skimr/pkgs/rlang/NEWS.md

CLI EDA for CSVs

# rlang 1.1.2

* Fixed an off-by-one typo in the traceback source column location (#1633).

* `abort()` now respects the base R global option,
  `options(show.error.messages = FALSE)` (#1630).

* `obj_type_friendly()` now only displays the first class of S3 objects (#1622).

* `expr_label()` now has back-compatility with respect to changes made by R version 4.4 and `is.atomic(NULL)` (#1655)

* Performance improvement in `.rlang_cli_compat()` (#1657).


# rlang 1.1.1

* `englue()` now allows omitting `{{`. This is to make it easier to
  embed in external functions that need to support either `{` and `{{`
  (#1601).

* Fix for CRAN checks.

* `stop_input_type()` now handles `I()` input literally in `arg`
  (#1607, @simonpcouch).

* `parse_expr()` and `parse_exprs()` are now faster when
  `getOption("keep.source")` is `TRUE` (#1603).


# rlang 1.1.0

## Life cycle changes

* `dots_splice()` is deprecated. This function was previously in
  the questioning lifecycle stage as we were moving towards the
  explicit `!!!` splicing style.

* `flatten()`, `squash()`, and their variants are deprecated in favour
  of `purrr::list_flatten()` and `purrr::list_c()`.

* `child_env()` is deprecated in favour of `env()` which has supported
  creating child environments for several years now.


## Main new features

* `last_error()` and `options(rlang_backtrace_on_error = "full")` now
  print the full backtrace tree by default (except for some hidden
  frames). The simplified backtraces tended to hide important context
  too often. Now we show intervening frames in a lighter colour so
  that they don't distract from the important parts of the backtraces
  but are still easily inspectable.

* `global_entrace()`, `last_warnings()`, and `last_messages()` now
  support knitr documents.

* New `rlang_backtrace_on_warning_report` global option. This is
  useful in conjunction with `global_entrace()` to get backtraces on
  warnings inside RMarkdown documents.

* `global_entrace()` and `entrace()` now stop entracing warnings and
  messages after 20 times. This is to avoid a large overhead when 100s
  or 1000s of warnings are signalled in a loop (#1473).

* `abort()`, `warn()`, and `inform()` gain an `.inherit` parameter.
  This controls whether `parent` is inherited. If `FALSE`,
  `cnd_inherits()` and `try_fetch()` do not match chained conditions
  across parents.

  It's normally `TRUE` by default, but if a warning is chained to an
  error or a message is chained to a warning or error (downgraded
  chaining), `.inherit` defaults to `FALSE` (#1573).

* `try_fetch()` now looks up condition classes across chained errors
  (#1534). This makes `try_fetch()` insensitive to changes of
  implementation or context of evaluation that cause a classed error
  to suddenly get chained to a contextual error.

* `englue()` gained `env`, `error_arg`, and `error_call` arguments to
  support being wrapped in another function (#1565).

* The data-masking documentation for arguments has been imported from
  dplyr. You can link to it by starting an argument documentation with
  this button:

  ```
  <[`data-masking`][rlang::args_data_masking]>
  ```

* `enquos()` and friends gain a `.ignore_null` argument (#1450).

* New `env_is_user_facing()` function to determine if an evaluation
  frame corresponds to a direct usage by the end user (from the global
  environment or a package being tested) or indirect usage by a third
  party function. The return value can be overridden by setting the
  `"rlang_user_facing"` global option.


## Miscellaneous fixes and features

* New `check_data_frame()` and `check_logical()` functions in
  `standalone-types-check.R` (#1587, @mgirlich).

* Added `allow_infinite` argument to `check_number_whole()` (#1588, @mgirlich).

* The lifecycle standalone file has been updated to match the modern
  lifecycle tools.

* `parse_expr()` now supports vectors of lines (#1540).

* Quosures can now be consistently concatenated to lists of quosures (#1446).

* Fixed a memory issue that caused excessive duplication in `list2()`
  and friends (#1491).

* Embraced empty arguments are now properly detected and trimmed by
  `quos()` (#1421).

* Fixed an edge case that caused `enquos(.named = NULL)` to return a
  named list (#1505).

* `expr_deparse()` now deparses the embrace operator `{{` on a single
  line (#1511).

* `zap_srcref()` has been rewritten in C for efficiency (#1513).

* `zap_srcref()` now supports expression vectors.

* The non-error path of `check_dots_unnamed()` has been rewritten in C
  for efficiency (#1528).

* Improved error messages in `englue()` (#1531) and in glue strings in
  the LHS of `:=` (#1526).

* `englue()` now requires size 1 outputs (#1492). This prevents
  surprising errors or inconsistencies when an interpolated input of
  size != 1 makes its way into the glue string.

* `arg_match()` now throws correct error when supplied a missing value
  or an empty vector (#1519).

* `is_integerish()` now handles negative doubles more consistently
  with positive ones (@sorhawell, #1530).

* New `check_logical()` in `standalone-types-check.R` (#1560).

* `quo_squash()` now squashes quosures in function position (#1509).

* `is_expression()` now recognises quoted functions (#1499).
  It now also recognises non-parsable attributes (#1475).

* `obj_address()` now supports the missing argument (#1521).

* Fixed a `check_installed()` issue with packages removed during the
  current R session (#1561).

* `new_data_mask()` is now slightly faster due to a smaller initial mask size
  and usage of the C level function `R_NewEnv()` on R >=4.1.0 (#1553).

* The C level `r_dyn_*_push_back()` utilities are now faster (#1542).

* The C level `r_lgl_sum()` and `r_lgl_which()` helpers are now faster
  (#1577, with contributions from @mgirlich).

* rlang is now compliant with `-Wstrict-prototypes` as requested by CRAN
  (#1508).


# rlang 1.0.6

* `as_closure(seq.int)` now works (#1468).

