fixed issues causing ‘lost braces’ (actually, superfluous braces) NOTEs from checks in R-devel occuring with some accented LaTeX characters. These NOTEs are not yet activated on CRAN but if/when they do, developers using Rd macros \insertRef and \insertCite can eliminate them by building their packages with Rdpack (> 2.5) and rbibutils (>= 2.2.16). Thanks to Sebastian Meyer for tracing down the issues.
fixed a Sweave expression in ‘man/predefined.Rd’ to not emit unnecessary braces (see the note above).
Rd macros insertCite and friends were calling insert_citeOnly with argument key instead of keys. This was not an error since partial matching is ok here but not good practice. Fixes issue #28 reported by Marcel Ramos.
changed the saved value in “tests/testthat/dummyArticle.rds” used in ‘test-bib.R’ with a value obtained with R-devel r84896 n previu versins of R there wasasuperfluous final ’ .’).
in some cases when square brackets were requested, rather than round ones, insertCiteOnly was using closing round parenthesis.
in some cases insertCiteOnly was not handling well the last cite when it was followed by free text.
new Rd macro \insertCited{} works like \insertAllCited but empties the references list after finishing its work. This means that the second and subsequent \insertCited in the same help page will list only citations done since the preceding \insertCited. Prompted by issue #27 to allow separate references lists for each method and the class in R6 documentation.
It is no longer necessary to remove backslashes from \& and similar in bibtex files (e.g. in “John Wiley & Sons”). The backslash was appearing in the rendered pdf manual but R-devel’s checks started to warn about it recently with something like checkRd: (-1) pcts-package.Rd:287: Escaped LaTeX specials: \& (starting June 2022). This could be avoided by removing the superfluous backslash in the bib file but this is annoying since this is proper syntax for the latter. Concerned packages should rebuild the tarball with Rdpack (> 2.3.1) or remove the offending backslashes from their bib files. (addresses issue #26)
\insertCite, two or more accented Latin characters were rendered with superfluous curly braces, see issue#25 (reported by Manuel López-Ibáñez) and https://stat.ethz.ch/pipermail/r-devel/2022-April/081604.html.Rdpack and could require Rdpack >= 2.3 to ensure that users installing their package don’t do that accidentally with Rdpack 2.2.\doi Rd macros resulted in warnings from R-devel’s checks for packages that use Rdpack built with the released versions of R. This fixes it. (fixes #24)Rdmacro \insertCite gets a new feature allowing to drop parentheses for parenthesised citations, analogous to natbib’s \citealp in latex. This makes it possible to use markup around the citations outside the arguments of the macro. This fixes github issue #23 by @ms609.
updated vignette “Inserting_bibtex_references” and other documentation.
the pkgdown site now has a Search button on the navigation bar.
fixed a test that was failing with versions of rbibutils before v2.1.1 (issue reported by Andreas Tille).
if the version of rbibutils is greater than v2.2.5, parsing bib files with rbibutils::readBib now uses texChars = "Rdpack". This ensures proper rendering of \'i' in references (by converting'ito'` which base R renders correctly). For detailed discussion of this see issue GeoBosh/bibutils#7 and NEWS.md in rbibutils v2.2.4 and v2.2.5.
Rdmacros generating references now drop the URL field if a doi field generating the same URL is present. This avoids repetition of the URL since rendering the doi field produces the URL anyway.
now references with identical authors are sorted by year.
fixed rebib() to process doi’s properly.
now checks on github (using github actions) are done on the three major platforms for the following versions of R: release, devel, oldrel, and 3.3.
now a warning (rather than error) is issued if package ‘testthat’ is not available for tests.
now bibtex files are parsed with option direct = TRUE when the version of rbibutils is >= ‘2.1.2’. This fixes GeoBosh/rbibutils#3 for Rdpack users (although that needs fixing also for direct = FALSE).
