-
The internal parsers are no longer part of the public roxygen2 interface.
-
@section
titles and@export
tags can now only span a single line to prevent common bugs. -
Parsing is stricter: many issues that were previously warnings are now errors. All errors should now give you the line number of the roxygen block associated with the error.
-
Every input is now checked to make sure that you have matching braces (e.g. every
{
has a matching}
). This should prevent frustrating errors that require carefully reading of.Rd
files. (#183) -
Roxygen2 will now never overwrite a file that was not generated by roxygen2. This means that the first time you use this version of roxygen, you'll need to delete all existing Rd files.
roxygenise()
gains a clean argument that will automatically remove any files previously created by roxygen2. -
Roxygen2 now adds a comment to all generated files so that you know they've been generated, and should not be hand edited.
-
Usage statements in generated roxygen statements non-longer contain non-ASCII characters and will be wrapped if long. (#180)
-
New
@field
tag for documenting the fields on a reference class. (#181) -
Roxygen2 no longer wraps the text in Rd files by default, i.e. the default option is
wrap = FALSE
now. To override it, you have to specify a fieldRoxygen: list(wrap = TRUE)
in DESCRIPTION. (#178)
It's now possible to document reference classes, using the "docstring"
convention described in ?setRefClass
. If you want to provide a short
paragraph description of what a method does, make the first component of the
message a string containing the description, e.g.:
setRefClass("A", methods = list(
f = function(a, b) {
"Take numbers \code{a} and \code{b} and add them together"
a + b
}
))
Unlike the documentation for R functions, the documentation for methods can be quite succinct.
Roxygen adopts the convention that documented methods are public, and will be listed in the man page for the object. Undocumented methods are private and will not be shown in the documentation. The methods for all superclasses are also listed, so that you don't need to flip through multiple pages of documentation to understand what you can do with an object. All documented methods will be placed in a bulleted list in a section titled "Methods", the method usage will be automatically prepended to the docstring.
-
Fixes for Rcpp 0.11.0 compatibility.
-
roxygenise()
now invisible returns a list of all files generated by individual roclets. This is useful for tools that want to figure out if there are extra files in theman/
directory. -
is_s3_generic()
now recognises group generics (#166). -
Don't try and add parameters for data objects (#165).
-
Sort output of families using C locale (#171).
-
@family
now escapes function names in references (#172).
Roxygen2 now fully supports S4 and RC (reference classes) - you should no
longer need to manually add @alias
or @usage
tags for S4 classes, methods
and generics, or for RC classes.
-
The default usage definitions are much better, generating the correct usage for data sets (#122), S3 methods (without additional
@method
tag), S4 generics, S4 methods, and for replacement (#119) and infix functions. Backslashes in function arguments in are correctly escaped. Usage statements also use a more sophisticated line wrapping algorithm so that they should cause fewer problems with the R CMD check line limit. (#89, #125). -
S4 classes, S4 methods, and RC classes are given better default topics, and the file names corresponding to those topics are shorter.
-
S4 methods will automatically inherit parameter documentation from their generic.
-
@slot name description
allows you to document the slots of a S4 class.
S3 support has also been improved: roxygen2 now figures out whether a function
is a S3 method or generic. (In the rare cases it does so incorrectly, use
@method
to manually describe the generic and class associated with a method).
This means you can remove existing uses of @method
, and can replace
@S3method
with @export
.
Roxygen now has support for package specific options through the Roxygen
field in the DESCRIPTION
. The value of the field should be R code that
results in a list. Currently only wrap
and roclet
values are supported:
-
Turn off Rd re-wrapping with adding
Roxygen: list(wrap = FALSE)
-
Change the default roclets by specifying
Roxygen: list(roclets = c("collate", "rd"))
Roxygen 3.0 also includes a number of minor fixes and improvements:
-
Infix functions are now escaped correctly in the
NAMESPACE
. (Thanks to @crowding, #111) -
roxygenise()
now works more likedevtools::document()
and only ever works in the current directory. The argumentsroxygen.dir
,overwrite
,copy.package
andunlink.target
have been deprecated due to potential data loss problems. -
The collate roclet is no longer a roclet: it processes R files using custom code (only statically, not dynamically) and is designed to be executed before the code is sourced. Run
update_collate()
to update the Collate directive based on@include
tags - if there are none present, a collate directive will not be generated. -
@useDynLib
now works with more possible specifications - if you include a comma in the tag value, the output will be passed as is. This means that@useDynLib mypackage, .registration = TRUE
will now generateuseDynLib(mypackage, .registration = TRUE)
in the NAMESPACE. (#124) -
inst
directory not created by default (#56). -
Explicitly depend on
utils
andmethods
packages to make roxygen compatible withRscript
(#72). Importdigest
package instead of depending on it. -
Always use C locale when sorting
NAMESPACE
file or tags in.Rd
files. This ensures a consistent ordering across systems (#127). -
Templates with extension
.r
are supported on case-sensitive file systems (#115). Template variables now actually work (#160, thanks to @bronaugh). -
Suppress default aliases, format and usage with
@aliases NULL
,@format NULL
and@usage NULL
.