* rlang no longer stores errors and backtraces in a `org:r-lib`
  environment on the search path.

* The low-level function `error_call()` is now exported (#1474).

* Fixed an issue that caused a failure about a missing `is_character`
  function when rlang is installed alongside an old version of vctrs (#1482).

* Fixed an issue that caused multiline calls in backtraces.

* The C API function `r_lgl_which()` now propagates the names of the input
  (#1471).

* The `pkg_version_info()` function now allows `==` for package
  version comparison (#1469, @kryekuzhinieri).


# rlang 1.0.5

* Fixed backtrace display with calls containing long lists of
  arguments (#1456).

* New `r_obj_type_friendly()` function in the C library (#1463). It
  interfaces with `obj_type_friendly()` from `compat-obj-type.R` via a
  C callable.


# rlang 1.0.4

* `is_installed()` no longer throws an error with irregular package
  names.

* `is_installed()` and `check_installed()` now properly detect that
  the base package is installed on older versions of R (#1434).


# rlang 1.0.3

* Child errors may now have empty messages to enable this pattern:

  ```
  Error in `my_function()`:
  Caused by error in `their_function()`:
  ! Message.
  ```

* The `rlib_bytes` class now uses prettyunits to format bytes. The
  bytes are now represented with decimal prefixes instead of binary
  prefixes.

* Supplying a frame environment to the `call` argument of `abort()`
  now causes the corresponding function call in the backtrace to be
  highlighted.

  In addition, if you store the argument name of a failing input in
  the `arg` error field, the argument is also highlighted in the
  backtrace.

  Instead of:

  ```
  cli::cli_abort("{.arg {arg}} must be a foobar.", call = call)
  ```

  You can now write this to benefit from arg highlighting:

  ```
  cli::cli_abort("{.arg {arg}} must be a foobar.", arg = arg, call = call)
  ```

* `abort(message = )` can now be a function. In this case, it is
  stored in the `header` field and acts as a `cnd_header()` method
  invoked when the message is displayed.

* New `obj_type_oo()` function in `compat-obj-type.R` (#1426).

* `friendly_type_of()` from `compat-obj-type.R` (formerly
  `compat-friendly-type.R`) is now `obj_type_friendly()`.

* `options(backtrace_on_error = "collapse")` and `print(trace,
  simplify = "collapse")` are deprecated. They fall back to `"none"`
  with a warning.

* `call_match()` now better handles `...` when `dots_expand = FALSE`.

* `list2(!!!x)` is now faster when `x` is a list. It is now returned
  as is instead of being duplicated into a new list.

* `abort()` gains a `.trace_bottom` argument to disambiguate from
  other `.frame`. This allows `cli::cli_abort()` to wrap `abort()` in
  such a way that `.internal` mentions the correct package to report
  the error in (#1386).

* The `transpose()` compat is now more consistent with purrr when
  inner names are not congruent (#1346).

* New `reset_warning_verbosity()` and `reset_message_verbosity()`
  functions. These reset the verbosity of messages signalled with
  `warn()` and `inform()` with the `.frequency` argument. This is
  useful for testing verbosity in your package (#1414).

* `check_dots_empty()` now allows trailing missing arguments (#1390).

* Calls to local functions that are not accessible through `::` or
  `:::` are now marked with `(local)` in backtraces (#1399).

* Error messages now mention indexed calls like `foo$bar()`.

* New `env_coalesce()` function to copy bindings from one environment
  to another. Unlike approaches based on looping with `[[<-`,
  `env_coalesce()` preserves active and lazy bindings.

* Chaining errors at top-level (directly in the console instead of in
  a function) no longer fails (#1405).

* Warning style is propagated across parent errors in chained error
  messages (#1387).

* `check_installed()` now works within catch-all `tryCatch(error = )`
  expressions (#1402, tidyverse/ggplot2#4845).

* `arg_match()` and `arg_match0()` now mention the correct call in
  case of type error (#1388).

* `abort()` and `inform()` now print messages to `stdout` in RStudio
  panes (#1393).

* `is_installed()` now detects unsealed namespaces (#1378). This fixes
  inconsistent behaviour when run within user onLoad hooks.

* Source references in backtraces and `last_error()`/`last_trace()` instructions
  are now clickable in IDEs that support links (#1396).

* `compat-cli.R` now supports `style_hyperlink()`.

* `abort(.homonyms = "error")` now throws the expected error (#1394).

* `env_binding_are_active()` no longer accidentally triggers active bindings
  (#1376).

* Fixed bug in `quo_squash()` with nested quosures containing the
  missing argument.


# rlang 1.0.2

* Backtraces of parent errors are now reused on rethrow. This avoids
  capturing the same backtrace twice and solves consistency problems
  by making sure both errors in a chain have the same backtrace.

* Fixed backtrace oversimplification when `cnd` is a base error in
  `abort(parent = cnd)`.

* Internal errors thrown with `abort(.internal = TRUE)` now mention
  the name of the package the error should be reported to.

* Backtraces are now separated from error messages with a `---` ruler
  line (#1368).

* The internal bullet formatting routine now ignores unknown names
  (#1364). This makes it consistent with the cli package, increases
  resilience against hard-to-detect errors, and increases forward
  compatibility.

* `abort()` and friends no longer calls non-existent functions
  (e.g. `cli::format_error()` or `cli::format_warning`) when the
  installed version of cli is too old (#1367, tidyverse/dplyr#6189).

* Fixed an OOB subsetting error in `abort()`.


# rlang 1.0.1

* New `rlang_call_format_srcrefs` global option (#1349). Similar to
  `rlang_trace_format_srcrefs`, this option allows turning off the
  display of srcrefs in error calls. This can be useful for
  reproducibility but note that srcrefs are already disabled
  within testthat by default.

* `abort(parent = NA)` is now supported to indicate an unchained
  rethrow. This helps `abort()` detect the condition handling context
  to create simpler backtraces where this context is hidden by
  default.

