requires any<subscript_t, r_lgl, r_int, r_str>
[[cppally::register]]
T cpp_negative_subset(T x, r_vector<subscript_t> y){
    return subset(x, y, /*invert=*/ true);
}

cpp_negative_subset(x, 0L) # Everything but 1st
#> [1]  2  3  4  5  6  7  8  9 10
cpp_negative_subset(x, x > 5) # Everything except where x > 5
#> [1] 1 2 3 4 5
cpp_negative_subset(x, integer()) # Everything
#>  [1]  1  2  3  4  5  6  7  8  9 10

Named subsetting is also supported. cppally internally hashes the vector names and performs hash lookups. For more info, see the ‘Vector Names’ vignette.

names(x) <- LETTERS[seq_along(x)]

cpp_subset(x, c("A", "J", "Z"))
#>    A    J <NA> 
#>    1   10   NA
cpp_negative_subset(x, "A")
#>  B  C  D  E  F  G  H  I  J 
#>  2  3  4  5  6  7  8  9 10

Concepts and Templates

One of the most powerful features of C++20 are concepts. These allow users to write human-readable templates and constraints.

When writing your own templates, it is necessary to place them in headers for cppally registration to work correctly.

Let’s practice by creating the abs() function in C++ using templates and the RNumber concept.


template <RNumber T>
[[cppally::register]]
T cpp_abs(T x){
  if (is_na(x)){
    return na<T>();
  } else if (x < 0){
    return -x;
  } else {
    return x;
  }
}

What’s nice is that it works correctly for integers and doubles while simultaneously preserving their type

cpp_abs(-4.2)
#> [1] 4.2
cpp_abs(-3L)
#> [1] 3

class(cpp_abs(-4.2)) # Double preserved
#> [1] "numeric"
class(cpp_abs(-3L)) # Integer preserved
#> [1] "integer"

This type of programming is historically tricky within the R C API and typically necessitates a switch statement that switches on the object’s type, handling each type separately. With our abs() template, the logic is correctly handled with one set of operations.

How it works

The top-line template <RNumber T> declares a template that encapsulates T, an RNumber - a concept that contains r_int, r_int64 and r_dbl

If x is NA then we immediately also return NA via na<T>() which is a templated function that returns NA of the input type T.

To correctly register templates, the ‘[[cppally::register]]’ tag must always go above the function name.

template <typename T>
[[cppally::register]] // <--- Here
T foo(T x){
  return x;
}

Templates without function arguments

Explicit instantiation (from R) is unfortunately not possible and template types must be deduced from supplied arguments.

template <typename T>
[[cppally::register]]
T foo(){
    return T();
}

Here foo() will not be compiled because the function has no arguments that let the compiler automatically deduce what T is. In C++ you would always call this function like so: foo<T>(). Unfortunately we can’t do that from R directly.

You may get a cryptic compiler error like this

error: no matching function for call to 'foo()'
[]<typename T>() -> decltype(cpp_to_r(::foo())) {

along with an equally cryptic note

note:   couldn't deduce template parameter 'T'
[]<typename T>() -> decltype(cpp_to_r(::foo())) {

Even though these kinds of templates can be written with cppally in C++, they cannot be exported to R.

An obvious and somewhat ugly workaround is to include a prototype argument that allows the template parameter to be deduced from.


// Return the default constructor result of RScalar types

template <RScalar T>
[[cppally::register]]
T scalar_default(T ptype){
    return T();
}

scalar_default(integer(1)) # Default is 0L
#> [1] 0
scalar_default(numeric(1)) # Default is 0.0
#> [1] 0
scalar_default(character(1)) # Default is ""
#> [1] ""

Exporting variadic templates are also not supported. The best alternative is to use lists (r_vector<r_sexp>).

In the above example we used the RScalar concept which includes all cppally scalar types (excluding r_sexp). For a list of all cppally concepts, please see the Annex

Attributes

Attributes can be manipulated via functions defined in the attr namespace.

