stringr 1.5.0

Breaking changes

• stringr functions now consistently implement the tidyverse recycling rules
  (#372). There are two main changes:

  • Only vectors of length 1 are recycled. Previously, (e.g.)
    str_detect(letters, c("x", "y")) worked, but it now errors.

  • str_c() ignores NULLs, rather than treating them as length 0
    vectors.

  Additionally, many more arguments now throw errors, rather than warnings,
  if supplied the wrong type of input.

• regex() and friends now generate class names with stringr_ prefix (#384).

• str_detect(), str_starts(), str_ends() and str_subset() now error
  when used with either an empty string ("") or a boundary(). These
  operations didn't really make sense (str_detect(x, "") returned TRUE
  for all non-empty strings) and made it easy to make mistakes when programming.

New features

• Many tweaks to the documentation to make it more useful and consistent.

• New vignette("from-base") by @sastoudt provides a comprehensive comparison
  between base R functions and their stringr equivalents. It's designed to
  help you move to stringr if you're already familiar with base R string
  functions (#266).

• New str_escape() escapes regular expression metacharacters, providing
  an alternative to fixed() if you want to compose a pattern from user
  supplied strings (#408).

• New str_equal() compares two character vectors using unicode rules,
  optionally ignoring case (#381).

• str_extract() can now optionally extract a capturing group instead of
  the complete match (#420).

• New str_flatten_comma() is a special case of str_flatten() designed for
  comma separated flattening and can correctly apply the Oxford commas
  when there are only two elements (#444).

• New str_split_1() is tailored for the special case of splitting up a single
  string (#409).

• New str_split_i() extract a single piece from a string (#278, @bfgray3).

• New str_like() allows the use of SQL wildcards (#280, @rjpat).

• New str_rank() to complete the set of order/rank/sort functions (#353).

• New str_sub_all() to extract multiple substrings from each string.

• New str_unique() is a wrapper around stri_unique() and returns unique
  string values in a character vector (#249, @seasmith).

• str_view() uses ANSI colouring rather than an HTML widget (#370). This
  works in more places and requires fewer dependencies. It includes a number
  of other small improvements:

  • It no longer requires a pattern so you can use it to display strings with
    special characters.
  • It highlights unusual whitespace characters.
  • It's vectorised over both stringandpattern` (#407).
  • It defaults to displaying all matches, making str_view_all() redundant
    (and hence deprecated) (#455).

• New str_width() returns the display width of a string (#380).

• stringr is now licensed as MIT (#351).

Minor improvements and bug fixes

• Better error message if you supply a non-string pattern (#378).

• A new data source for sentences has fixed many small errors.

• str_extract() and str_exctract_all() now work correctly when pattern
  is a boundary().

• str_flatten() gains a last argument that optionally override the
  final separator (#377). It gains a na.rm argument to remove missing
  values (since it's a summary function) (#439).

• str_pad() gains use_width argument to control whether to use the total
  code point width or the number of code points as "width" of a string (#190).

• str_replace() and str_replace_all() can use standard tidyverse formula
  shorthand for replacement function (#331).

• str_starts() and str_ends() now correctly respect regex operator
  precedence (@carlganz).

• str_wrap() breaks only at whitespace by default; set
  whitespace_only = FALSE to return to the previous behaviour (#335, @rjpat).

• word() now returns all the sentence when using a negative start parameter
  that is greater or equal than the number of words. (@pdelboca, #245)