* When `parent` is supplied, `abort()` now loops over callers to
  detect the condition handler frame. This makes it easier to wrap or
  extract condition handlers in functions without supplying `.frame`.

* When `parent` is supplied and `call` points to the condition setup
  frame (e.g. `withCallingHandlers()` or `try_fetch()`), `call` is
  replaced with the caller of that setup frame. This provides a more
  helpful default call.

* `is_call()` is now implemented in C for performance.

* Fixed performance regression in `trace_back()`.

* Fixed a partial matching issue with `header`, `body`, and `footer`
  condition fields.

* `eval_tidy()` calls are no longer mentioned in error messages.


# rlang 1.0.0

## Major changes

This release focuses on the rlang errors framework and features
extensive changes to the display of error messages.

* `abort()` now displays errors as fully bulleted lists. Error headers
  are displayed with a `!` prefix. See
  <https://rlang.r-lib.org/reference/topic-condition-customisation.html>
  to customise the display of error messages.

* `abort()` now displays a full chain of messages when errors are
  chained with the `parent` argument. Following this change, you
  should update dplyr to version 1.0.8 to get proper error messages.

* `abort()` now displays function calls in which a message originated
  by default. We have refrained from showing these calls until now to
  avoid confusing messages when an error is thrown from a helper
  function that isn't relevant to users.

  To help with these cases, `abort()` now takes a `call` argument that
  you can set to `caller_env()` or `parent.frame()` when used in a
  helper function. The function call corresponding to this environment
  is retrieved and stored in the condition.

* cli formatting is now supported. Use `cli::cli_abort()` to get
  advanced formatting of error messages, including indented bulleted
  lists. See <https://rlang.r-lib.org/reference/topic-condition-formatting.html>.

* New `try_fetch()` function for error handling. We recommend to use
  it for chaining errors. It mostly works like `tryCatch()` with a few
  important differences.

  - Compared to `tryCatch()`, `try_fetch()` preserves the call
    stack. This allows full backtrace capture and allows `recover()`
    to reach the error site.

  - Compared to `withCallingHandler()`, `try_fetch()` is able to
    handle stack overflow errors (this requires R 4.2, unreleased at
    the time of writing).

* The tidy eval documentation has been fully rewritten to reflect
  current practices. Access it through the "Tidy evaluation" and
  "Metaprogramming" menus on <https://rlang.r-lib.org>.


## Breaking changes

* The `.data` object exported by rlang now fails when subsetted
  instead of returning `NULL`. This new error helps you detect when
  `.data` is used in the wrong context.

  We've noticed several packages failing after this change because
  they were using `.data` outside of a data-masking context. For
  instance the `by` argument of `dplyr::join()` is not data-masked.
  Previously `dplyr::join(by = .data$foo)` would silently be
  interpreted as `dplyr::join(by = NULL)`. This is now an error.

  Another issue is using `.data` inside `ggplot2::labs(...)`. This is
  not allowed since `labs()` isn't data-masked.

* `call_name()` now returns `NULL` instead of `"::"` for calls of the
  form `foo::bar`.

  We've noticed some packages do not check for `NULL` results from
  `call_name()`. Note that many complex calls such as `foo()()`,
  `foo$bar()` don't have a "name" and cause a `NULL` result. This is
  why you should always check for `NULL` results when using
  `call_name()`.

  We've added the function `is_call_simple()` to make it easier to
  work safely with `call_name()`. The invariant is that `call_name()`
  always returns a string when `is_call_simple()` returns `TRUE`.
  Conversely it always returns `NULL` when `is_call_simple()` retuns
  `FALSE`.

* `is_expression()` now returns `FALSE` for manually constructed
  expressions that can't be created by the parser. It used to return
  `TRUE` for any calls, including those that contain injected objects.

  Consider using `is_call()` or just remove the expression check. In
  many cases it is fine letting all objects go through when an
  expression is expected. For instance you can inject objects directly
  inside dplyr arguments:

  ```
  x <- seq_len(nrow(data))
  dplyr::mutate(data, col = !!x)
  ```

* If a string is supplied to `as_function()` instead of an object
  (function or formula), the function is looked up in the global
  environment instead of the calling environment. In general, passing
  a function name as a string is brittle. It is easy to forget to pass
  the user environment to `as_function()` and sometimes there is no
  obvious user environment. The support for strings should be
  considered a convenience for end users only, not for programmers.

  Since environment forwarding is easy to mess up, and since the
  feature is aimed towards end users, `as_function()` now defaults to
  the global environment. Supply an environment explicitly if that is
  not correct in your case.

* `with_handlers()`, `call_fn()`, and `friendly_type()` are deprecated.

* The `action` argument of `check_dots_used()`, `check_dots_unnamed()`,
  and `check_dots_empty()` is deprecated in favour of the new `error`
  argument which takes an error handler.

* Many functions deprecated in rlang 0.2.0 and 0.3.0 have
  been removed from the package.


## Fixes and features

### tidyeval

* New `englue()` operator to allow string-embracing outside of dynamic
  dots (#1172).

* New `data_sym()` and `data_syms()` functions to create calls of the
  form `.data$foo`.

* `.data` now fails early when it is subsetted outside of a data mask
  context. This provides a more informative error message (#804, #1133).

* `as_label()` now better handles calls to infix operators (#956,
  r-lib/testthat#1432). This change improves auto-labelled expressions
  in data-masking functions like `tibble()`, `mutate()`, etc.

* The `{{` operator is now detected more strictly (#1087). If
  additional arguments are supplied through `{`, it is no longer
  interpreted as an injection operator.

* The `.ignore_empty` argument of `enexprs()` and `enquos()` no longer
  treats named arguments supplied through `...` as empty, consistently
  with `exprs()` and `quos()` (#1229).