Example: Adding names to a list


[[cppally::register]]
r_vector<r_sexp> set_list_names(r_vector<r_sexp> x, r_vector<r_str> names){
  x.set_names(names);
  return x;
}

set.seed(42)
norm_samples <- lapply(1:5, \(x) rnorm(10, mean = x))
set_list_names(norm_samples, paste0("sample_", 1:5))
#> $sample_1
#>  [1] 2.3709584 0.4353018 1.3631284 1.6328626 1.4042683 0.8938755 2.5115220
#>  [8] 0.9053410 3.0184237 0.9372859
#> 
#> $sample_2
#>  [1]  3.3048697  4.2866454  0.6111393  1.7212112  1.8666787  2.6359504
#>  [7]  1.7157471 -0.6564554 -0.4404669  3.3201133
#> 
#> $sample_3
#>  [1] 2.693361 1.218692 2.828083 4.214675 4.895193 2.569531 2.742731 1.236837
#>  [9] 3.460097 2.360005
#> 
#> $sample_4
#>  [1] 4.455450 4.704837 5.035104 3.391074 4.504955 2.282991 3.215541 3.149092
#>  [9] 1.585792 4.036123
#> 
#> $sample_5
#>  [1] 5.205999 4.638943 5.758163 4.273295 3.631719 5.432818 4.188607 6.444101
#>  [9] 4.568554 5.655648

More useful attribute helpers

get_attrs() - Returns a list of attributes (possibly r_vector<r_sexp>(r_null))
set_attrs() - Sets attributes to ones specified. Note: replaces any current attributes
clear_attrs() - Removes all attributes
set_attr() - Set a single attribute
get_attr() - Get a single attribute
inherits1() - Does object inherit class?
inherits_any() - Does object inherit at least one of the specified classes?
inherits_all() - Does object inherit all of the specified classes?
modify_attrs() - Modifies current attributes but doesn’t remove any existing ones

Regular sequences

There are two core functions for generating regular sequences - sequence() and seq(). seq() behaves exactly like the R equivalent base::seq(), and sequence() behaves like the R equivalent base::sequence(), with the exception that it accepts scalar arguments instead of vector ones, and also works with decimal increments.

seq(r_dbl(1), r_dbl(5), r_dbl(0.5))

#> [1] 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

sequence(5, /*from = */ r_int(0), /*by = */ r_int(-1))

#> [1]  0 -1 -2 -3 -4

We can also use sequence() to easily replicate base::seq_len()


[[cppally::register]]
r_vector<r_int> cpp_seq_len(r_size_t n){
  return sequence(n, /* from = */ r_int(1), /* by = */ r_int(1));
}

cpp_seq_len(5)
#> [1] 1 2 3 4 5

It is also straightforward to replicate base::sequence() with pmap()


template <typename T>
requires (any<T, r_int, r_int64, r_dbl>)
[[cppally::register]]
r_vector<r_sexp> cpp_sequences(r_vector<r_int> size, r_vector<T> from, r_vector<T> by){
    return pmap([](auto a, auto b, auto c){
        return as<r_sexp>(sequence(a, b, c));
    }, size, from, by);
}

cpp_sequences(1:3, from = 0L, by = 1L) |> 
  unlist()
#> [1] 0 0 1 0 1 2

# Same as base R
sequence(1:3, from = 0L, by = 1L)
#> [1] 0 0 1 0 1 2

Using multiple OpenMP threads

To set the number of global OpenMP threads, use cppally::set_threads().

cppally::set_threads(4); // Sets the number of threads to 4

To see the number of currently available threads, use cppally::get_threads().

From R we can then register these C++ function and use them.


[[cppally::register]]
void cpp_set_threads(int n){
  set_threads(n);
}

[[cppally::register]]
int cpp_get_threads(){
  return get_threads();
}


cpp_set_threads(2) # Set to 2 threads
cpp_get_threads()  # Number of threads now available
#> [1] 2

cpp_set_threads(1) # Reset back to 1
cpp_get_threads()  # Should return 1
#> [1] 1

Because this is unique to each dll file, setting threads in one R package doesn’t affect another. Once threads have been set, all cppally code that can make use of them, will use them. This means you should speed improvements across many cppally functions.

