Benchmark of stringr replacement

Context
- Session info
- Helpers
Benchmarking

Context

As part of #1549, we removed stringr dependency by replacing all functions used with base R equivalent.

This document aim to be a benchmark of each replacement. We’ll use bench to do the benchmark, and compare stringr function with its replacement. Each benchmark will be done in the context of there initial usage in knitr.

Session info

xfun::session_info(c("knitr", "bench", "stringr"))

## R version 4.2.0 (2022-04-22 ucrt)
## Platform: x86_64-w64-mingw32/x64 (64-bit)
## Running under: Windows 10 x64 (build 22621)
## 
## Locale:
##   LC_COLLATE=French_France.utf8  LC_CTYPE=French_France.utf8   
##   LC_MONETARY=French_France.utf8 LC_NUMERIC=C                  
##   LC_TIME=French_France.utf8    
## 
## Package version:
##   bench_1.1.2     cli_3.6.0       evaluate_0.20   fansi_1.0.3    
##   glue_1.6.2      graphics_4.2.0  grDevices_4.2.0 highr_0.10     
##   knitr_1.41.9    lifecycle_1.0.3 magrittr_2.0.3  methods_4.2.0  
##   pillar_1.8.1    pkgconfig_2.0.3 profmem_0.6.0   rlang_1.0.6    
##   stats_4.2.0     stringi_1.7.12  stringr_1.5.0   tibble_3.1.8   
##   tools_4.2.0     utf8_1.2.2      utils_4.2.0     vctrs_0.5.1    
##   xfun_0.36       yaml_2.3.6

Helpers

knitr_example = function(...) system.file('examples', ..., package = 'knitr')

Benchmarking

`stringr::str_replace_all()`

Commit	PR	File
9c92eff1	#2174	utils-vignettes.R#L158

Before:

x[!i] = stringr::str_replace_all(x[!i], p$inline.code, '') # remove inline code

After:

x[!i] = gsub(p$inline.code, '', x[!i], perl = TRUE) # remove inline code

file = knitr_example('knitr-minimal.Rnw')
x = xfun::read_utf8(file)
p = knitr:::detect_pattern(x, tolower(xfun::file_ext(file)))
p = knitr:::all_patterns[[p]]
p1 = p$chunk.begin; p2  = p$chunk.end
m = knitr:::group_indices(grepl(p1, x), grepl(p2, x))
i = m %% 2 == 0

res = bench::mark(
  stringr = stringr::str_replace_all(x[!i], p$inline.code, ''),
  new = gsub(p$inline.code, '', x[!i], perl = TRUE),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1.29   1.38      1        1226.     5.99
## 2 new         1      1         1.34        1      1

ggplot2::autoplot(res)

## Le chargement a nécessité le package : tidyr

`stringr::str_trim()`

Commit

PR

File

Before:

labels = stringr::str_trim(gsub(lab, '\\3', sapply(groups, `[`, 1)))

After:

x[!i] = gsub(p$inline.code, '', x[!i], perl = TRUE) # remove inline code

path = "https://raw.githubusercontent.com/yihui/knitr-examples/46c8d1db0cf0c9ab04432444079927324c4c3688/113-foo.R"
lines = xfun::read_utf8(path)
lab = knitr:::.sep.label
idx = cumsum(grepl(lab, lines))
groups = unname(split(lines, idx))

content = gsub(lab, '\\3', sapply(groups, `[`, 1))

res = bench::mark(
  stringr = stringr::str_trim(content),
  new = trimws(content),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         2.33       Inf     7.76
## 2 new         2.48   2.54      1          NaN     1

ggplot2::autoplot(res)

Commit	PR	File
8fa7d17	#2177	utils.R#L77

Before:

stringr::str_trim(stringr::str_split(string, ';|,')[[1]])