* Fixed a hang when a quosure inheriting from a data mask is evaluated
  in the mask again.

* Fixed performance issue when splicing classes that explicitly
  inherit from list with `!!!` (#1140, r-lib/vctrs#1170).

* Attributes of quosure lists are no longer modified by side effect
  (#1142).

* `enquo()`, `enquos()` and variants now support numbered dots like
  `..1` (#1137).

* Fixed a bug in the AST rotation algorithm that caused the `!!`
  operator to unexpectedly mutate injected objects (#1103).

* Fixed AST rotation issue with `!!` involving binary operators (#1125).


### rlang errors

* `try_fetch()` is a flexible alternative to both `tryCatch()` and
  `withCallingHandlers()` (#503). It is also more efficient than
  `tryCatch()` and creates leaner backtraces.

* New `cnd_inherits()` function to detect a class in a chain of errors
  (#1293).

* New `global_entrace()` function, a user-friendly helper for
  configuring errors in your RProfile. Call it to enrich all base
  errors and warnings with an rlang backtrace. This enables
  `last_error()`, `last_warnings()`, `last_messages()`, and
  `backtrace_on_error` support for all conditions.

* New `global_handle()` function to install a default configuration of
  error handlers. This currently calls `global_entrace()` and
  `global_prompt_install()`. Expect more to come.

* The "Error:" part of error messages is now printed by rlang instead
  of R. This introduces several cosmetic and informative changes in
  errors thrown by `abort()`:

  - The `call` field of error messages is now displayed, as is the
    default in `base::stop()`. The call is only displayed if it is a
    simple expression (e.g. no inlined function) and the arguments are
    not displayed to avoid distracting from the error message. The
    message is formatted with the tidyverse style (`code` formatting
    by the cli package if available).

  - The source location is displayed (as in `base::stop()`) if `call`
    carries a source reference. Source locations are not displayed
    when testthat is running to avoid brittle snapshots.

  - Error headers are always displayed on their own line, with a `"!"`
    bullet prefix.

  See <https://rlang.r-lib.org/reference/topic-condition-customisation.html>
  to customise this new display.

* The display of chained errors created with the `parent` argument of
  `abort()` has been improved. Chains of errors are now displayed at
  throw time with the error prefix "Caused by error:".

* The `print()` method of rlang errors (commonly invoked with
  `last_error()`) has been improved:
    - Display calls if present.
    - Chained errors are displayed more clearly.

* `inform()` and `warn()` messages can now be silenced with the global
  options `rlib_message_verbosity` and `rlib_warning_verbosity`.

* `abort()` now outputs error messages to `stdout` in interactive
  sessions, following the same approach as `inform()`.

* Errors, warnings, and messages generated from rlang are now
  formatted with cli. This means in practice that long lines are
  width-wrapped to the terminal size and user themes are applied.
  This is currently only the case for rlang messages.

  This special formatting is not applied when `abort()`, `warn()`, and
  `inform()` are called from another namespace than rlang.
  See <https://rlang.r-lib.org/reference/topic-condition-formatting.html>
  if you'd like to use cli to format condition messages in your
  package.

* `format_error_bullets()` (used as a fallback instead of cli) now
  treats:

  - Unnamed elements as unindented line breaks (#1130)
  - Elements named `"v"` as green ticks (@rossellhayes)
  - Elements named `" "` as indented line breaks
  - Elements named `"*"` as normal bullets
  - Elements named `"!"` as warning bullets

  For convenience, a fully unnamed vector is interpreted as a vector
  of `"*"` bullets.

* `abort()` gains a `.internal` argument. When set to `TRUE`, a footer
  bullet is added to `message` to let the user know that the error is
  internal and that they should report it to the package authors.

* `abort()`, `warn()`, and `inform()` gain a `body` argument to supply
  additional bullets in the error message.

* rlang conditions now have `as.character()` methods. Use this generic
  on conditions to generate a whole error message, including the
  `Error:` prefix. These methods are implemented as wrappers around
  `cnd_message()`.

* `header` and `footer` methods can now be stored as closures in
  condition fields of the same name.

* `cnd_message()` gains a `prefix` argument to print the message with
  a full prefix, including `call` field if present and parent messages
  if the condition is chained.

* `cnd_message()` gains an `inherit` argument to control whether to
  print the messages of parent errors.

* Condition constructors now check for duplicate field names (#1268).

* `cnd_footer()` now returns the `footer` field by default, if any.

* `warn()` and `inform()` now signal conditions of classes
  `"rlang_warning"` and `"rlang_message"` respectively.

* The `body` field of error conditions can now be a character vector.

* The error returned by `last_error()` is now stored on the search
  path as the `.Last.error` binding of the `"org:r-lib"`
  environment. This is consistent with how the processx package
  records error conditions. Printing the `.Last.error` object is now
  equivalent to running `last_error()`.

* Added `is_error()`, `is_warning()`, and `is_message()` predicates (#1220).

* `interrupt()` no longer fails when interrupts are suspended (#1224).

* `warn()` now temporarily sets the `warning.length` global option to
  the maximum value (8170). The default limit (1000 characters) is
  especially easy to hit when the message contains a lot of ANSI
  escapes, as created by the crayon or cli packages (#1211).


### Backtraces

* `entrace()` and `global_entrace()` now log warnings and messages
  with backtraces attached. Run `last_warnings()` or `last_messages()`
  to inspect the warnings or messages emitted during the last command.

* Internal errors now include a winch backtrace if installed. The user
  is invited to install it if not installed.

* Display of rlang backtraces for expected errors in dynamic reports
  (chunks where `error = TRUE` in knitted documents and RStudio
  notebooks) is now controlled by the
  `rlang_backtrace_on_error_report` option. By default, this is set to
  `"none"`.

  The display of backtraces for _unexpected_ errors (in chunks where
  `error` is unset or set to `FALSE`) is still controlled by
  `rlang_backtrace_on_error`.