Stats functions

cppally provides some stats functions which are both common and difficult to implement efficiently and safely.

sum() - Sum of values
mean() - Average of values
range() - Min and max range of values
var() - Variance
quantile() - Sample quantiles³


template <RNumber T>
[[cppally::register]]
r_dbl cpp_sum(r_vector<T> x, bool na_rm){
  return as<r_dbl>(sum(x, na_rm));
}
template <RNumber T>
[[cppally::register]]
r_vector<T> cpp_range(r_vector<T> x, bool na_rm){
  return range(x, na_rm);
}
template <RNumber T>
[[cppally::register]]
r_dbl cpp_var(r_vector<T> x, bool na_rm){
  return var(x, na_rm);
}
template <RNumber T>
[[cppally::register]]
r_vector<r_dbl> cpp_quantile(r_vector<T> x, r_vector<r_dbl> probs, bool na_rm, bool names){
  return quantile(x, probs, na_rm, names); 
}

library(bench)

set.seed(1)
x <- sample.int(10, 5e05, replace = TRUE)
x[sample.int(10^5, 10^4)] <- NA # Add many NAs

# Sum (ignoring NAs)
mark(
  cppally_sum = cpp_sum(x, na_rm = TRUE),
  base_sum = sum(x, na.rm = TRUE)
)
#> # A tibble: 2 × 6
#>   expression       min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_sum   88.7µs   89.7µs    11029.        0B        0
#> 2 base_sum     183.5µs  186.8µs     4894.        0B        0

# Sum (not ignoring NAs)
mark(
  cppally_sum = cpp_sum(x, na_rm = FALSE),
  base_sum = sum(x, na.rm = FALSE)
)
#> # A tibble: 2 × 6
#>   expression       min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_sum    721ns    781ns  1160593.        0B        0
#> 2 base_sum       130ns    150ns  5698038.        0B        0

# Range (ignoring NAs)
mark(
  cppally_range = cpp_range(x, na_rm = TRUE),
  base_range = range(x, na.rm = TRUE)
)
#> # A tibble: 2 × 6
#>   expression         min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>    <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_range 294.45µs 299.18µs     3314.        0B       0 
#> 2 base_range      1.99ms   2.15ms      384.    11.4MB     249.

# Range (not ignoring NAs)
mark(
  cppally_range = cpp_range(x, na_rm = FALSE),
  base_range = range(x, na.rm = FALSE)
)
#> # A tibble: 2 × 6
#>   expression         min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>    <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_range    731ns    781ns  1210758.        0B      0  
#> 2 base_range       287µs    817µs     1339.    1.91MB     70.4

# Variance (ignoring NAs)
mark(
  cppally_var = cpp_var(x, na_rm = TRUE),
  base_var = var(x, na.rm = TRUE)
)
#> # A tibble: 2 × 6
#>   expression       min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_var 486.38µs 497.39µs     1920.        0B      0  
#> 2 base_var      2.79ms   2.86ms      323.    5.74MB     51.7

# Variance (not ignoring NAs)
mark(
  cppally_var = cpp_var(x, na_rm = FALSE),
  base_var = var(x, na.rm = FALSE)
)
#> # A tibble: 2 × 6
#>   expression       min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_var    721ns    771ns  1209930.        0B       0 
#> 2 base_var       193µs    785µs     1289.    3.81MB     116.