After:

trimws(stringr::str_split(string, ';|,')[[1]])

string = ' .5,.6 , .7; .9 '
splitted = stringr::str_split(string, ';|,')[[1]]

res = bench::mark(
  stringr = stringr::str_trim(splitted),
  new = trimws(splitted),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         2.32       Inf     7.74
## 2 new         2.47   2.48      1          NaN     1

ggplot2::autoplot(res)

`stringr::str_dup()`

Commit	PR	File
1ce8286	#2186	parser.R#L348 \|

Before:

cat(' ', stringr::str_dup('~', getOption('width') - 10L), '\n')

After:

rep_str = function(x, n, sep = '') paste(rep(x, n), collapse = sep)
cat(' ', rep_str('~', getOption('width') - 10L), '\n')

res = bench::mark(
  stringr = stringr::str_dup('~', getOption('width') - 10L),
  new = paste(rep('~', getOption('width') - 10L), collapse = ''),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         1.20      5.96      Inf
## 2 new         1.17   1.14      1         1         NaN

ggplot2::autoplot(res)

Later, usage of stringr::str_dup() and its replacement rep_str() have been removed completely in (https://github.com/yihui/knitr/commit/0efb7914a1be93460e510bb283f9d4d72c49b360) as code was simplified in https://github.com/yihui/knitr/commit/07bf3adf49f67bd528be4e6a60b3284a7ea6f8a9 and https://github.com/yihui/knitr/commit/7ef7be00711de949fcf2a299d6381189669bb753 so no more impact.

`stringr::str_split()`

Commit

PR

File

Before:

x = stringr::str_split(x, '\n')

After:

# patch strsplit() to split '' into '' instead of character(0)
str_split = function(x, split, ...) {
  y = strsplit(x, split, ...)
  y[x == ''] = list('')
  y
}

x = str_split(x, '\n')

con = url("https://raw.githubusercontent.com/yihui/knitr-examples/46c8d1db0cf0c9ab04432444079927324c4c3688/084-pandoc.pandoc")
x = read.dcf(con)
close(con)
x = x[!is.na(x)]

str_split = function(x, split, ...) {
  y = strsplit(x, split, ...)
  y[x == ''] = list('')
  y
}
res = bench::mark(
  stringr = stringr::str_split(x, '\n'),
  new = str_split(x, '\n'),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     2.48   2.48      1         2.38     1   
## 2 new         1      1         2.38      1        1.19

ggplot2::autoplot(res)

`stringr::str_sub()`

Commit	PR	File
5a2cb72	#2195	utils-string.R block.R#L564 utils.R#L92-L93

Before:

stringr::str_sub(input, loc[i, 1], loc[i, 2]) = if (length(res)) {
      paste(hook(res), collapse = '')
    } else ''

After:

# replace parts of a string with new values; `pos` is a matrix of positions and
# each row is a pair of [start, end]
str_replace = function(x, pos, value) {
  if (length(x) != 1) stop("Only a character scalar is supported.")
  # extract parts of the string that are outside [start, end]
  m = rbind(pos[, 1] - 1, pos[, 2] + 1)
  m = matrix(c(1, m, nchar(x)), nrow = 2)
  y = substring(x, m[1, ], m[2, ])
  paste(rbind(y, c(value, '')), collapse = '')
}

str_replace(input, block$location, ans)

input = "inline `r (function() 1)()`"
location = matrix(c(8,27), ncol = 2, byrow = TRUE)
ans = "1"

str_replace = function(x, pos, value) {
  if (length(x) != 1) stop("Only a character scalar is supported.")
  # extract parts of the string that are outside [start, end]
  m = rbind(pos[, 1] - 1, pos[, 2] + 1)
  m = matrix(c(1, m, nchar(x)), nrow = 2)
  y = substring(x, m[1, ], m[2, ])
  paste(rbind(y, c(value, '')), collapse = '')
}

