cheapr

In cheapr, ‘cheap’ means fast and memory-efficient, and that’s exactly the philosophy that cheapr aims to follow.

Installation

You can install cheapr like so:

install.packages("cheapr")

or you can install the development version of cheapr:

remotes::install_github("NicChr/cheapr")

Some common operations that cheapr can do much faster and more efficiently include:

• Counting, finding, removing and replacing NA and scalar values

• Creating factors

• Creating multiple sequences in a vectorised way

• Sub-setting vectors and data frames efficiently

• Safe, flexible and fast greatest common divisor and lowest common multiple

• Lags/leads

• Lightweight integer64 support

• In-memory Math (no copies, vectors updated by reference)

• Summary statistics of data frame variables

• Binning of continuous data

Let’s first load the required packages

library(cheapr)
library(bench)

num_na() is a useful function to efficiently return the number of NA values and can be used in a variety of problems.

Almost all the NA handling functions in cheapr can make use of multiple cores on your machine through openMP.

x <- rep(NA, 10^6)

# 1 core by default
mark(num_na(x), sum(is.na(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 num_na(x)        986µs   1.04ms      954.    2.41KB      0  
#> 2 sum(is.na(x))    780µs   1.75ms      575.    3.81MB     45.6
# 4 cores
options(cheapr.cores = 4)
mark(num_na(x), sum(is.na(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 num_na(x)        267µs  338.4µs     2881.        0B      0  
#> 2 sum(is.na(x))    782µs   1.75ms      577.    3.81MB     48.3
options(cheapr.cores = 4)

Efficient NA counts by row/col

m <- matrix(x, ncol = 10^3)
# Number of NA values by row
mark(row_na_counts(m), 
     rowSums(is.na(m)))
#> # A tibble: 2 × 6
#>   expression             min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>        <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 row_na_counts(m)    1.34ms   2.66ms      374.    9.11KB      0  
#> 2 rowSums(is.na(m))   3.48ms   3.69ms      267.    3.82MB     23.9
# Number of NA values by col
mark(col_na_counts(m), 
     colSums(is.na(m)))
#> # A tibble: 2 × 6
#>   expression             min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>        <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 col_na_counts(m)   721.5µs  894.9µs     1123.     9.1KB      0  
#> 2 colSums(is.na(m))   2.63ms   2.83ms      351.    3.82MB     31.9

is_na is a multi-threaded alternative to is.na

x <- rnorm(10^6)
x[sample.int(10^6, 10^5)] <- NA
mark(is.na(x), is_na(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 is.na(x)      739µs   1.79ms      564.    3.81MB     66.5
#> 2 is_na(x)      425µs 848.45µs     1184.    3.82MB    109.

### posixlt method is much faster
hours <- as.POSIXlt(seq.int(0, length.out = 10^6, by = 3600),
                    tz = "UTC")
hours[sample.int(10^6, 10^5)] <- NA

mark(is.na(hours), is_na(hours))
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
#> # A tibble: 2 × 6
#>   expression        min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 is.na(hours)    1.22s    1.22s     0.821      61MB    0.821
#> 2 is_na(hours)   5.26ms    7.5ms   111.       13.9MB   13.9

It differs in 2 regards:

• List elements are regarded as NA when either that element is an NA value or it is a list containing only NA values.
• For data frames, is_na returns a logical vector where TRUE defines an empty row of only NA values.

# List example
is.na(list(NA, list(NA, NA), 10))
#> [1]  TRUE FALSE FALSE
is_na(list(NA, list(NA, NA), 10))
#> [1]  TRUE  TRUE FALSE

# Data frame example
df <- data.frame(x = c(1, NA, 3),
                 y = c(NA, NA, NA))
df
#>    x  y
#> 1  1 NA
#> 2 NA NA
#> 3  3 NA
is_na(df)
#> [1] FALSE  TRUE FALSE
is_na(df)
#> [1] FALSE  TRUE FALSE
# The below identity should hold
identical(is_na(df), row_na_counts(df) == ncol(df))
#> [1] TRUE

is_na and all the NA handling functions fall back on calling is.na() if no suitable method is found. This means that custom objects like vctrs rcrds and more are supported.