# Quantile (ignoring NAs)
# Since quantile(na.rm = F) errors when NAs are present, we can compare against it
mark(
  cppally_quantile = cpp_quantile(x, c(0.25, 0.75), na_rm = TRUE, names = FALSE), 
  base_quantile = quantile(x, c(0.25, 0.75), na.rm = TRUE, names = FALSE)
)
#> # A tibble: 2 × 6
#>   expression            min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>       <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_quantile 803.98µs   1.27ms      787.    1.91MB     35.0
#> 2 base_quantile      5.71ms   5.84ms      168.    13.3MB    101.
# Many probabilities
mark(
  cppally_quantile = cpp_quantile(x, seq(0, 1, 0.001), na_rm = TRUE, names = FALSE), 
  base_quantile = quantile(x, seq(0, 1, 0.001), na.rm = TRUE, names = FALSE)
)
#> # A tibble: 2 × 6
#>   expression            min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>       <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_quantile  828.4µs   1.32ms     801.     1.94MB     44.7
#> 2 base_quantile      15.2ms  15.42ms      60.8   13.43MB     37.1

# Using multiple threads

cpp_set_threads(4)

# Sum (ignoring NAs)
mark(
  cppally_sum = cpp_sum(x, na_rm = TRUE),
  base_sum = sum(x, na.rm = TRUE)
)
#> # A tibble: 2 × 6
#>   expression       min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_sum     48µs   56.1µs    17509.        0B        0
#> 2 base_sum       181µs    186µs     5263.        0B        0

# Sum (not ignoring NAs)
mark(
  cppally_sum = cpp_sum(x, na_rm = FALSE),
  base_sum = sum(x, na.rm = FALSE)
)
#> # A tibble: 2 × 6
#>   expression       min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_sum    661ns    771ns  1090736.        0B        0
#> 2 base_sum       130ns    141ns  3846969.        0B        0

# Range (ignoring NAs)
mark(
  cppally_range = cpp_range(x, na_rm = TRUE),
  base_range = range(x, na.rm = TRUE)
)
#> # A tibble: 2 × 6
#>   expression         min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>    <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_range  74.93µs  76.54µs    12673.        0B       0 
#> 2 base_range      1.99ms   2.05ms      481.    11.4MB     328.

# Range (not ignoring NAs)
mark(
  cppally_range = cpp_range(x, na_rm = FALSE),
  base_range = range(x, na.rm = FALSE)
)
#> # A tibble: 2 × 6
#>   expression         min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>    <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_range    741ns    781ns  1197256.        0B       0 
#> 2 base_range       287µs    293µs     3394.    1.91MB     162.

# Variance (ignoring NAs)
mark(
  cppally_var = cpp_var(x, na_rm = TRUE),
  base_var = var(x, na.rm = TRUE)
)
#> # A tibble: 2 × 6
#>   expression       min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_var 195.69µs 200.58µs     4873.        0B      0  
#> 2 base_var      2.81ms   2.84ms      332.    5.72MB     52.3

# Variance (not ignoring NAs)
mark(
  cppally_var = cpp_var(x, na_rm = FALSE),
  base_var = var(x, na.rm = FALSE)
)
#> # A tibble: 2 × 6
#>   expression       min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_var    731ns    771ns  1223153.        0B       0 
#> 2 base_var       165µs    183µs     5315.    3.81MB     483.

# Reset to single-threaded
cpp_set_threads(1)

A note on cppally::sum

cppally::sum() is written to use SIMD instructions for performance where possible. It also uses higher width integers for accuracy when dealing with integer vectors. For example, if summing a 32-bit integer vector, it uses intermediate 64-bit integers, and likewise if summing a 64-bit integer vector, it uses intermediate 128-bit integers (if your compiler supports it). Doing it this way avoids integer overflow entirely for vectors whose length is less than 2^32. It also avoids intermediate integer overflow that would have arisen if same-width integers were used. For example, for x <- .Machine$integer.max * c(-1L, -1L, 1L, 1L), the sum of x is trivially 0, but the intermediate addition of x[1] and x[2] would result in overflow if one were to use 32-bit integers.