# wrapper to approximate what is done internally
old_replace = function(input, loc, ans) {
  stringr::str_sub(input, loc[1, 1], loc[1, 2]) = ans
  return(input)
}

res = bench::mark(
  stringr = old_replace(input, location, ans),
  new = str_replace(input, location, ans),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         1.21      9.19     1.21
## 2 new         1.25   1.20      1         1        1

ggplot2::autoplot(res)

Commit	PR	File
5a2cb72	#2195	header.R#L59-L60 header.R#L90-L91

Before:

tmp = stringr::str_sub(doc[i], l[, 1], l[, 2])
stringr::str_sub(doc[i], l[,1], l[,2]) = paste0(tmp, '\n', make_header_html())

After :

tmp = substr(doc[i], l[, 1], l[, 2])
doc[i] = str_replace(doc[i], l, paste0(tmp, '\n', make_header_html()))

b = knitr::all_patterns$html$header.begin
doc = c("<head>", "<!--content-->", "</head>", "<body>", "</body>")
i = grep(b, doc)
l = stringr::str_locate(doc[i], b)

res = bench::mark(
  stringr = stringr::str_sub(doc[i], l[, 1], l[, 2]),
  new = substr(doc[i], l[, 1], l[, 2]),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     3.48   3.45      1          Inf     1   
## 2 new         1      1         3.59       NaN     3.59

ggplot2::autoplot(res)

tmp = stringr::str_sub(doc[i], l[, 1], l[, 2])
header = knitr:::make_header_html()

old_fun = function(doc, l, tmp, header) {
  stringr::str_sub(doc[i], l[,1], l[,2]) = paste0(tmp, header)
  doc
}

new_fun = function(doc, l, tmp, header) {
  doc[i] = str_replace(doc[i], l, paste0(tmp, header))
  doc
}

res = bench::mark(
  stringr = old_fun(doc, l, tmp, header),
  new = new_fun(doc, l, tmp, header),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         1.13       Inf     1   
## 2 new         1.19   1.17      1          NaN     1.33

ggplot2::autoplot(res)

Before:

tmp = stringr::str_sub(doc[i], l[, 1], l[, 2])
stringr::str_sub(doc[i], l[,1], l[,2]) = paste0(tmp, make_header_latex(doc))

After :

tmp = substr(doc[i], l[, 1], l[, 2])
doc[i] = str_replace(doc[i], l, paste0(tmp, make_header_latex(doc)))

b = knitr::all_patterns$tex$header.begin
doc = c("\\documentclass[opt]{article}", "some tex content")
i = grep(b, doc)
l = stringr::str_locate(doc[i], b)

res = bench::mark(
  stringr = stringr::str_sub(doc[i], l[, 1], l[, 2]),
  new = substr(doc[i], l[, 1], l[, 2]),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     3.55   3.39      1          Inf     1   
## 2 new         1      1         3.42       NaN     3.42

ggplot2::autoplot(res)

tmp = stringr::str_sub(doc[i], l[, 1], l[, 2])
header = knitr:::make_header_latex(doc)

old_fun = function(doc, l, tmp, header) {
  stringr::str_sub(doc[i], l[,1], l[,2]) = paste0(tmp, header)
  doc
}

new_fun = function(doc, l, tmp, header) {
  doc[i] = str_replace(doc[i], l, paste0(tmp, header))
  doc
}

res = bench::mark(
  stringr = old_fun(doc, l, tmp, header),
  new = new_fun(doc, l, tmp, header),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         1.20       Inf     1   
## 2 new         1.25   1.25      1          NaN     1.25

ggplot2::autoplot(res)

Variant replacement `str_insert()` instead or `str_replace()`

Commit	PR	File
5a2cb72	#2195	header.R#L59-L60 header.R#L90-L91

Before:

tmp = stringr::str_sub(doc[i], l[, 1], l[, 2])
stringr::str_sub(doc[i], l[,1], l[,2]) = paste0(tmp, '\n', make_header_html())

