When building a package, I received the following warning:
* checking PDF version of manual ... WARNING
LaTeX errors when creating PDF version.
This typically indicates Rd problems.
I have no idea how to even begin diagnosing this. Is there a tool that tells me what .Rd file the problem is in?
I get no warnings about any of my Rd files in the checking documentation step....
Try R CMD Rd2pdf mypackage to create the manual, and possibly also set the --no-clean option to keep the temporary files. This should allow you to debug the LaTeX code triggering the error.
Though #Dirk's answer also helped me to fix the problem I would like to add a bit which might especially help recent updaters. That is, people who haven't encountered other LaTeX / R troubles after the update to 3.1.3 yet. The problem is little bit more general than just building. For me, on OS X the problem was that R CMD Rd2pdf as well as the R CMD CHECK expected texi2dvi to be in /usr/local/bin while it was in /usr/bin.
A symlink helped to fix the problem. On terminal type:
# to check whether the same issue exists for you
which texi2dvi
# if so
cd /usr/local/bin
ln -s /usr/bin/texi2dvi
Of course if the first line returns something else, you need to adapt the symlink here.
Concluding from the comments and from my own experience the problem often seems to be that some TeX fonts are missing, most often
inconsolata.sty and
upquote.sty
First you have to find the right directory where TeX fonts are stored - in my case this is:
C:\Program Files\R\R-3.3.0\share\texmf\tex\latex
Then you can download them here:
https://www.ctan.org/tex-archive/fonts/inconsolata/?lang=en
https://www.ctan.org/tex-archive/macros/latex/contrib/upquote?lang=en
Just copy them to the respective folder and in many cases the problem will be solved (in my case too). This should work for all operating systems.
...and another reason is that you haven't installed MikTex yet.
Download MikTex from here and follow the dialog prompts to install. I found the defaults to be reasonable and worked well for me.
Try to build your R package again. It should be OK now.
In my case, I had no error when running devtools::check() nor devtools::document() but when running R CMD check mypackage_version.tar.gz I got an error:
* checking PDF version of manual ... WARNING
LaTeX errors when creating PDF version.
This typically indicates Rd problems.
LaTeX errors found:
* checking PDF version of manual without hyperrefs or index ... ERROR
In this question in RStudio Community they point to a problem with LATEX installation.
I have the LATEX installation suggested in R markdown cookbook: TinyTex.
I fixed the issue by running in the R console
tinytex::latexmk(file = "../mypackage.Rcheck/mypackage-manual.tex")
This command automatically updated my LATEX installation so the output file mypackage-manual.pdf was created. After this, I did not get any other error related to PDF when running R CMD check:
* checking PDF version of manual ... OK
* DONE
If you are on Ubuntu just install Tex Live by this command:
apt-get install texlive and restart Rstudo if you use it.
First, #dirk-eddelbuettel's approach in the current question identified the missing tex package (which was "makeindex" in my case).
system("R CMD Rd2pdf --no-preview --output=./documentation-peek.pdf ." )
# ... <omitted pages of output> ...
# Warning in sys2(makeindex, shQuote(idxfile)) : '"makeindex"' not found
# Error in texi2dvi(file = file, pdf = TRUE, clean = clean, quiet = quiet, :
# unable to run 'makeindex' on 'Rd2.idx'
# Error in running tools::texi2pdf()
Then #pedro-lima's answer in https://stackoverflow.com/a/69968791/1082435 worked for my specific case.
tinytex::tlmgr_install("makeindex")
Oftentimes this error occur because of Unicode characters in the package. In this cases, the error message might look like this.
See the inputenc package documentation for explanation.
Type H <return> for immediate help.
! Package inputenc Error: Unicode character (U+009D)
(inputenc) not set up for use with LaTeX.
You can find any Unicode characters in your package using tools::showNonASCIIfile(). Here's a simple way to check for these characters in your functions and documentation:
# functions
functions <- list.files(path = './R', all.files = T, recursive = T, full.names = T)
lapply(X=functions, FUN = tools::showNonASCIIfile)
# documentation
docs <- list.files(path = './man', all.files = T, recursive = T, full.names = T)
lapply(X=docs, FUN = tools::showNonASCIIfile)
Related
I'm working on a very recent Windows 10 build, RStudio 1.3.959 and I've just installed the latest MikTex.