Unique values

n_unique() is a convenient helper to calculate the number of unique values. unique() is also highly performant.


template <RVector T>
[[cppally::register]]
r_int cpp_n_unique(T x){
  return as<r_int>(n_unique(x));
}

template <RVector T>
[[cppally::register]]
T cpp_unique(T x, bool sort){
  return unique(x, sort);
}

x <- sample(1:100, 10^5, replace = TRUE)

mark(
  base_n_unique = length(unique(x)),
  cppally_n_unique = cpp_n_unique(x)
)
#> # A tibble: 2 × 6
#>   expression            min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>       <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 base_n_unique       367µs    379µs     2618.    1.38MB     82.0
#> 2 cppally_n_unique    102µs    103µs     9565.        0B      0

mark(
  base_unique = unique(x),
  cppally_unique = cpp_unique(x, sort = FALSE)
)
#> # A tibble: 2 × 6
#>   expression          min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>     <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 base_unique       364µs    376µs     2633.    1.38MB     84.1
#> 2 cppally_unique    112µs    114µs     8684.      448B      0

mark(
  base_sorted_unique = sort(unique(x)),
  cppally_sorted_unique = cpp_unique(x, sort = TRUE)
)
#> # A tibble: 2 × 6
#>   expression                 min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>            <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 base_sorted_unique       381µs    393µs     2466.    1.38MB     77.0
#> 2 cppally_sorted_unique    113µs    117µs     7769.      896B      0

Sorting

cppally also offers order() and sort() for sorting vectors.

A hybrid approach is used, incorporating various sorting algorithms like counting sort, LSD radix sort, insertion sort, etc.



template <RSortableVector T>
[[cppally::register]]
r_vector<r_int> cpp_order(T x){
  
  // Add 1 to return 1-indexed permutation to match base::order()
  // In general C++ code, always use 0-indexed as cppally indexing is 
  // always 0-indexed.
  
  return order(x) + 1; 
}

template <RSortableVector T>
[[cppally::register]]
T cpp_sort(T x){
  return sort(x);
}

cppally::sort always sorts NA values last and so is equivalent to base::sort(..., method = "radix", na.last = TRUE).

r_radix_sort <- function(x){
  sort(x, method = "radix", na.last = TRUE)
}
r_radix_order <- function(x){
  order(x, method = "radix", na.last = TRUE)
}

# Small data
mark(cpp_sort(c(3, 2, 1)), r_radix_sort(c(3, 2, 1)))
#> # A tibble: 2 × 6
#>   expression                    min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>               <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cpp_sort(c(3, 2, 1))      761.1ns    881ns   938567.        0B      0  
#> 2 r_radix_sort(c(3, 2, 1))   14.8µs   16.5µs    58442.        0B     23.4
mark(cpp_order(c(3, 2, 1)), r_radix_order(c(3, 2, 1)))
#> # A tibble: 2 × 6
#>   expression                     min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>                <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cpp_order(c(3, 2, 1))     792.09ns 862.05ns   999563.        0B      0  
#> 2 r_radix_order(c(3, 2, 1))   6.74µs   7.57µs   126120.        0B     25.2

# Integer data
ints <- sample.int(20, 2e05, replace = TRUE)
mark(cpp_sort(ints), r_radix_sort(ints))
#> # A tibble: 2 × 6
#>   expression              min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>         <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cpp_sort(ints)        575µs    699µs     1363.    1.53MB     45.7
#> 2 r_radix_sort(ints)    607µs    730µs     1152.    1.53MB     40.1
mark(cpp_order(ints), r_radix_order(ints))
#> # A tibble: 2 × 6
#>   expression               min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>          <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cpp_order(ints)        492µs    692µs     1609.     781KB     29.5
#> 2 r_radix_order(ints)    306µs    349µs     2365.     781KB     39.7

# Double-floating point precision data
dbls <- as.double(ints)
mark(cpp_sort(dbls), r_radix_sort(dbls))
#> # A tibble: 2 × 6
#>   expression              min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>         <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cpp_sort(dbls)     958.02µs   1.08ms      928.    2.29MB     63.7
#> 2 r_radix_sort(dbls)   2.57ms   2.66ms      374.    2.29MB     27.5
mark(cpp_order(dbls), r_radix_order(dbls))
#> # A tibble: 2 × 6
#>   expression               min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>          <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cpp_order(dbls)      912.6µs  951.1µs      963.     781KB     21.5
#> 2 r_radix_order(dbls)    2.1ms    2.2ms      434.     781KB     10.5