Cheap data frame summaries with overview

Inspired by the excellent skimr package, overview() is a cheaper alternative designed for larger data.

df <- data.frame(
  x = sample.int(100, 10^7, TRUE),
  y = factor_(sample(LETTERS, 10^7, TRUE)),
  z = rnorm(10^7)
)
overview(df, hist = TRUE)
#> obs: 10000000 
#> cols: 3 
#> 
#> ----- Numeric -----
#>   col   class n_missing p_complete n_unique  mean    p0   p25 p50  p75 p100
#> 1   x integer         0          1      100 50.52     1    26  51   76  100
#> 2   z numeric         0          1 10000000     0 -4.95 -0.67   0 0.67 4.97
#>    iqr    sd  hist
#> 1   50 28.87 ▇▇▇▇▇
#> 2 1.35     1 ▁▂▇▂▁
#> 
#> ----- Categorical -----
#>   col  class n_missing p_complete n_unique n_levels min max
#> 1   y factor         0          1       26       26   A   Z
mark(overview(df))
#> # A tibble: 1 × 6
#>   expression        min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 overview(df)    1.26s    1.26s     0.793    2.09KB        0

Cheaper and consistent subsetting with sset

sset(iris, 1:5)
#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
#> 1          5.1         3.5          1.4         0.2  setosa
#> 2          4.9         3.0          1.4         0.2  setosa
#> 3          4.7         3.2          1.3         0.2  setosa
#> 4          4.6         3.1          1.5         0.2  setosa
#> 5          5.0         3.6          1.4         0.2  setosa
sset(iris, 1:5, j = "Species")
#>   Species
#> 1  setosa
#> 2  setosa
#> 3  setosa
#> 4  setosa
#> 5  setosa

# sset always returns a data frame when input is a data frame

sset(iris, 1, 1) # data frame
#>   Sepal.Length
#> 1          5.1
iris[1, 1] # not a data frame
#> [1] 5.1

x <- sample.int(10^6, 10^4, TRUE)
y <- sample.int(10^6, 10^4, TRUE)
mark(sset(x, x %in_% y), sset(x, x %in% y), x[x %in% y])
#> # A tibble: 3 × 6
#>   expression              min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>         <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 sset(x, x %in_% y)   82.5µs    122µs     8488.    88.1KB     6.66
#> 2 sset(x, x %in% y)   152.8µs    230µs     4388.   285.3KB     8.80
#> 3 x[x %in% y]         130.5µs    210µs     4824.   324.4KB    14.2

sset uses an internal range-based subset when i is an ALTREP integer sequence of the form m:n.

mark(sset(df, 0:10^5), df[0:10^5, , drop = 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 sset(df, 0:10^5)            132.5µs  518.8µs     1944.    1.53MB    26.8 
#> 2 df[0:10^5, , drop = FALSE]    6.1ms   7.28ms      136.    4.83MB     6.57

It also accepts negative indexes

mark(sset(df, -10^4:0), 
     df[-10^4:0, , drop = FALSE],
     check = FALSE) # The only difference is the row names
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
#> # A tibble: 2 × 6
#>   expression                       min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>                  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 sset(df, -10^4:0)             47.8ms   49.5ms     15.8      152MB     5.94
#> 2 df[-10^4:0, , drop = FALSE]  748.5ms  748.5ms      1.34     776MB     4.01

The biggest difference between sset and [ is the way logical vectors are handled. The two main differences when i is a logical vector are:

• NA values are ignored, only the locations of TRUE values are used.
• i must be the same length as x and is not recycled.