After :

doc[i] = str_insert(doc[i], l[, 2], paste0('\n', make_header_html()))

b = knitr::all_patterns$html$header.begin
doc = c("<head>", "<!--content-->", "</head>", "<body>", "</body>")
i = grep(b, doc)
l = stringr::str_locate(doc[i], b)

header = knitr:::make_header_html()

str_insert = function(x, i, value) {
  if (i <= 0) return(paste0(value, x))
  n = nchar(x)
  if (n == 0 || i >= n) return(paste0(x, value))
  paste0(substr(x, 1, i), value, substr(x, i + 1, n))
}

old_str_sub = function(doc, l, header) {
  tmp = stringr::str_sub(doc[i], l[, 1], l[, 2])
  stringr::str_sub(doc[i], l[,1], l[,2]) = paste0(tmp, header)
  doc
}

new_str_replace = function(doc, l, header) {
  tmp = substr(doc[i], l[, 1], l[, 2])
  doc[i] = str_replace(doc[i], l, paste0(tmp, header))
  doc
}

new_str_insert = function(doc, l, header) {
  doc[i] = str_insert(doc[i], l[, 2], paste0('\n', header))
  doc
}

res = bench::mark(
  stringr = old_str_sub(doc, l, header),
  str_replace = new_str_replace(doc, l, header),
  str_insert = new_str_insert(doc, l, header),
  min_time = Inf,
  # we get a small difference in output
  check = FALSE
)
summary(res, relative = TRUE)

## # A tibble: 3 × 6
##   expression    min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr>  <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr      2.81   2.74      1          Inf     1.28
## 2 str_replace  2.47   2.41      1.11       NaN     1.89
## 3 str_insert   1      1         2.34       NaN     1

ggplot2::autoplot(res)

Special edge case here: Our replacement function is not completely equivalent.

New version gives <head>\n<style type="text/css">\n\n</style>
stringr version gives <head><style type="text/css">\n\n</style>

Before:

tmp = stringr::str_sub(doc[i], l[, 1], l[, 2])
stringr::str_sub(doc[i], l[,1], l[,2]) = paste0(tmp, make_header_latex(doc))

After :

doc[i] = str_insert(doc[i], l[, 2], make_header_latex(doc))

b = knitr::all_patterns$tex$header.begin
doc = c("\\documentclass[opt]{article}", "some tex content")
i = grep(b, doc)
l = stringr::str_locate(doc[i], b)

header = knitr:::make_header_latex(doc)

old_str_sub = function(doc, l, header) {
  tmp = stringr::str_sub(doc[i], l[, 1], l[, 2])
  stringr::str_sub(doc[i], l[,1], l[,2]) = paste0(tmp, header)
  doc
}

new_str_replace = function(doc, l, header) {
  tmp = substr(doc[i], l[, 1], l[, 2])
  doc[i] = str_replace(doc[i], l, paste0(tmp, header))
  doc
}

new_str_insert = function(doc, l, header) {
  doc[i] = str_insert(doc[i], l[, 2], header)
  doc
}

res = bench::mark(
  stringr = old_str_sub(doc, l, header),
  str_replace = new_str_replace(doc, l, header),
  str_insert = new_str_insert(doc, l, header),
  min_time = Inf
)

summary(res, relative = TRUE)

## # A tibble: 3 × 6
##   expression    min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr>  <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr      3.34   3.31      1          Inf      Inf
## 2 str_replace  3.05   3.02      1.14       NaN      Inf
## 3 str_insert   1      1         3.44       NaN      NaN

ggplot2::autoplot(res)

`stringr::str_detect()`

Commit	PR	File
1a0f2cc	#2202	pandoc.R#L125

Before:

if (length(pat) && any(stringr::str_detect(text, pat))) return(p)