In viewRd, the default for the help type now is options("help_type") (previously the default was "text").
for roxygen2 users, added a note in README and Rdpack-package.Rd not to indent by exactly four spaces \insertAllCited{} relative to @references when markdown mode is activated. This is to avoid it being translated by roxygen2 as verbatim (\preformatted) text, see issue (#18).
moved package gbRd to Suggests.
fixed uses of in several examples (necessitated by a change in R-devel circa February 2021).
dramatic speed up of processing of bibliography references, most noticeable for users with large number of help pages and large REFERENCES.bib. (prompted by Kisung You, issue #17)
amended the warning message from bibliography macros to mention duplicated keys.
added entry for rbibutils in REFERENCES.bib.
corrected wrong citation of rbibutils.
bibtex from Suggests and elsewhere.imported rbibutils and made suitable adjustments to the code.
moved bibtex to Suggests (since bibtex has been orphaned for several years).
gbutils, which is not among the dependencies of Rdpack. This was raising a NOTE on one of the CRAN testing machines.viewRd() now loads also Rd macros declared by the package to which the rendered Rd file belongs. Previously only macros from Rdpack were loaded but now there is at least one other package, mathjaxr, which defines Rd macros.
moved grDevices to Suggests.
edited some documentation files.
in documentation of S4 classes, such as "classname-class.Rd", reprompt() was sometimes inserting entries for slots already listed in the Slots section. The same bug was causing the new slots not to be reported properly to the user.
in some cases reprompt() failed to process properly \S4method entries in the Usage section of Rd files. Now fixed.
reprompt() now handles \S4method statements for replacement methods. As for other functions and methods (S3 and S4), it is sufficient to put a declaration with empty argument list in the Usage section and reprompt() will insert the correct formal arguments for the method (they may be different from those of the generic).
reprompt() now gives a more helpful error message when type is invalid.
corrected some minor typo’s in the documentation.
README and the documentation of reprompt() and Rdpack-package now give more details on reprompt()ing replacement functions.
README.md and README.org now give the correct install.packages instruction for CRAN (pull request #10 from @katrinleinweber).
some examples were leaving a stray file, dummyfun.Rd, after R CMD check.
Updated the vignette about \insertFig, \printExample and \runExamples, to reflect the lifting in R 3.6.0 of some limitations of Rd processing in previous R versions.
Vignette ‘Inserting BibTeX references’ now includes a section on bibliography styles. This section was previously only in README.
The fix in Rdpack 0.10-3 (see below) for an issue introduced in R-devel in Oct 2018 will be made permanent, at least for now. This fix resolves also a similar issue in package pkgdown, see the discussion at https://github.com/GeoBosh/Rdpack/issues/9 for details and further links.
\insertAllCited{} to appear in a single paragraph in the html rendering of the Rd documentation.added pkgdown site to DESCRIPTION.
README and the vignette about evaluated examples now state that R-devel no longer gives warnings about \Sexpr not being a top level section. This means that macro runExamples which creates section ‘Examples’ containing code and results of evaluation will be useable in CRAN packages.
new function Rdo_fetch() gets the Rd object representing a help page from an installed or source package. It works also for packages under devtool’s developer’s mode
The site created with pkgdown contains fewer errors now. (Note that the documentation builds without error with R’s tools).
Started moving tests from my local setup to testthat.
now REFERENCES.bib is read-in using the declared encoding for the corresponding package. If there is no declared encoding, “UTF-8” is assumed.
Now macros \insertCite and \insertCiteOnly use the correct results=rd instead of results=Rd. This was not catched by Rs building tools but caused errors when processed with pkgdown::build_site(). Fixes issue#8. Also fixes r-lib/pkgdown#784. Thanks to Jay Hesselberth for uncovering this.
now references produced by the citation macros for BibTeX entries of type @book and @incollection treat field ‘series’ similarly to other BibTeX styles, including JSS (issue#7 raised by Kisung You). Note that even though the underlying base R tools are based on JSS, they treat this field differently.
Bugfix: now Rd macro \printExample evaluates the expressions in the correct environment.
help page of get_usage gets a fairly complete Details section with numerous evaluated examples.
now run_examples() escapes % by default, before returning the text. This is needed for text that is to be included in an Rd file. It can be turned off by setting the new argument escape to FALSE.
now reprompt() gives a more informative error message if an Rd file describes a non-existent S4 class. This is not captured by Rs tools. It can happen during development if a class is removed.
now reports printed by reprompt() about methods documented in Usage sections, but no longer existing, are more readable. This is due to a new print method for (the mostly internal) class “f_usage”.
Many features of Rdpack are best demonstrated on a package. The new package RdpackTester under ./inst/examples now makes this easier.
now the help page “predefined.Rd” does not print some tables twice in the pdf manual. (This was due to using \if{latex}{}{}' instead of’ for those tables.)
also in “predefined.Rd”, removed illegal use of vertical bars (in column specifications of tabular environments) from the pure LaTeX code in the \ifelse clause(s) and wrapped them in \out{}.
in “get_sig_text.Rd”, replace help() with utils::help() to avoid warnings from more stringent R CMD check. Similarly, in “Rdpack-package.Rd” replace packageDescription() with utils::packageDescription(). This may be needed in \Sexpr’s more generally (TODO: check if these would still be needed if the symbols are imported by the package.)