# Examples with NAs
x <- c(1, 5, NA, NA, -5)
x[x > 0]
#> [1]  1  5 NA NA
sset(x, x > 0)
#> [1] 1 5

# Example with length(i) < length(x)
sset(x, TRUE)
#> Error in check_length(i, length(x)): i must have length 5

# This is equivalent 
x[TRUE]
#> [1]  1  5 NA NA -5
# to..
sset(x)
#> [1]  1  5 NA NA -5

Vector and data frame lags with lag_()

set.seed(37)
lag_(1:10, 3) # Lag(3)
#>  [1] NA NA NA  1  2  3  4  5  6  7
lag_(1:10, -3) # Lead(3)
#>  [1]  4  5  6  7  8  9 10 NA NA NA

# Using an example from data.table
library(data.table)
#> Warning: package 'data.table' was built under R version 4.4.1
dt <- data.table(year=2010:2014, v1=runif(5), v2=1:5, v3=letters[1:5])

# Similar to data.table::shift()

lag_(dt, 1) # Lag 
#>     year         v1    v2     v3
#>    <int>      <num> <int> <char>
#> 1:    NA         NA    NA   <NA>
#> 2:  2010 0.54964085     1      a
#> 3:  2011 0.07883715     2      b
#> 4:  2012 0.64879698     3      c
#> 5:  2013 0.49685336     4      d
lag_(dt, -1) # Lead
#>     year         v1    v2     v3
#>    <int>      <num> <int> <char>
#> 1:  2011 0.07883715     2      b
#> 2:  2012 0.64879698     3      c
#> 3:  2013 0.49685336     4      d
#> 4:  2014 0.71878731     5      e
#> 5:    NA         NA    NA   <NA>

With lag_ we can update variables by reference, including entire data frames

# At the moment, shift() cannot do this
lag_(dt, set = TRUE)
#>     year         v1    v2     v3
#>    <int>      <num> <int> <char>
#> 1:    NA         NA    NA   <NA>
#> 2:  2010 0.54964085     1      a
#> 3:  2011 0.07883715     2      b
#> 4:  2012 0.64879698     3      c
#> 5:  2013 0.49685336     4      d

dt # Was updated by reference
#>     year         v1    v2     v3
#>    <int>      <num> <int> <char>
#> 1:    NA         NA    NA   <NA>
#> 2:  2010 0.54964085     1      a
#> 3:  2011 0.07883715     2      b
#> 4:  2012 0.64879698     3      c
#> 5:  2013 0.49685336     4      d

lag2_ is a more generalised variant that supports vectors of lags, custom ordering and run lengths.

lag2_(dt, order = 5:1) # Reverse order lag (same as lead)
#>     year         v1    v2     v3
#>    <int>      <num> <int> <char>
#> 1:  2010 0.54964085     1      a
#> 2:  2011 0.07883715     2      b
#> 3:  2012 0.64879698     3      c
#> 4:  2013 0.49685336     4      d
#> 5:    NA         NA    NA   <NA>
lag2_(dt, -1) # Same as above
#>     year         v1    v2     v3
#>    <int>      <num> <int> <char>
#> 1:  2010 0.54964085     1      a
#> 2:  2011 0.07883715     2      b
#> 3:  2012 0.64879698     3      c
#> 4:  2013 0.49685336     4      d
#> 5:    NA         NA    NA   <NA>
lag2_(dt, c(1, -1)) # Alternating lead/lag
#>     year         v1    v2     v3
#>    <int>      <num> <int> <char>
#> 1:    NA         NA    NA   <NA>
#> 2:  2011 0.07883715     2      b
#> 3:  2010 0.54964085     1      a
#> 4:  2013 0.49685336     4      d
#> 5:  2012 0.64879698     3      c
lag2_(dt, c(-1, 0, 0, 0, 0)) # Lead e.g. only first row
#>     year         v1    v2     v3
#>    <int>      <num> <int> <char>
#> 1:  2010 0.54964085     1      a
#> 2:  2010 0.54964085     1      a
#> 3:  2011 0.07883715     2      b
#> 4:  2012 0.64879698     3      c
#> 5:  2013 0.49685336     4      d