* The `last_error()` reminder is no longer displayed in RStudio
  notebooks.

* A `knitr::sew()` method is registered for `rlang_error`. This makes
  it possible to consult `last_error()` (the call must occur in a
  different chunk than the error) and to set
  `rlang_backtrace_on_error_report` global options in knitr to display
  a backtrace for expected errors.

  If you show rlang backtraces in a knitted document, also set this in
  a hidden chunk to trim the knitr context from the backtraces:

  ```
  options(
    rlang_trace_top_env = environment()
  )
  ```

  This change replaces an ad hoc mechanism that caused bugs in corner
  cases (#1205).

* The `rlang_trace_top_env` global option for `trace_back()` now
  detects when backtraces are created within knitr. If the option is
  not set, its default value becomes `knitr::knit_global()` when knitr
  is in progress (as determined from `knitr.in.progress` global
  option). This prevents the knitr evaluation context from appearing
  in the backtraces (#932).

* Namespace changes are now emboldened in backtraces (#946).

* Functions defined in the global environments or in local execution
  environments are now displayed with a space separator in backtraces
  instead of `::` and `:::`. This avoids making it seem like these
  frame calls are valid R code ready to be typed in (#902).

* Backtraces no longer contain inlined objects to avoid performance
  issues in edge cases (#1069, r-lib/testthat#1223).

* External backtraces in error chains are now separately displayed (#1098).

* Trace capture now better handles wrappers of calling handler in case
  of rethrown chained errors.

* Backtraces now print dangling srcrefs (#1206). Paths are shortened
  to show only three components (two levels of folder and the file).

* The root symbol in backtraces is now slightly different so that it
  can't be confused with a prompt character (#1207).


### Argument intake

* `arg_match()` gains a `multiple` argument for cases where zero or
  several matches are allowed (#1281).

* New function `check_required()` to check that an argument is
  supplied. It produces a more friendly error message than `force()`
  (#1118).

* `check_dots_empty()`, `check_dots_used()`, and
  `check_dots_unnamed()` have been moved from ellipsis to rlang. The
  ellipsis package is deprecated and will eventually be archived.

  We have added `check_dots_empty0()`. It has a different UI but is
  almost as efficient as checking for `missing(...)`. Use this in very
  low level functions where a couple microseconds make a difference.

* The `arg_nm` argument of `arg_match0()` must now be a string or
  symbol.

* `arg_match()` now mentions the supplied argument (#1113).

* `is_installed()` and `check_installed()` gain a `version` argument (#1165).

* `check_installed()` now consults the
  `rlib_restart_package_not_found` global option to determine whether
  to prompt users to install packages. This also disables the restart
  mechanism (see below).

* `check_installed()` now signals errors of class
  `rlib_error_package_not_found` with a
  `rlib_restart_package_not_found` restart. This allows calling
  handlers to install the required packages and restart the check
  (#1150).

* `is_installed()` and `check_installed()` now support
  DESCRIPTION-style version requirements like `"rlang (>= 1.0)"`.
  They also gain `version` and `compare` arguments to supply requirements
  programmatically.

* `check_installed()` gains an `action` argument that is called when
  the user chooses to install and update missing and outdated packages.

* New `check_exclusive()` function to check that only one argument of
  a set is supplied (#1261).


### R APIs

* `on_load()` and `run_on_load()` lets you run `.onLoad()` expressions
  from any file of your package. `on_package_load()` runs expressions
  when another package is loaded. (#1284)

* The new predicate `is_call_simple()` indicates whether a call has a
  name and/or a namespace. It provides two invariants:

  - If `is_call_simple(x)` is `TRUE`, `call_name()` always returns a
    string.

  - If `is_call_simple(x, ns = TRUE)` is `TRUE`, `call_ns()` always
    returns a string.

* `call_name()` and `call_ns()` now return `NULL` with calls of the
  form `foo::bar` (#670).

* New `current_call()`, `caller_call()`, and `frame_call()`
  accessors. New `frame_fn()` accessor.

* `env_has()` and the corresponding C-level function no longer force
  active bindings (#1292).

* New `names2<-` replacement function that never adds missing values
  when names don't have names (#1301).

* `zap_srcref()` now preserves attributes of closures.

* Objects headers (as printed by `last_error()`, `env_print()`, ...)
  are now formatted using the `cls` class of the cli package.

* `as_function()` gains `arg` and `call` arguments to provide
  contextual information about erroring inputs.

* `is_expression()` now returns `FALSE` for manually constructed
  expressions that cannot be created by the R parser.

* New C callable `rlang_env_unbind()`. This is a wrapper around
  `R_removeVarFromFrame()` on R >= 4.0.0. On older R this wraps the R
  function `base::rm()`. Unlike `rm()`, this function does not warn
  (nor throw) when a binding does not exist.

* `friendly_type_of()` now supports missing arguments.

* `env_clone()` now properly clones active bindings and avoids forcing
  promises (#1228). On R < 4.0, promises are still forced.

* Fixed an `s3_register()` issue when the registering package is a
  dependency of the package that exports the generic (#1225).

* Added `compat-vctrs.R` file for robust manipulation of data frames
  in zero-deps packages.

* Added `compat-cli.R` file to format message elements consistently
  with cli in zero-deps packages.

* `compat-purrr.R` now longer includes `pluck*` helpers; these used a defintion
  of pluck that predated purrr (#1159). `*_cpl()` has also been removed.
  The `map*` wrappers now call `as_function()` so that you can pass short
  anonymous functions that use `~` (#1157).

* `exprs_auto_name()` gains a `repair_auto` argument to make automatic
  names unique (#1116).

* The `.named` argument of `dots_list()` can now be set to `NULL` to
  give the result default names. With this option, fully unnamed
  inputs produce a fully unnamed result with `NULL` names instead of a
  character vector of minimal `""` names (#390).