# Floating point data with decimals
dcmls <- rnorm(length(dbls))
mark(cpp_sort(dcmls), r_radix_sort(dcmls))
#> # A tibble: 2 × 6
#>   expression               min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>          <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cpp_sort(dcmls)       2.37ms   2.79ms      359.    2.29MB     23.9
#> 2 r_radix_sort(dcmls)   4.31ms   4.49ms      220.    2.29MB     15.4
mark(cpp_order(dcmls), r_radix_order(dcmls))
#> # A tibble: 2 × 6
#>   expression                min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>           <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cpp_order(dcmls)       2.46ms   2.54ms      388.     781KB     8.35
#> 2 r_radix_order(dcmls)   4.05ms   4.13ms      237.     781KB     6.25

# String data
strs <- round(dcmls, 2) |> paste0()
mark(cpp_sort(strs), r_radix_sort(strs))
#> # A tibble: 2 × 6
#>   expression              min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>         <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cpp_sort(strs)       1.51ms   1.61ms      612.    3.06MB     73.2
#> 2 r_radix_sort(strs)   1.25ms   1.57ms      663.    2.29MB     59.1
mark(cpp_order(strs), r_radix_order(strs))
#> # A tibble: 2 × 6
#>   expression               min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>          <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cpp_order(strs)       1.04ms   1.18ms      835.    1.53MB     44.2
#> 2 r_radix_order(strs) 501.45µs 514.31µs     1923.   781.3KB     43.3

Grouping

Some areas of analysis require working with groups, with sometimes very large numbers of them. cppally provides a powerful class groups, along with make_groups(), a function that creates groups from vectors.

The structure of groups is:

r_vector<r_int> ids - The group IDs
int n_groups - Number of unique groups
bool ordered - Do the group IDs specify a sorting order, or are they by order-of-first-appearance?
bool sorted - Are the group IDs sorted? (This can also be true for order-of-first-appearance IDs)
r_vector<r_int> starts() - Returns a vector of start locations of each unique group, signifying the location in the data at which each group initially appeared
r_vector<r_int> counts() - Returns a vector of frequency counts of each unique group
r_vector<r_int> order() - Returns a 0-indexed permutation vector that can be used to sort group IDs

Note: When the groups are in sorted-order (where sorted is true), member functions like starts() and counts() use a hybrid search strategy combining linear, gallop, and binary searching. This strategy ensures that search time complexity scales with the number of groups and not the size of the data.



template <RVector T>
[[cppally::register]]
r_vector<r_sexp> group_info(T x){
  groups g = make_groups(x, /*ordered=*/ false);
  return make_vec<r_sexp>(
    arg("group_id") = g.ids,
    arg("n_groups") = g.n_groups,
    arg("ordered") = g.ordered,
    arg("sorted") = g.sorted,
    arg("starts") = g.starts(),
    arg("counts") = g.counts(),
    arg("order") = g.order(),
    arg("group_names") = group_names(x, g)
  );
}

set.seed(1)
x <- sample(c("A", "B", "C"), 20, replace = TRUE)

# Convert into group info

letter_groups <- group_info(x)

letter_groups$group_id # Group IDs
#>  [1] 0 1 0 2 0 1 1 2 2 1 1 0 0 0 2 2 2 2 1 0
letter_groups$n_groups # Number of groups
#> [1] 3
letter_groups$ordered  # Groups were collected using order-of-first-appearance
#> [1] FALSE
letter_groups$sorted   # Are group IDs in sorted order?
#> [1] FALSE
letter_groups$starts   # Data start locations (0-indexed)
#> [1] 0 1 3
letter_groups$counts   # Group counts
#> [1] 7 6 7
letter_groups$order    # Order permutation (0-indexed)
#>  [1]  0  2  4 11 12 13 19  1  5  6  9 10 18  3  7  8 14 15 16 17
letter_groups$group_names # Names of unique groups
#> [1] "A" "C" "B"