\eqn{}.new macro \printExample for inclusion of example computations in narrative sections, such as Details. The code is evaluated and printed similarly to R sessions but the code is not prefixed and the output is prefixed with comment symbols, eg.
2+2
##: 4new experimental macro \runExamples for use as top level section in an Rd file as a replacement of section \examples. \runExamples{code} evaluates the code and creates section \examples containing the code and the results (similarly to \printExample). So, \runExamples{2 + 2} produces
\examples{
2 + 2
##: 4
}The generated section examples is processed by R’s documentation tools (almost) as if it was there from the outset.
new experimental macro \insertFig to create a figure with R code and include it in the documentation. The related macro \makeFig just creates a graphics file, which can be included with the standard Rd command \figure.
new vignette gives a brief description of the new macros.
\insertRef and \insertAllCited (see below) - in the rendered News on CRAN the backslashed words had disappeared.\insertRef and \insertAllCited macros now support bibstyles for formatting references (feature requested by Jamie Halliday, issue#5). Use Rdpack (>= 0.8) in Imports: to use this feature. Currently only long author names are supported but complete support for styles can be added trivially if requested.
updates to the documentation, in particular the bulk of Rdpack-package.Rd was from 2011!
improvements to handling of free form citations in textual mode:
(bugfix) now the whole citation is not parenthesised in textual mode,
the handling for textual mode was incomplete in that additional text after the citation was not put inside the parentheses along with the year.
updates to the documentation.
fix a bug in Rdo_locate_core_section().
consolidated the changes introduced since the previous CRAN release of Rdpack (it was 0.5-5) in preparation for the next release. Users of the new macros for citation can use Rdpack (>=0.7) in the Imports: field of file “DESCRIPTION” to ensure that they are available.
comprehensive overhaul of handling of errors and warnings during processing of references and citations. In particular, such errors should (in most cases) produce only warnings during R CMD build and R CMD INSTALL, and not prevent the package from being built and installed.
Unresolved BibTeX keys produce warnings during building and installation of the package, but not errors. Dummy entries are inserted in the documentation explaining what was amiss (currently with ‘author’ A Adummy).
new Rd macros for citations
\insertCite inserts citation(s) for one or more keys and records the keys for \insertAllCited (see below).
\insertCiteOnly is similar to \insertCite but does not record the keys.
\insertNoCite records the keys but does not produce a citation.
\insertAllCited prints a bibliography including all references recorded by \insertCite and \insertNoCite.
new entries in this file will use markdown syntax.
updates to the documentation.
get_bibentries() gets a new argument “everywhere”. When is is TRUE, the default, unescaped percents are escaped in all bibtex fields, otherwise the replacement is done only in field “URL”.
removed several print() statements which were accidentally left in the code in v. 0.5-6.
updated help page of get_bibentries() and included examples.
cleaned up the imports in NAMESPACE.
insert_ref() and Rd macro \insertRef should now work ok in the presence of percent encoded symbols in URL field of a BibTeX entry. Closes issue “Unexpected END_OF_INPUT error (URL parsing?)”, see Rdpack issue 3, raised by jdnewmil.
get_bibentries() now takes care of percent encoded symbols in URLs. It now returns an object from class “bibentryRd”, which inherits from “bibentry” but has its own print method which escapes or unescapes the percent signs in URLs depending on the requested output style. This was also reported by by jdnewmil in issue#3 referenced above.
get_bibentries() now tries to load the bib file from the development directory of argument “package”, in case it is in development mode under “devtools”. (Even though the search is with system.file(), which devtools replaces with its own version, the bib file still needs to be looked for in “inst/”, i.e. it is not in the root directory as is the case in installed packages.)
Streamlined the help page of reprompt() and Rdpack-package.
new argument, edit', inreprompt()` opens the Rd file after updating it.
new function ereprompt() (e' for _edit_)* opensreprompt()with suitable defaults to replace the original Rd file and open it in an editor, otherwise equivalent toreprompt()`.
now if infile does not exist, reprompt(), and hence ereprompt(), strips the directory part, finds the root of the package directory, and looks for the file under man/ (this uses package rprojroot’). In particular, if the working directory is anywhere under the package root, infile can be simply the name of the Rd file, without path.
now README.md is generated from README.org. I changed the layout and amended the information in it.