* `is_named2()` is a variant of `is_named()` that always returns
  `TRUE` for empty vectors (#191). It tests for the property that each
  element of a vector is named rather than the presence of a `names`
  attribute.

* New `rlib_bytes` class imported from the bench package (#1117).
  It prints and parses human-friendly sizes.

* The `env` argument of `as_function()` now defaults to the global
  environment. Its previous default was the caller of `as_function()`,
  which was rarely the correct environment to look in. Since it's hard
  to remember to pass the user environment and it's sometimes tricky
  to keep track of it, it's best to consider string lookup as a
  convenience for end users, not for developers (#1170).

* `s3_register()` no longer fails when generic does not exist. This
  prevents failures when users don't have all the last versions of
  packages (#1112).

* Formulas are now deparsed according to the tidyverse style guide
  (`~symbol` without space and `~ expression()` with a space).

* New `hash_file()`, complementing `hash()`, to generate 128-bit hashes for
  the data within a file without loading it into R (#1134).

* New `env_cache()` function to retrieve a value or create it with a
  default if it doesn't exist yet (#1081).

* `env_get()` and `env_get_list()` gain a `last` argument. Lookup
  stops in that environment. This can be useful in conjunction with
  `base::topenv()`.

* New `call_match()` function. It is like `match.call()` but also
  supports matching missing arguments to their defaults in the function
  definition (#875).

  `call_standardise()` is deprecated in favour of `call_match()`.

* `expr_deparse()` now properly escapes `\` characters in symbols,
  argument names, and vector names (#1160).

* `friendly_type_of()` (from `compat-friendly-type.R`) now supports
  matrices and arrays (#141).

* Updated `env_print()` to use `format_error_bullets()` and consistent
  tidyverse style (#1154).

* `set_names()` now recycles names of size 1 to the size of the input,
  following the tidyverse recycling rules.

* `is_bare_formula()` now handles the `scoped` argument
  consistently. The default has been changed to `TRUE` for
  compatibility with the historical default behaviour (#1115).

* The "definition" API (`dots_definitions()` etc.) has been archived.

* New `is_complex()` predicates to complete the family (#1127).

* The C function `r_obj_address()` now properly prefixes addresses
  with the hexadecimal prefix `0x` on Windows (#1135).

* `obj_address()` is now exported.

* `%<~%` now actually works.

* `XXH3_64bits()` from the XXHash library is now exposed as C callable
  under the name `rlang_xxh3_64bits()`.


# rlang 0.4.12

* Fix for CRAN checks.


# rlang 0.4.11

* Fix for CRAN checks.

* Fixed a gcc11 warning related to `hash()` (#1088).


# rlang 0.4.10

* New `hash()` function to generate 128-bit hashes for arbitrary R objects
  using the xxHash library. The implementation is modeled after
  [xxhashlite](https://github.com/coolbutuseless/xxhashlite), created
  by @coolbutuseless.

* New `check_installed()` function. Unlike `is_installed()`, it asks
  the user whether to install missing packages. If the user accepts,
  the packages are installed with `pak::pkg_install()` if available,
  or `utils::install.packages()` otherwise. If the session is non
  interactive or if the user chooses not to install the packages, the
  current evaluation is aborted (#1075).

* rlang is now licensed as MIT (#1063).

* Fixed an issue causing extra empty lines in `inform()` messages with
  `.frequency` (#1076, @schloerke).

* `expr_deparse()` now correctly wraps code using `::` and `:::`
  (#1072, @krlmlr).


# rlang 0.4.9

## Breaking changes

* Dropped support for the R 3.2 series.


## New features

* `inject()` evaluates its argument with `!!`, `!!!`, and `{{`
  support.

* New `enquo0()` and `enquos0()` operators for defusing function
  arguments without automatic injection (unquotation).

* `format_error_bullets()` is no longer experimental. The `message`
  arguments of `abort()`, `warn()`, and `inform()` are automatically
  passed to that function to make it easy to create messages with
  regular, info, and error bullets. See `?format_error_bullets` for
  more information.

* New `zap_srcref()` function to recursively remove source references
  from functions and calls.

* A new compat file for the zeallot operator `%<-%` is now available
  in the rlang repository.

* New `%<~%` operator to define a variable lazily.

* New `env_browse()` and `env_is_browsed()` functions. `env_browse()`
  is equivalent to evaluating `browser()` within an environment. It
  sets the environment to be persistently browsable (or unsets it if
  `value = FALSE` is supplied).

* Functions created from quosures with `as_function()` now print in a
  more user friendly way.

* New `rlang_print_backtrace` C callable for debugging from C
  interpreters (#1059).


## Bugfixes and improvements

* The `.data` pronoun no longer skips functions (#1061). This solves a
  dplyr issue involving rowwise data frames and list-columns of
  functions (tidyverse/dplyr#5608).

* `as_data_mask()` now intialises environments of the correct size to
  improve efficiency (#1048).

* `eval_bare()`, `eval_tidy()` (#961), and `with_handlers()` (#518)
  now propagate visibility.

* `cnd_signal()` now ignores `NULL` inputs.

* Fixed bug that prevented splicing a named empty vector with the
  `!!!` operator (#1045).

* The exit status of is now preserved in non-interactive sessions when
  `entrace()` is used as an `options(error = )` handler (#1052,
  rstudio/bookdown#920).

* `next` and `break` are now properly deparsed as nullary operators.


# rlang 0.4.8

* Backtraces now include native stacks (e.g. from C code) when the
  [winch](https://r-prof.github.io/winch/) package is installed and
  `rlang_trace_use_winch` is set to `TRUE` (@krlmlr).

* Compatibility with upcoming testthat 3 and magrittr 2 releases.

* `get_env()` now returns the proper environment with primitive
  functions, i.e. the base namespace rather than the base environment
  (r-lib/downlit#32).