Greatest common divisor and smallest common multiple

gcd2(5, 25)
#> [1] 5
scm2(5, 6)
#> [1] 30

gcd(seq(5, 25, by = 5))
#> [1] 5
scm(seq(5, 25, by = 5))
#> [1] 300

x <- seq(1L, 1000000L, 1L)
mark(gcd(x))
#> # A tibble: 1 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 gcd(x)        1.2µs    1.5µs   626912.        0B        0
x <- seq(0, 10^6, 0.5)
mark(gcd(x))
#> # A tibble: 1 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 gcd(x)       51.5ms   51.8ms      19.2        0B        0

Creating many sequences

As an example, to create 3 sequences with different increments,
the usual approach might be to use lapply to loop through the increment values together with seq()

# Base R
increments <- c(1, 0.5, 0.1)
start <- 1
end <- 5
unlist(lapply(increments, \(x) seq(start, end, x)))
#>  [1] 1.0 2.0 3.0 4.0 5.0 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 1.0 1.1 1.2 1.3 1.4
#> [20] 1.5 1.6 1.7 1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3
#> [39] 3.4 3.5 3.6 3.7 3.8 3.9 4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 5.0

In cheapr you can use seq_() which accepts vector arguments.

seq_(start, end, increments)
#>  [1] 1.0 2.0 3.0 4.0 5.0 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 1.0 1.1 1.2 1.3 1.4
#> [20] 1.5 1.6 1.7 1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3
#> [39] 3.4 3.5 3.6 3.7 3.8 3.9 4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 5.0

Use add_id = TRUE to label the individual sequences.

seq_(start, end, increments, add_id = TRUE)
#>   1   1   1   1   1   2   2   2   2   2   2   2   2   2   3   3   3   3   3   3 
#> 1.0 2.0 3.0 4.0 5.0 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 1.0 1.1 1.2 1.3 1.4 1.5 
#>   3   3   3   3   3   3   3   3   3   3   3   3   3   3   3   3   3   3   3   3 
#> 1.6 1.7 1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 
#>   3   3   3   3   3   3   3   3   3   3   3   3   3   3   3 
#> 3.6 3.7 3.8 3.9 4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 5.0

If you know the sizes of your sequences beforehand, use sequence_()

seq_sizes <- c(3, 5, 10)
sequence_(seq_sizes, from = 0, by = 1/3, add_id = TRUE) |> 
  enframe_()
#> # A tibble: 18 × 2
#>    name  value
#>    <chr> <dbl>
#>  1 1     0    
#>  2 1     0.333
#>  3 1     0.667
#>  4 2     0    
#>  5 2     0.333
#>  6 2     0.667
#>  7 2     1    
#>  8 2     1.33 
#>  9 3     0    
#> 10 3     0.333
#> 11 3     0.667
#> 12 3     1    
#> 13 3     1.33 
#> 14 3     1.67 
#> 15 3     2    
#> 16 3     2.33 
#> 17 3     2.67 
#> 18 3     3

You can also calculate the sequence sizes using seq_size()