I'm trying to put together an R package using RStudio. I can build the package and the Function documentation comments are being converted into /man/*.Rd files. These are then successfully displayed when one executes ?function_name in the RStudio console window.
Unfortunately, I'm having very little luck building the PDF package API documentation (not to be mistaken as the vignette; which I can build). I've looked over a good few tutorials but they all stop short of instructing how one builds the final PDF API document that one expects with every R package.
I've tried:
Build[Windows]->More->Document ... which execute devtools::document(roclets = c('rd', 'collate', 'namespace', 'vignette'))
Build[Windows]->More->Build Source Package ... which executes devtools::document(roclets = c('rd', 'collate', 'namespace', 'vignette')) followed by devtools::build(binary = TRUE, args = c('--preclean'))
Build[Windows]->More->Build Binary Package ... which executes devtools::document(roclets = c('rd', 'collate', 'namespace', 'vignette')) followed by devtools::build(binary = TRUE, args = c('--preclean'))
All three function as expected but still no final package manual pdf file.
Doing some digging on Stack I noticed someone used the command:
devtools::build_manual()
I'm convinced this is what I need. However, when I execute that line of code I get the error:
Converting Rd files to LaTeX ...
Warning in sys2(makeindex, shQuote(idxfile)) : '"makeindex"' not found
Error in texi2dvi(file = file, pdf = TRUE, clean = clean, quiet = quiet, :
unable to run 'makeindex' on 'Rd2.idx'
Warning in sys2(makeindex, shQuote(idxfile)) : '"makeindex"' not found
Error in texi2dvi(file = file, pdf = TRUE, clean = clean, quiet = quiet, :
unable to run 'makeindex' on 'Rd2.idx'
Error in running tools::texi2pdf()
Error: Failed to build manual
Which has left me none the wiser, although it's quite clear that something's a bit upset by the absence of makeindex. Help is much appreciated.
If I understand correctly, you mean the standard reference manual. For example ggplot2 reference manual. In my experience, this is thrown together by CRAN when you submit. All the manual is (as far as I can tell) a collection of things that a well documented package should have, such as a DESCRIPTION file, a NAMESPACE file, and the various .RD files for actual documentation.
Even when looking at the public github for ggplot2 we see that they do not have the manual in their repository. Additionally, there isn't any evidence in their .git*ignore files to suggest they made the manual themselves.
However, if you want to make this yourself devtools::build_manual() is the correct function call.
I was able to make the manual with a preexisting package on github. I would suggest trying to reinstall your devtools package and make sure there are no warnings or errors. It may be helpful to run a session as administrator to insure things get installed correctly.
Good luck!
I am using R version 3.4 with RStudio 1.1.453 on MacOS High Serra.
R CMD check packagename_0.1.1.tar.gz --as-cran produces a WARNING if badges are present in README.md. This causes consequences like travis CI to fail (since it treats warnings as errors)
Is there a workaround to get the check to pass without removing the badges?
Here is an example of the WARNING I see
* checking top-level files ... WARNING
Conversion of ‘README.md’ failed:
pandoc: Could not fetch https://www.r-pkg.org/badges/version/bigrquery
TlsExceptionHostPort (HandshakeFailed (Error_Misc "user error (unexpected type received. expecting handshake and got: Alert [(AlertLevel_Fatal,HandshakeFailure)])")) "www.r-pkg.org" 443
Related: pandoc: Could not fetch http://www.r-pkg.org/badges/version/package TlsExceptionHostPort (HandshakeFailed
This is a workaround to embed (static) badges in pages. They do not auto-update. The purpose is to avoid offline derived errors.
Take a look at readme files in:
https://gitlab.com/ferroao/idiogramFISH. This also handles errors while installing with devtools in windows, a process that does not have online connectivity (see vignette index.Rmd). Look at DESCRIPTION also if you want to see dependencies and vignette builders.
Generate the .md from the .Rmd (Rmarkdown).
Months on, I discovered the problem. When I installed anaconda, it changed the version of pandoc my system used - which pandoc should return something like /usr/local/bin/pandoc (and not like /Users/st/anaconda3/bin/pandoc).