* `entrace()` no longer handles non-rlang errors that carry a
  backtrace. This improves compatibility with packages like callr.

* Backtraces of unhandled errors are now displayed without truncation
  in non-interactive sessions (#856).

* `is_interactive()` no longer consults "rstudio.notebook.executing"
  option (#1031).


# rlang 0.4.7

* `cnd_muffle()` now returns `FALSE` instead of failing if the
  condition is not mufflable (#1022).

* `warn()` and `inform()` gain a `.frequency` argument to control how
  frequently the warning or message should be displayed.

* New `raw_deparse_str()` function for converting a raw vector into a
  string of hexadecimal characters (@krlmlr, #978).

* The backtraces of chained errors are no longer decomposed by error
  context. Instead, the error messages are displayed as a tree to
  reflect the error ancestry, and the deepest backtrace in the ancestry
  is displayed.

  This change simplifies the display (#851) and makes it possible to
  rethow errors from a calling handler rather than an exiting handler,
  which we now think is more appropriate because it allows users to
  `recover()` into the error.

* `env_bind()`, `env_bind_active()`, `env_bind_lazy()`, `env_get()`,
  and `env_get_list()` have been rewritten in C.

* `env_poke()` now supports `zap()` sentinels for removing bindings
  (#1012) and has better support for characters that are not
  representable in the local encoding.

* `env_poke()` has been rewritten in C for performance.

* The unicode translation warnings that appeared on Windows with R 4.0
  are now fixed.

* `env_unbind(inherit = TRUE)` now only removes a binding from the
  first parent environment that has a binding. It used to remove the
  bindings from the whole ancestry. The new behaviour doesn't
  guarantee that a scope doesn't have a binding but it is safer.

* `env_has()` is now rewritten in C for performance.

* `dots_list()` gains a `.named` argument for auto-naming dots (#957).

* It is now possible to subset the `.data` pronoun with quosured
  symbols or strings (#807).

* Expressions like `quote(list("a b" = 1))` are now properly deparsed
  by `expr_deparse()` (#950).

* `parse_exprs()` now preserves names (#808). When a single string
  produces multiple expressions, the names may be useful to figure out
  what input produced which expression.

* `parse_exprs()` now supports empty expressions (#954).

* `list2(!!!x)` no longer evaluates `x` multiple times (#981).

* `is_installed()` now properly handles a `pkg` argument of length > 1.
  Before this it silently tested the first element of `pkg` only
  and thus always returned `TRUE` if the first package was installed
  regardless of the actual length of `pkg`. (#991, @salim-b)

* `arg_match0()` is a faster version of `arg_match()` for use when performance
  is at a premium (#997, @krlmlr).


# rlang 0.4.6

* `!!!` now uses a combination of `length()`, `names()`, and `[[` to splice
  S3 and S4 objects. This produces more consistent behaviour than `as.list()`
  on a wider variety of vector classes (#945, tidyverse/dplyr#4931).


# rlang 0.4.5

* `set_names()`, `is_formula()`, and `names2()` are now implemented in
  C for efficiency.

* The `.data` pronoun now accepts symbol subscripts (#836).

* Quosure lists now explicitly inherit from `"list"`. This makes them
  compatible with the vctrs package (#928).

* All rlang options are now documented in a centralised place, see
  `?rlang::faq-options` (#899, @smingerson).

* Fixed crash when `env_bindings_are_lazy()` gets improper arguments (#923).

* `arg_match()` now detects and suggests possible typos in provided
  arguments (@jonkeane, #798).

* `arg_match()` now gives an error if argument is of length greater
  than 1 and doesn't exactly match the values input, similar to base
  `match.arg` (#914, @AliciaSchep)


# rlang 0.4.4

* Maintenance release for CRAN.


# rlang 0.4.3

* You can now use glue syntax to unquote on the LHS of `:=`. This
  syntax is automatically available in all functions taking dots with
  `list2()` and `enquos()`, and thus most of the tidyverse. Note that
  if you use the glue syntax in an R package, you need to import glue.

  A single pair of braces triggers normal glue interpolation:

  ```r
  df <- data.frame(x = 1:3)

  suffix <- "foo"
  df %>% dplyr::mutate("var_{suffix}" := x * 2)
  #>   x var_foo
  #> 1 1       2
  #> 2 2       4
  #> 3 3       6
  ```

  Using a pair of double braces is for labelling a function argument.
  Technically, this is shortcut for `"{as_label(enquo(arg))}"`. The
  syntax is similar to the curly-curly syntax for interpolating
  function arguments:

  ```r
  my_wrapper <- function(data, var, suffix = "foo") {
    data %>% dplyr::mutate("{{ var }}_{suffix}" := {{ var }} * 2)
  }
  df %>% my_wrapper(x)
  #>   x x_foo
  #> 1 1     2
  #> 2 2     4
  #> 3 3     6

  df %>% my_wrapper(sqrt(x))
  #>   x sqrt(x)_foo
  #> 1 1    2.000000
  #> 2 2    2.828427
  #> 3 3    3.464102
  ```

* Fixed a bug in magrittr backtraces that caused duplicate calls to
  appear in the trace.

* Fixed a bug in magrittr backtraces that caused wrong call indices.

* Empty backtraces are no longer shown when `rlang_backtrace_on_error`
  is set.

* The tidy eval `.env` pronoun is now exported for documentation
  purposes.

* `warn()` and `abort()` now check that either `class` or `message`
  was supplied. `inform()` allows sending empty message as it is
  occasionally useful for building user output incrementally.

* `flatten()` fails with a proper error when input can't be flattened (#868, #885).

* `inform()` now consistently appends a final newline to the message
  (#880).

* `cnd_body.default()` is now properly registered.

* `cnd_signal()` now uses the same approach as `abort()` to save
  unhandled errors to `last_error()`.