After:

if (length(pat) && any(grepl(pat, text, perl = TRUE))) return(p)

pat = knitr:::all_patterns$md$chunk.begin
text = xfun::read_utf8(knitr_example("knitr-spin.Rmd"))

res = bench::mark(
  stringr = stringr::str_detect(text, pat),
  new = grepl(pat, text, perl = TRUE),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         2.13      76.1      Inf
## 2 new         2.20   2.24      1          1        NaN

ggplot2::autoplot(res)

`stringr::str_wrap()`

Commit	PR	File
1a0f2cc	#2202	citation.R#L135

Before:

b[-idx] = stringr::str_wrap(b[-idx], width, 2, 4)

After:

str_wrap = function(...) {
  res = strwrap(..., simplify = FALSE)
  unlist(lapply(res, one_string))
  }
b[-idx] = str_wrap(b[-idx], width, 2, 4)

x = "knitr"
lib.loc = NULL
tweak = TRUE
prefix = 'R-'
.tweak.bib = knitr:::.tweak.bib
citation = function(...) utils::citation(..., lib.loc = lib.loc)
bib = sapply(x, function(pkg) {
  cite = citation(pkg, auto = if (pkg != 'base') {
    meta = packageDescription(pkg, lib.loc = lib.loc)
    # don't use the CRAN URL if the package has provided its own URL
    if (identical(meta$Repository, 'CRAN') && !is.null(meta$URL)) {
      # however, the package may have provided multiple URLs, in which case we
      # still use the CRAN URL
      if (!grepl('[, ]', meta$URL)) meta$Repository = NULL
    }
    meta
  })
  entry = toBibtex(cite)
  entry[1] = sub('\\{,$', sprintf('{%s%s,', prefix, pkg), entry[1])
  entry
}, simplify = FALSE)

str_wrap = function(...) {
  res = strwrap(..., simplify = FALSE)
  unlist(lapply(res, knitr:::one_string))
}
res = bench::mark(
  stringr = stringr::str_wrap(bib$knitr, 20, 2, 4),
  new = str_wrap(bib$knitr, 20, 2, 4),
  min_time = Inf, 
  check = FALSE # functions are not completely equivalent
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         5.20      1.74     4.87
## 2 new         5.12   5.48      1         1        1

ggplot2::autoplot(res)

Special edge case here: Our replacement function is not completely equivalent.

New version gives note = {R package\n version\n 1.41.9},
stringr version gives note = {R package\n version 1.41.9},

Commit	PR	File
1a0f2cc	#2202	output.R#L503 \|

Before:

stringr::str_wrap(message, width = getOption('width'))

After:

str_wrap = function(...) {
  res = strwrap(..., simplify = FALSE)
  unlist(lapply(res, one_string))
  }
str_wrap(message, width = getOption('width'))

warn_msg = "Warning function: This is a warning message quite loooooooooooong"

str_wrap = function(...) {
  res = strwrap(..., simplify = FALSE)
  unlist(lapply(res, knitr:::one_string))
}
res = bench::mark(
  stringr = stringr::str_wrap(warn_msg, 10),
  new = str_wrap(warn_msg, 10),
  min_time = Inf, 
  check = FALSE # functions are not completely equivalent
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         2.12       Inf     2.33
## 2 new         2.14   2.23      1          NaN     1

ggplot2::autoplot(res)

Special edge case here: Our replacement function is not completely equivalent.

New version gives Warning\nfunction:\nThis is\na warning\nmessage\nquite\nloooooooooooong
stringr version gives Warning\nfunction:\nThis is a\nwarning\nmessage\nquite\nloooooooooooong

`stringr::str_pad()`

Commit	PR	File
1a0f2cc	#2202	block.R#L619

Before:

paste0('## ----', stringr::str_pad(label, max(getOption('width') - 11L, 0L), 'right', '-'), '----', code)