In my case, I edited by .bash_profile to exclude ananconda, closed and reopened the terminal and everything worked. (I also reinstalled pandoc from here, but I don't think that had anything to do with fixing it - it did show me the default installation location though)
I recently updated R to the latest release: 3.4.0. R is installed on a network location H:/. Now something weird is happening when I try to install a local binary package:
filename <- paste0("R:/path/independeR_", versions, ".zip")
install.packages(filename,
repos = NULL, type = "source",
lib = gsub("\\\\\\\\networkpath/home/[[:alpha:]]*/",
"H:/", .libPaths()[1]))
Both H:/ and R:/ are network locations. In .libPaths() the default location is in the H:/ location, but it shows with the entire network adress. In the call to install.packages I substituted this.
The code above fails, with the following output:
'\\networkpath\home\JDUB~PN6\DOCU~UZL\R\R-34~TN4.0' is not recognized as an internal or external command,
operable program or batch file.
Warning in install.packages :
running command '"//networkpath/home/JDUB~PN6/DOCU~UZL/R/R-34~TN4.0/bin/x64/R" CMD INSTALL -l "H:\Documents\R\R-3.4.0\library" "R:/path/independeR_0.1.8.zip"' had status 1
Warning in install.packages :
installation of package ‘R:/path/independeR_0.1.8.zip’ had non-zero exit status
There is two things that surprise me here. The directorynames are all jumbled up (DOCU~UZL instead of Documents etc), but for some reason Command promt seems fine with that. The more interesting thing is the following:
When I try to put the command "//networkpath/home/JDUB~PN6/DOCU~UZL/R/R-34~TN4.0/bin/x64/R" CMD INSTALL -l "H:\Documents\R\R-3.4.0\library" "R:/path/independeR_0.1.8.zip" directly into Command Prompt, the output is very similar:
C:\Users\jdubbeldam>"//networkpath/home/JDUB~PN6/DOCU~UZL/R/R-34~TN4.0/b
in/x64/R" CMD INSTALL -l "H:\Documents\R\R-3.4.0\library" "R:/path/independeR_0.1.8.zip"
'\\networkpath\home\JDUB~PN6\DOCU~UZL\R\R-34~TN4.0' is not recognized as
an internal or external command,
operable program or batch file.
For some reason CMD seems to cut off the path to the command halfway through. I guess that this is because the command is too long. When I try the same command, but with H:/Documents/R/R-3.4.0/bin/x64/R, the installation goes just fine.
I would like to be able to automatically install this package from a script, so I would like a solution to this problem from within R. Is there a way to get R to use the shorter H:/Documents/R/R-3.4.0/bin/x64/R?
I was having problems with updating packages, while searching I found that there is a reported bug affecting getting the timestamps from files in Windows 10:
https://bugs.r-project.org/bugzilla3/show_bug.cgi?id=17261
Starting in R 3.4.0, file.info sometimes returns for mtime,
atime and ctime for directories. It seems to have something to do
with sharing. This affects functions that use file.mtime like
update.packages.
Reproduce:
Try file.info() with a random directory. If it returns a legitimate
file time, open a windows explorer window and navigate to the
directory, then run the call again and it will return s.
In some cases it returns s even if the directory is not open (or
in the path of an open explorer window), but this is not consistent.
With a comment of Tomas Kalibera:
Thank you for the report, this is a known bug that has been fixed
recently in R-devel
I found that when I closed the file explorer windows the timestamps worked correctly and didn't show up as NA's.
This may be a lead for why your code isn't working.
I managed to make the following work. However, I think it is extremely ugly, and would still like to see if it is possible to do it someway else.
filename <- paste0("R:/path/independeR_", versions, ".zip")
cmd <- file.path(gsub("//networkpath/home/[[:alnum:]]*/", "H:/",
gsub("//networkpath/home/[[:alnum:]]*~[[:alnum:]]*/",
"H:/", R.home())), "bin/x64/R")
libname <- gsub("\\\\\\\\networkpath/home/[[:alpha:]]*/",
"H:/", .libPaths()[1])
call <- paste(paste0('"', cmd, '"'),
"CMD",
"INSTALL",
"-l",
paste0('"', libname, '"'),
"--no-lock",
paste0('"', filename, '"'),
sep = " ")
system(call)
My texi2dvi is apparently in a place where R CMD Rd2pdf doesn't expect it. Mine is at /usr/local/bin/texi2dvi, and it's being looked for at /usr/local/opt/texinfo/bin/texi2dvi:
[KenMacBook:~/git] % \R CMD Rd2pdf missing
Hmm ... looks like a package
Converting Rd files to LaTeX
Creating pdf output from LaTeX ...