seq_size(start, end, increments)
#> [1]  5  9 41

‘Cheaper’ Base R alternatives

which

x <- rep(TRUE, 10^6)
mark(cheapr_which = which_(x),
     base_which = which(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 cheapr_which   2.03ms   3.41ms      282.    3.81MB     2.09
#> 2 base_which    629.4µs   2.62ms      397.    7.63MB     9.62
x <- rep(FALSE, 10^6)
mark(cheapr_which = which_(x),
     base_which = which(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 cheapr_which    230µs    290µs     3230.        0B      0  
#> 2 base_which      456µs    480µs     2055.    3.81MB     20.1
x <- c(rep(TRUE, 5e05), rep(FALSE, 1e06))
mark(cheapr_which = which_(x),
     base_which = which(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 cheapr_which   1.24ms      2ms      488.    1.91MB     2.08
#> 2 base_which      781µs   1.74ms      588.    7.63MB    12.0
x <- c(rep(FALSE, 5e05), rep(TRUE, 1e06))
mark(cheapr_which = which_(x),
     base_which = which(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 cheapr_which   2.98ms   4.25ms      236.    3.81MB     2.09
#> 2 base_which    938.1µs    3.1ms      330.    9.54MB     9.22
x <- sample(c(TRUE, FALSE), 10^6, TRUE)
x[sample.int(10^6, 10^4)] <- NA
mark(cheapr_which = which_(x),
     base_which = which(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 cheapr_which   1.93ms   2.56ms      374.    1.89MB     2.09
#> 2 base_which     3.19ms    4.2ms      240.     5.7MB     2.07

factor

x <- sample(seq(-10^3, 10^3, 0.01))
y <- do.call(paste0, expand.grid(letters, letters, letters, letters))
mark(cheapr_factor = factor_(x), 
     base_factor = factor(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 cheapr_factor   9.19ms   9.74ms    102.      4.59MB        0
#> 2 base_factor   490.32ms 494.23ms      2.02   27.84MB        0
mark(cheapr_factor = factor_(x, order = FALSE), 
     base_factor = factor(x, levels = 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 cheapr_factor   4.37ms   5.08ms    192.      1.53MB        0
#> 2 base_factor    752.8ms  752.8ms      1.33   22.79MB        0
mark(cheapr_factor = factor_(y), 
     base_factor = factor(y))
#> # A tibble: 2 × 6
#>   expression         min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>    <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cheapr_factor 208.06ms 211.32ms     4.60     5.23MB        0
#> 2 base_factor      3.23s    3.23s     0.309   54.35MB        0
mark(cheapr_factor = factor_(y, order = FALSE), 
     base_factor = factor(y, levels = unique(y)))
#> # A tibble: 2 × 6
#>   expression         min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>    <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cheapr_factor   4.97ms   8.01ms     119.     3.49MB        0
#> 2 base_factor    53.61ms  59.46ms      16.9   39.89MB        0

intersect & setdiff

x <- sample.int(10^6, 10^5, TRUE)
y <- sample.int(10^6, 10^5, TRUE)
mark(cheapr_intersect = intersect_(x, y, dups = FALSE),
     base_intersect = intersect(x, y))
#> # A tibble: 2 × 6
#>   expression            min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>       <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cheapr_intersect   2.71ms   2.89ms      337.    1.18MB     2.30
#> 2 base_intersect     4.91ms   5.28ms      182.    5.16MB     0
mark(cheapr_setdiff = setdiff_(x, y, dups = FALSE),
     base_setdiff = setdiff(x, y))
#> # A tibble: 2 × 6
#>   expression          min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>     <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cheapr_setdiff   2.97ms   3.15ms      315.    1.76MB     0   
#> 2 base_setdiff     4.97ms   5.51ms      179.    5.71MB     2.29

%in_% and %!in_%

mark(cheapr = x %in_% y,
     base = x %in% y)
#> # A tibble: 2 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cheapr       1.75ms   1.83ms      539.  781.34KB     0   
#> 2 base         2.52ms   3.08ms      325.    2.53MB     2.24
mark(cheapr = x %!in_% y,
     base = !x %in% y)
#> # A tibble: 2 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cheapr       1.68ms   1.82ms      543.  787.84KB     0   
#> 2 base         2.82ms   3.25ms      306.    2.91MB     2.25

as_discrete

as_discrete is a cheaper alternative to cut

x <- rnorm(10^7)
b <- seq(0, max(x), 0.2)
mark(cheapr_cut = as_discrete(x, b, left = FALSE), 
     base_cut = cut(x, b))
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
#> # A tibble: 2 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 cheapr_cut    216ms    219ms      4.54    38.2MB     0   
#> 2 base_cut      642ms    642ms      1.56   267.1MB     1.56