After:

paste0('## ----', label, strrep('-', max(getOption('width') - 11L - nchar(label), 0L)), '----', code)

label = "my-first-chunk"

res = bench::mark(
  stringr = paste0('## ----', stringr::str_pad(label, max(getOption('width') - 11L, 0L), 'right', '-'), '----'),
  new = paste0('## ----', label, strrep('-', max(getOption('width') - 11L - nchar(label), 0L)), '----'),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     13.0   12.9       1         Inf     1   
## 2 new          1      1        13.9       NaN     1.16

ggplot2::autoplot(res)

`stringr::str_count()`

Commit	PR	File
1a0f2cc	#2202	utils.R#L583-L590

Before:

line_count_str = function(x) stringr::str_count(x, '\n') + 1L

After:

line_count = function(x) {
  res = gregexpr('\n', x, fixed = TRUE)
  unlist(lapply(res, function(x) {
    n = length(x)
    if (n == 1 && x == -1) n = 0
    n + 1
  }))
}

text = xfun::read_utf8(knitr_example("knitr-spin.Rmd"))

res = bench::mark(
  stringr = line_count_str(text),
  new = line_count(text),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         2.83       1        1  
## 2 new         2.27   2.97      1         23.7     18.4

ggplot2::autoplot(res)

`stringr::str_extract_all`

Commit	PR	File
cc3b92a	#2205	block.R#L587

Before:

eval(parse_only(unlist(stringr::str_extract_all(code, 'read_chunk\\(([^)]+)\\)'))))

After:

str_extract = function(x, pattern) {
  m = gregexpr(pattern, x, perl = TRUE)
  regmatches(x, m)
}
eval(parse_only(unlist(str_extract(code, 'read_chunk\\(([^)]+)\\)'))))

code = "knitr::read_chunk('113-foo.R')"

str_extract = function(x, pattern) {
  m = gregexpr(pattern, x, perl = TRUE)
  regmatches(x, m)
}
res = bench::mark(
  stringr = stringr::str_extract_all(code, 'read_chunk\\(([^)]+)\\)'),
  new = str_extract(code, 'read_chunk\\(([^)]+)\\)'),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         3.75      1        1.50
## 2 new         3.81   3.98      1         4.83     1

ggplot2::autoplot(res)

Commit	PR	File
cc3b92a	#2205	template.R#L123

Before:

mat = stringr::str_extract_all(txt, delim)[[1L]]

After:

str_extract = function(x, pattern) {
  m = gregexpr(pattern, x, perl = TRUE)
  regmatches(x, m)
}
mat = str_extract(txt, delim)[[1L]]

txt = 'This is the value of `x`: {{x}}'
delim = c('{{', '}}')
delim = gsub('([.|()\\^{}+$*?]|\\[|\\])', '\\\\\\1', delim)
delim = paste0(delim[1L], '((.|\n)+?)', delim[2L])

str_extract = function(x, pattern) {
  m = gregexpr(pattern, x, perl = TRUE)
  regmatches(x, m)
}
res = bench::mark(
  stringr = stringr::str_extract_all(txt, delim),
  new = str_extract(txt, delim),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         4.06       1       1.35
## 2 new         4.01   4.30      1         93.8     1

ggplot2::autoplot(res)

`stringr::str_locate()`

Commit	PR	File
cc3b92a	#2205	header.R#L58 header.R#L89

Before:

stringr::str_locate(doc[i], b)

After:

location = function(x) {
  len = attr(x, 'match.length')
  if (length(x) == 1 && x == -1) x = integer()
  cbind(start = x, end = x + len - 1L)
}
str_locate = function(x, pattern, all = TRUE) {
  out = (if (all) gregexpr else regexpr)(pattern, x, perl = TRUE)
  if (all) lapply(out, location) else location(out)
}
str_locate(doc[i], b, FALSE)

b = knitr::all_patterns$html$header.begin
doc = c("<head>", "<!--content-->", "</head>", "<body>", "</body>")
i = grep(b, doc)
location = function(x) {
  len = attr(x, 'match.length')
  if (length(x) == 1 && x == -1) x = integer()
  cbind(start = x, end = x + len - 1L)
}
str_locate = function(x, pattern, all = TRUE) {
  out = (if (all) gregexpr else regexpr)(pattern, x, perl = TRUE)
  if (all) lapply(out, location) else location(out)
}

res = bench::mark(
  stringr = stringr::str_locate(doc[i], b),
  new = str_locate(doc[i], b, FALSE),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         3.59       1       7.17
## 2 new         3.55   3.74      1         75.8     1

ggplot2::autoplot(res)