Error in texi2dvi(file = file, pdf = TRUE, clean = clean, quiet = quiet, :
Running 'texi2dvi' on 'Rd2.tex' failed.
Messages:
sh: /usr/local/opt/texinfo/bin/texi2dvi: No such file or directory
Output:
Error in texi2dvi(file = file, pdf = TRUE, clean = clean, quiet = quiet, :
Running 'texi2dvi' on 'Rd2.tex' failed.
Messages:
sh: /usr/local/opt/texinfo/bin/texi2dvi: No such file or directory
Output:
Error in running tools::texi2pdf()
I can work around this by running R_TEXI2DVICMD=/usr/local/bin/texi2dvi R CMD Rd2pdf, and then the docs are built correctly.
I'd like to put that setting in my .Rprofile so that things like RStudio (which won't read my .zshrc) and other random R sessions will see the setting. But neither of the following seems to have any effect in my .Rprofile:
Sys.setenv(R_TEXI2DVICMD='/usr/local/bin/texi2dvi')
options(texi2dvi='/usr/local/bin/texi2dvi')
I'm guessing .Rprofile doesn't get read by R CMD commands, is that correct? Is there an appropriate place to put my settings?
UPDATE:
Since Dirk doubts my doubting of .RProfile for affecting R CMD Rd2pdf :-), here's my evidence:
[KenMacBook:~/git] % tail -n2 ~/.Rprofile
Sys.setenv(TEXI2DVI='/no/where')
cat("End of RProfile\n")
[KenMacBook:~/git] % Rscript -e '2+2'
End of RProfile
[1] 4
[KenMacBook:~/git] % R CMD Rd2pdf missing
Hmm ... looks like a package
Converting Rd files to LaTeX
Creating pdf output from LaTeX ...
Error in texi2dvi(file = file, pdf = TRUE, clean = clean, quiet = quiet, :
Running 'texi2dvi' on 'Rd2.tex' failed.
Messages:
sh: /usr/local/opt/texinfo/bin/texi2dvi: No such file or directory
Output:
Error in texi2dvi(file = file, pdf = TRUE, clean = clean, quiet = quiet, :
Running 'texi2dvi' on 'Rd2.tex' failed.
Messages:
sh: /usr/local/opt/texinfo/bin/texi2dvi: No such file or directory
Output:
Error in running tools::texi2pdf()
Notice that the file's settings are respected in a normal R session, but setting TEXI2DVI has no effect here.
That seems wrong as /usr/local/bin/texi2dvi should be in the $PATH.
I have
edd#max:~$ grep texi2dvi /etc/R/Renviron
## used for options("texi2dvi")
R_TEXI2DVICMD=${R_TEXI2DVICMD-${TEXI2DVI-'/usr/bin/texi2dvi'}}
edd#max:~$
Note that if you want to set the TEXI2DVI environment variable, you probably have to do start before you start R -- think ~/.bash_profile
Here is an example explicitly setting TEXI2DVI:
edd#max:/tmp$ TEXI2DVI=/no/where R CMD Rd2pdf Rcpp-package.Rd
Converting Rd files to LaTeX ...
Rcpp-package.Rd
Creating pdf output from LaTeX ...
Error in texi2dvi(file = file, pdf = TRUE, clean = clean, quiet = quiet, :
Running 'texi2dvi' on 'Rd2.tex' failed.
Messages:
sh: 1: /no/where: not found
Output:
Error in texi2dvi(file = file, pdf = TRUE, clean = clean, quiet = quiet, :
Running 'texi2dvi' on 'Rd2.tex' failed.
Messages:
sh: 1: /no/where: not found
Output:
Error in running tools::texi2pdf()
edd#max:/tmp$
As you can see, it is respected.
Edit: Also let's not forget Renviron and Renviron.site so you have plenty of choices so set his.
Edit 2: As you seem to doubt ~/.Rprofile:
edd#max:~$ tail -1 .Rprofile
cat("End of .Rprofile\n")
edd#max:~$ Rscript -e '2+2'
End of .Rprofile
[1] 4
edd#max:~$
Had the same problem and figured out how to fix it - I think it has something to do with a previous installation of macports interfering with the path when R has been installed using brew (assuming you're on OSX).