-
Correctly use keyword
datasets
notdataset
(Fixes #60) -
Reference classes no longer given incorrect docType (data).
-
Use unicode escapes in test files so tests pass on all platforms.
-
Work around bug in
gsub
in C locale by manually specifyingEncoding()
.
-
Package docType will automatically add package alias, if needed. (Fixes #4)
-
Data docType will automatically add
datasets
keyword, default usage, and default format. (Fixes #5). Data docType automatically added to data objects. -
New
@encoding
tag for manually setting non-ASCII encodings when needed. (Fixes #7)
-
write.description()
now tries much harder to respect users' original DESCRIPTION field formatting instead of forcibly re-wrapping certain fields at 60 characters. -
@details
and@description
now work correctly -
@useDynLib
now works correctly:@useDynLib packageName routine1 routine2
produces
useDynLib(packageName, routine1) useDynLib(packageName, routine2)
in the NAMESPACE file, instead of separate (wrong) useDynLib statements as before.
-
All namespace import directives now behave in the same way as the export directives, producing multiple single directives instead one multiple directive:
@importClassesFrom pkg a b
now producesimportClassesFrom(pkg, a)
andimportClassesFrom(pkg, b)
-
In example files included with
@example
you can now use infix operators (e.g. %*%) or other things with %, because they will be preceded by a backslash in the Rd file. This behaviour was already in place for examples directly included with@examples
. -
Aliases are no longer quoted, and % is escaped with a backslash (Fixes #24). Names also have % escaped (Fixes #50)
-
Replacement functions (e.g.
foo<-
) now get correct usage statements:foo() <- value
instead offoo()<-value
. (Fixes #38) -
Functions with no arguments now correctly get usage statements (Fixes #35)
-
Indentation in examples now preserved (Fixes #27)
-
roxygen2 will replace characters that are not valid in filenames with a character substitute, e.g.
[]
becomessub
,<-
becomesset
(Fixes #6) -
Usage strings use non-breaking spaces to prevent string default values containing whitespace to be split across multiple lines. This may cause problems in the unlikely event that you have default value containing a non-breaking space (`"\uA0"') (Fixes #21)
-
Functions with quoted names now get correct usage statements (Fixes #41)
-
Objects that no longer exist are not documented (Fixes #42)
-
Errors now display file name and line number of roxygen block to help you find the problem. Thanks to code contributions from Renaud Gaujoux. (Fixes #13)
-
Documentation with no untagged text but with
@title
,@description
and@details
tags now produces correct output.
-
package dependencies loaded automatically
-
added support for the
@source
tag
-
NAMESPACE
file no longer needs to exist -
Collate
field inDESCRIPTION
no longer needs to exist -
=
now recognised as way of assigning functions -
x$y <- function() {...}
no longer causes error -
@example
no longer added extra new-lines. -
Correct directory normalisation under windows fixes broken test.
-
A special thanks goes to Yihui Xie who contributed all of the fixes and improvements (bar one) in this version!
-
now works with run-time details to give more accurate output. This requires that the source code that roxygen is documenting be loaded prior to documentation. roxygen will attempt to do so, but you need to ensure required packages are loaded.
Run-time data fixes some long standing bugs where roxygen couldn't correctly figure out function usage. We are not aware of any cases where you still need to use the
@usage
tag. -
written in idiomatic R, and uses S3 instead of a homegrown class system.
-
roclets build up an internal data structure instead of writing to disk directly. This means that you can now use the
@rdname
tag to merge documentation for multiple functions into one file, and that only unique namespace directives are written toNAMESPACE
(which makes@importFrom
much more useful). -
some features have been removed, and may or may not (based on your feedback) be reincluded. These include the callgraph roclet, and
R CMD roxygen
, which only worked on some systems. -
a templating system: use the
@template
tag to insert abrew
template stored inman-roxygen
. Template variables can be set using@templateVar name value
and retrieved from within the template with<%= name %>
-
extensive use of caching to make repeated runs as fast as possible. To clear caches and guarantee a complete rebuild, use
clear_caches()
. -
parsing of "introduction" (the text before the first tag) has changed. Now the title consists of the first paragraph (i.e. all text before the first empty line), the second paragraph is the description and all others are put in the details. Any component can be overridden with
@title
,@description
and@details
as appropriate.
-
@name
is always output as an alias, even if@aliases
are used. -
@export
correctly uses@method
to generate@S3method
namespace directive
-
@rdname filename
sets the output filename (without extension). Use for functions non-alphanumeric functions (e.g.[<-
) or if you want to document multiple functions in one file -
@template templatename
includes a documentation template (see above) -
@section Section title: contents
includes a section with any title. Don't forget the colon! That separates the title of the section from its contents. -
@description
and@details
tags allow you to specify description and details components in a template -
@family family name
automatically adds see-also cross-references between all functions in a family. A function can belong to multiple families. -
@inheritParams name
allows you to inherit the documentation for parameters from another function, either within the current package (function
) or in any other installed package (package:function
). Currently only supports single inheritance (i.e. you can't inherit from a function that inherits from another function), but you can have multiple @inheritParams tags. -
@format
has been implemented; it existed in the roxygen package but was actually ignored