Commit	PR	File
cc3b92a	#2205	parser.R#L358

Before:

loc = stringr::str_locate_all(input, inline.code)[[1]]

After:

location = function(x) {
  len = attr(x, 'match.length')
  if (length(x) == 1 && x == -1) x = integer()
  cbind(start = x, end = x + len - 1L)
}
str_locate = function(x, pattern, all = TRUE) {
  out = (if (all) gregexpr else regexpr)(pattern, x, perl = TRUE)
  if (all) lapply(out, location) else location(out)
}
loc = str_locate(input, inline.code)[[1]]

input = 'Inline expressions such as `r "the following"`'
inline.code = knitr:::all_patterns$md$inline.code

res = bench::mark(
  stringr = stringr::str_locate_all(input, inline.code),
  new = str_locate(input, inline.code),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         3.05      1        1   
## 2 new         3.22   3.30      1         1.63     2.30

ggplot2::autoplot(res)

Commit	PR	File
cc3b92a	#2205	utils.R#L94

Before:

idx = stringr::str_locate(txt, hb) # locate documentclass

After:

location = function(x) {
  len = attr(x, 'match.length')
  if (length(x) == 1 && x == -1) x = integer()
  cbind(start = x, end = x + len - 1L)
}
str_locate = function(x, pattern, all = TRUE) {
  out = (if (all) gregexpr else regexpr)(pattern, x, perl = TRUE)
  if (all) lapply(out, location) else location(out)
}
idx = str_locate(txt, hb, FALSE) # locate documentclass

input = xfun::read_utf8(knitr_example("child/knitr-main.Rnw"))
patterns = knitr:::all_patterns$tex
db = patterns$document.begin
hb = patterns$header.begin
idx2 = grep(db, input)[1]
idx1 = grep(hb, input)[1]
txt = knitr:::one_string(input[idx1:(idx2 - 1L)])

location = function(x) {
  len = attr(x, 'match.length')
  if (length(x) == 1 && x == -1) x = integer()
  cbind(start = x, end = x + len - 1L)
}
str_locate = function(x, pattern, all = TRUE) {
  out = (if (all) gregexpr else regexpr)(pattern, x, perl = TRUE)
  if (all) lapply(out, location) else location(out)
}
res = bench::mark(
  stringr = stringr::str_locate(txt, hb),
  new = str_locate(txt, hb, FALSE),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         3.22       Inf     6.44
## 2 new         3.17   3.36      1          NaN     1

ggplot2::autoplot(res)

`stringr::str_locate_all()`

Commit	PR	File
cc3b92a	#2205	template.R#L121

Before:

loc = stringr::str_locate_all(txt, delim)[[1L]]

After:

location = function(x) {
  len = attr(x, 'match.length')
  if (length(x) == 1 && x == -1) x = integer()
  cbind(start = x, end = x + len - 1L)
}
str_locate = function(x, pattern, all = TRUE) {
  out = (if (all) gregexpr else regexpr)(pattern, x, perl = TRUE)
  if (all) lapply(out, location) else location(out)
}
loc = str_locate(txt, delim)[[1L]]