In real analysis, you normally wouldn’t need all of this metadata. For example, to return a data frame of unique groups (keys) and their associated counts, we only need starts() and counts().


template <RVector T>
[[cppally::register]]
r_df group_counts(T x){
  groups g = make_groups(x, /*ordered=*/ false);
  
  return make_df(
    arg("key") = subset(x, g.starts()),
    arg("count") = g.counts()
  );
}

group_counts(x)
#>   key count
#> 1   A     7
#> 2   C     6
#> 3   B     7

To illustrate how fast make_groups() is, let’s compare cppally-derived group counts of a relatively large sample to base-R derived group counts.

# 10k groups, 100k data points
x <- sample.int(10^4, 10^5, replace = TRUE)
mark(
  cppally_group_counts = group_counts(x)$count,
  base_table = table(factor(x, levels = unique(x))) |> as.integer()
)
#> # A tibble: 2 × 6
#>   expression                min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>           <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cppally_group_counts  216.6µs 222.29µs     4458.   514.1KB     59.4
#> 2 base_table             5.02ms   5.11ms      195.     6.2MB     42.1

Annex

Unsupported classes in R-registered templates

r_sym and r_function are both unsupported in R-registered template functions. This only applies when used as template arguments. They will work as normal where function arguments are non-templated.

Good:

template <RStringType T>
[[cppally::register]]
r_lgl foo(r_sym x, T y){
    return as<r_str>(x) == y;
}

Bad:

template <RStringType T, typename Symbol>
requires (is<Symbol, r_sym>)
[[cppally::register]]
r_lgl foo(Symbol x, T y){
    return as<r_str>(x) == y;
}

All core cppally concepts

RNumber - Includes r_int, r_int64 and r_dbl
RIntegerType - Includes r_lgl, r_int, r_int64
RIntegerNumber - Includes r_int and r_int64
RMathType - Includes r_lgl, r_int, r_int64 and r_dbl
RStringType - Includes r_str and r_str_view
RScalar - Includes r_lgl, r_int, r_int64, r_dbl, r_str, r_str_view, r_cplx, r_raw, r_date and r_psxct
RVal - Includes anything a cppally vector (r_vector<>) can contain: RScalar +r_sexp
RVector - Includes r_vector<T> where T is an RVal
RFactor - Factors
RDataFrame - Data frames
RComposite - Includes vectors, factors and data frames
RTimeType - Includes r_date and r_psxct
RNumericType - Numeric types, including RMathType and RTimeType
RSortableType - Includes RNumericType and RStringType (strings can also be sorted)
RAtomicVector - A vector that contains RScalar elements
CppallyType - Any R type defined by R, including RVal, RVector, RFactor, RDataFrame, RSymbol
CppType - Anything that is not an CppallyType
CastableToRScalar - Anything that can be constructed or cast into an RScalar (which also includes RScalar)

Other useful type traits

unwrap_t - Returns the underlying unwrapped type
as_r_scalar_t - Returns the equivalent RScalar type
as_r_composite_t - Returns the equivalent RComposite type
common_r_t - Returns the common cppally type between 2 types

Accessing the underlying types and values

While it is generally not recommended to access the underlying values, you can do so with unwrap() which returns the underlying C/C++ value. For example, unwrap(r_int(5)) will return an int of value 5.

To access the underlying type, use unwrap_t<> which always aligns with unwrap()

int one = unwrap(r_int(1));
double two = unwrap(r_dbl(2));

// Confirming that unwrap_t produces the required types as well
static_assert(std::is_same_v<int, unwrap_t<r_int>>);
static_assert(std::is_same_v<double, unwrap_t<r_dbl>>);

// unwrap_t<r_str> = SEXP
// cppally recommends against using BOTH the cppally API and the R C API together.
unwrap_t<r_str> r_c_api_blank_str = unwrap(r_str(""));