* Parsable constants like `NaN` and `NA_integer_` are now deparsed by
  `expr_deparse()` in their parsable form (#890).

* Infix operators now stick to their LHS when deparsed by
  `expr_deparse()` (#890).
  

# rlang 0.4.2

* New `cnd_header()`, `cnd_body()` and `cnd_footer()` generics. These
  are automatically called by `conditionMessage.rlang_error()`, the
  default method for all rlang errors.

  Concretely, this is a way of breaking up lazy generation of error
  messages with `conditionMessage()` into three independent
  parts. This provides a lot of flexibility for hierarchies of error
  classes, for instance you could inherit the body of an error message
  from a parent class while overriding the header and footer.

* The reminder to call `last_error()` is now less confusing thanks to
  a suggestion by @markhwhiteii.

* The functions prefixed in `scoped_` have been renamed to use the
  more conventional `local_` prefix. For instance, `scoped_bindings()`
  is now `local_bindings()`. The `scoped_` functions will be
  deprecated in the next significant version of rlang (0.5.0).

* The `.subclass` argument of `abort()`, `warn()` and `inform()` has
  been renamed to `class`. This is for consistency with our
  conventions for class constructors documented in
  https://adv-r.hadley.nz/s3.html#s3-subclassing.

* `inform()` now prints messages to the standard output by default in
  interactive sessions. This makes them appear more like normal output
  in IDEs such as RStudio. In non-interactive sessions, messages are
  still printed to standard error to make it easy to redirect messages
  when running R scripts (#852).

* Fixed an error in `trace_back()` when the call stack contains a
  quosured symbol.

* Backtrace is now displayed in full when an error occurs in
  non-interactive sessions. Previously the backtraces of parent errors
  were left out.


# rlang 0.4.1

* New experimental framework for creating bulleted error messages. See
  `?cnd_message` for the motivation and an overwiew of the tools we
  have created to support this approach. In particular, `abort()` now
  takes character vectors to assemble a bullet list. Elements named
  `x` are prefixed with a red cross, elements named `i` are prefixed
  with a blue info symbol, and unnamed elements are prefixed with a
  bullet.

* Capture of backtrace in the context of rethrowing an error from an
  exiting handler has been improved. The `tryCatch()` context no
  longer leaks in the high-level backtrace.

* Printing an error no longer recommends calling `last_trace()`,
  unless called from `last_error()`.

* `env_clone()` no longer recreates active bindings and is now just an
  alias for `env2list(as.list(env))`. Unlike `as.list()` which returns
  the active binding function on R < 4.0, the value of active bindings
  is consistently used in all versions.

* The display of rlang errors derived from parent errors has been
  improved. The simplified backtrace (as printed by
  `rlang::last_error()`) no longer includes the parent errors. On the
  other hand, the full backtrace (as printed by `rlang::last_trace()`)
  now includes the backtraces of the parent errors.

* `cnd_signal()` has improved support for rlang errors created with
  `error_cnd()`. It now records a backtrace if there isn't one
  already, and saves the error so it can be inspected with
  `rlang::last_error()`.

* rlang errors are no longer formatted and saved through
  `conditionMessage()`. This makes it easier to use a
  `conditionMessage()` method in subclasses created with `abort()`,
  which is useful to delay expensive generation of error messages
  until display time.

* `abort()` can now be called without error message. This is useful
  when `conditionMessage()` is used to generate the message at
  print-time.

* Fixed an infinite loop in `eval_tidy()`. It occurred when evaluating
  a quosure that inherits from the mask itself.

* `env_bind()`'s performance has been significantly improved by fixing a bug
  that caused values to be repeatedly looked up by name.

* `cnd_muffle()` now checks that a restart exists before invoking
  it. The restart might not exist if the condition is signalled with a
  different function (such as `stop(warning_cnd)`).

* `trace_length()` returns the number of frames in a backtrace.

* Added internal utility `cnd_entrace()` to add a backtrace to a
  condition.

* `rlang::last_error()` backtraces are no longer displayed in red.

* `x %|% y` now also works when `y` is of same length as `x` (@rcannood, #806).

* Empty named lists are now deparsed more explicitly as
  `"<named list>"`.

* Fixed `chr()` bug causing it to return invisibly.


# rlang 0.4.0

## Tidy evaluation

### Interpolate function inputs with the curly-curly operator

The main change of this release is the new tidy evaluation operator
`{{`.  This operator abstracts the quote-and-unquote idiom into a
single interpolation step:

```
my_wrapper <- function(data, var, by) {
  data %>%
    group_by({{ by }}) %>%
    summarise(average = mean({{ var }}, na.rm = TRUE))
}
```

`{{ var }}` is a shortcut for `!!enquo(var)` that should be easier on
the eyes, and easier to learn and teach.

Note that for multiple inputs, the existing documentation doesn't
stress enough that you can just pass dots straight to other tidy eval
functions. There is no need for quote-and-unquote unless you need to
modify the inputs or their names in some way:

```
my_wrapper <- function(data, var, ...) {
  data %>%
    group_by(...) %>%
    summarise(average = mean({{ var }}, na.rm = TRUE))
}
```


### More robust `.env` pronoun

Another improvement to tidy evaluation should make it easier to use
the `.env` pronoun. Starting from this release, subsetting an object
from the `.env` pronoun now evaluates the corresponding symbol. This
makes `.env` more robust, in particular in magrittr pipelines. The
following example would previously fail:

```
foo <- 10
mtcars %>% mutate(cyl = cyl * .env$foo)
```

This way, using the `.env` pronoun is now equivalent to unquoting a
constant objects, but with an easier syntax:

```
mtcars %>% mutate(cyl = cyl * !!foo)
```

Note that following this change, and despite its name, `.env` is no
longer referring to a bare environment. Instead, it is a special
shortcut with its own rules. Similarly, the `.data` pronoun is not
really a data