txt = 'This is the value of `x`: {{x}}'
delim = c('{{', '}}')
delim = gsub('([.|()\\^{}+$*?]|\\[|\\])', '\\\\\\1', delim)
delim = paste0(delim[1L], '((.|\n)+?)', delim[2L])
location = function(x) {
  len = attr(x, 'match.length')
  if (length(x) == 1 && x == -1) x = integer()
  cbind(start = x, end = x + len - 1L)
}
str_locate = function(x, pattern, all = TRUE) {
  out = (if (all) gregexpr else regexpr)(pattern, x, perl = TRUE)
  if (all) lapply(out, location) else location(out)
}

res = bench::mark(
  stringr = stringr::str_locate_all(txt, delim),
  new = str_locate(txt, delim),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         3.81       1       1   
## 2 new         3.85   3.96      1         93.8     1.05

ggplot2::autoplot(res)

`stringr::str_match_all()`

Commit	PR	File
cc3b92a	#2205	parser.R#L360

Before:

code = stringr::str_match_all(input, inline.code)[[1L]]

After:

str_match = function(x, pattern) {
  # gregexec() was added in R 4.1.0; for lower versions of R, use fallback
  if (is.function(gregexec <- baseenv()[['gregexec']])) {
    m = gregexec(pattern, x, perl = TRUE)
  } else {
    x = unlist(str_extract(x, pattern))
    m = regexec(pattern, x, perl = TRUE)
  }
  do.call(cbind, regmatches(x, m))
}
code = t(str_match(input, inline.code))

input = 'Inline expressions such as `r "the following"`'
inline.code = knitr:::all_patterns$md$inline.code
str_match = function(x, pattern) {
  if (is.function(gregexec <- baseenv()[['gregexec']])) {
    m = gregexec(pattern, x, perl = TRUE)
  } else {
    x = unlist(str_extract(x, pattern))
    m = regexec(pattern, x, perl = TRUE)
  }
  do.call(cbind, regmatches(x, m))
}

res = bench::mark(
  stringr = { 
    code = stringr::str_match_all(input, inline.code)[[1L]]
    code[is.na(code)] = ''
    code
  },
  new = t(str_match(input, inline.code)),
  min_time = Inf
)
summary(res, relative = TRUE)

## # A tibble: 2 × 6
##   expression   min median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
## 1 stringr     1      1         4.45      1        1.11
## 2 new         4.26   4.75      1         6.75     1

ggplot2::autoplot(res)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

stringr-replace.md

stringr-replace.md

Benchmark of stringr replacement

Context

Session info

Helpers

Benchmarking

`stringr::str_replace_all()`

`stringr::str_trim()`

`stringr::str_dup()`

`stringr::str_split()`

`stringr::str_sub()`

Variant replacement `str_insert()` instead or `str_replace()`

`stringr::str_detect()`

`stringr::str_wrap()`

`stringr::str_pad()`

`stringr::str_count()`

`stringr::str_extract_all`

`stringr::str_locate()`

`stringr::str_locate_all()`

`stringr::str_match_all()`

Files

stringr-replace.md

Latest commit

History

stringr-replace.md

File metadata and controls

Benchmark of stringr replacement

Context

Session info

Helpers

Benchmarking

stringr::str_replace_all()

stringr::str_trim()

stringr::str_dup()

stringr::str_split()

stringr::str_sub()

Variant replacement str_insert() instead or str_replace()

stringr::str_detect()

stringr::str_wrap()

stringr::str_pad()

stringr::str_count()

stringr::str_extract_all

stringr::str_locate()

stringr::str_locate_all()

stringr::str_match_all()

`stringr::str_replace_all()`

`stringr::str_trim()`

`stringr::str_dup()`

`stringr::str_split()`

`stringr::str_sub()`

Variant replacement `str_insert()` instead or `str_replace()`

`stringr::str_detect()`

`stringr::str_wrap()`

`stringr::str_pad()`

`stringr::str_count()`

`stringr::str_extract_all`

`stringr::str_locate()`

`stringr::str_locate_all()`

`stringr::str_match_all()`