README.* now get links to georgisemacs for an emacs function to reprompt() the Rd file at point.
viewRd() now works also when the file is from a package under devtools’ development mode on any platform (e.g. RStudio, Emacs/ESS, Rgui).
new RStudio add-in ’Reprompt`, contributed by Duncan Murdoch.
If the file being edited in the code editor is an Rd file it calls reprompt(). If the file is as R source file, it looks for the help page of a selected object name and works on it if found, otherwise creates one.
added the version of Rdpack to the abstract of the vignette. This seems more informative than giving the compilation date.
now reprompt() doesn’t give spurious warnings about unknown Rd macros when the Rd file contains user defined Rd macros. These warnings were harmless and rare, but alarming and annoying.
fixed a bug in inspect_args() which caused the warning “In is.na(x) : is.na() applied to non-(list or vector) of type ‘NULL’” in reprompt(), when the signature of a function in “Usage” section was empty.
added the github url to DESCRIPTION.
now give a more informative warning if a key is missing from REFERENCES.bib (thanks to Kisung You for suggesting this).
tools::parse_Rd() rather than parse_Rd() (before this was not necessary since ‘gbRd’ depended on ‘tools’).new function viewRd() parses and shows an Rd file in a source package.
May be particularly helpful for devtools developers.
revamped the vignette, added additional explanations of possible issues.
new functions makeVignetteReference() and vigbib() generate Bibtex references for vignettes in a package.
Added the requirement to Import Rdpack to the help page of insert_ref(). (It was out of sync with the vignette.)
Updated Description in DESCRIPTION, which was also out of sync with the vignette.
Bug fix in the vignette: changed \@references to @references in the roxygen2 example.
Edited and amended vignette “Inserting_bibtex_references”, including:
additional explanation and an example,
requirement to import Rdpack and a mention of travis etc.
paragraph about devtools.
changed the URL for parse_Rd.pdf in REFERENCES.bib to https.
new facility for inserting references from BibTeX files - just put the line:
RdMacros: Rdpackin the DESCRIPTION file of your package and the references in inst/REFERENCES.bib. Then put \inserRef{key}{yourpackage} in documentation chunks to insert item key from the bib file. This works with both manually created Rd files and roxygen comments, see the documentation for details.
from now on, put (CRAN) next to versions published on CRAN (as above)
a minor correction in file NEWS.
reprompt() and rebib().don’t export parse_text
corrected bug in reprompt for S4 classes - new slots were not handled correctly (see slots.R for details).
reprompt for S4 methods* if the methods documentation describes methods that do not exist, print an informative message for the user. (Such methods are also printed for debugging purposes but in non-user friendly form.)
included methods in Imports: - around R-devel version 2015-07-01 r68620 not including it triggers warnings)
new functions .whichtageq and .whichtagin; replaced most calls to toolsdotdotdotRdTags with these.
removed toolsdotdotdotRdTags
removed ::: calls from code; copied some functions from tools to achieve this (see “threedots.R”). Renamed the functions replacing ::: with dotdotdot.
export by name (not the generic export pattern); preparation for more selective export of functions in the future.
new functions: Rdo_get_argument_names, Rdo_get_item_labels
dependence bibtex becomes “Imports”. tools and gbRd remain (for now) “Depends” since functions from them are used in some examples without require()’s.
rebib(): now outfile="" can be used to request overwriting infile. (a small convenience; before the change, one could do this by giving the outfile and infile the same values.)
Bug fix: predefined.Rd contained \tabular environments with vertical bars, |, in the format specification. This is not documented in “Writing R exts” but works for LaTeX and remained unnoticed by me and R CMD check. However, rendering the help page for the objects documented in “predefined.Rd” gave an error in html and text mode. Package installation failed only if html was requested at build time.
small changes in the documentation
new major feature: processing references from Bibtex files. The top user level function is rebib(), which updates section “references” in an Rd file. promptPackageSexpr() has been updated to use this feature. inspect_Rdbib() can be used in programming.
new auxiliary functions for work with bibentry objects.
new convenience programming functions for Rd objects.
some small bug fixes.
some gaps in the documentation filled.
reprompt() was not handling properly S4 replacement methods. Changed the parsing of the arguments to rectify this. Some other functions also needed correction to handle this.Depends field in DESCRIPTION file updated to require R 2.15.0 or later. (because of a few uses of paste0() in recent code.)