Run the following in terminal:
defaults write com.apple.finder AppleShowAllFiles TRUE
Then go Apple > Force Quit > Finder > Relaunch. You'll now be able to see hidden files.
In your user directory there may be a file named .profile, in this file I commented out the line (i.e. put a # in front of it, as shown).
#export PATH=/opt/local/bin:/opt/local/sbin:$PATH
Then navigate to your R.home() (get this by running R.home() in R)
R.home()
[1] "/usr/local/Cellar/r/3.2.4_1/R.framework/Resources"
And modify the following line in Renviron
R_TEXI2DVICMD=${R_TEXI2DVICMD-${TEXI2DVI-'/usr/local/bin/texi2dvi'}}
This fixed it for me.
To put finder back to usual run
defaults write com.apple.finder AppleShowAllFiles FALSE
in the terminal and relaunch it.
Hope that helps.
The back-and-forth with Dirk produced some solutions, but they're pretty buried, so I summarize them here.
Diagnosis: etc/Renviron contains stale info.
My /usr/local/Cellar/r/3.2.2_1/R.framework/Versions/3.2/Resources/etc/Renviron file (installed using Homebrew) contains this line:
R_TEXI2DVICMD=${R_TEXI2DVICMD-${TEXI2DVI-'/usr/local/opt/texinfo/bin/texi2dvi'}}
That's a remnant of someone (possibly me, possibly Homebrew's R package creator) who installed MacTeX in the default location, and then that path got frozen in time in the Renviron file. My texi2dvi is now at /usr/local/bin/texi2dvi, so this value needs to be overridden somehow.
1) $HOME/.Rprofile and $HOME/.Renviron won't help.
They don't take effect soon enough for R to notice them. R sets options("texi2dvi") based on the environment it sees at startup,
% tail -n2 ~/.Rprofile
Sys.setenv(TEXI2DVI='/no/where')
options(texi2dvi='/no/where/else')
% cat ~/.Renviron
TEXI2DVI=/no/where/at/all
% R CMD Rd2pdf myPackageDirectory # Still no joy
Hmm ... looks like a package
Converting Rd files to LaTeX
Creating pdf output from LaTeX ...
Error in texi2dvi(file = file, pdf = TRUE, clean = clean, quiet = quiet, :
Running 'texi2dvi' on 'Rd2.tex' failed.
Messages:
sh: /usr/local/opt/texinfo/bin/texi2dvi: No such file or directory
...
See "Initialization at Start of an R Session" for more info about startup files, though as shown above, the information in that document about overriding R_HOME/etc/Renviron is either incorrect or incomplete for this situation. Perhaps it should be amended in the section about R_CHECK_ENVIRON and R_BUILD_ENVIRON to also include something about R CMD RD2*, but I'm not sure whether that's what's going on, I only know this isn't a solution.
2) $HOME/.zshrc (and friends) won't help.
On OS X, your shell startup file is not consulted when you launch GUI apps. You could use defaults write or launchctl setenv to change the TEXI2DVI variable so that it's set when R launches, but you'd also have to stick it in your shell startup file for processes not started by launchd, which is icky. I also version my dotfiles, and I don't like sticking this bit of configuration in the launchctl ether where I can't easily remember it's there. But launchctl is presumably one solution to this.
3) etc/Renviron.site won't help.
This one is surprising - I expected it to work:
% cat /usr/local/Cellar/r/3.2.2_1/R.framework/Resources/etc/Renviron.site
R_TEXI2DVICMD=/usr/local/bin/texi2dvi
TEXI2DVI=/usr/local/bin/texi2dvi
% \R CMD Rd2pdf myPackageDirectory
Hmm ... looks like a package
Converting Rd files to LaTeX
Creating pdf output from LaTeX ...
Error in texi2dvi(file = file, pdf = TRUE, clean = clean, quiet = quiet, :
Running 'texi2dvi' on 'Rd2.tex' failed.
Messages:
sh: /usr/local/opt/texinfo/bin/texi2dvi: No such file or directory
So Renviron.site's settings aren't taking effect here. I'm getting impatient, so I didn't try to diagnose why.
3) Editing etc/Renviron as a last resort
So this finally works:
% grep TEXI2DVI /usr/local/Cellar/r/3.2.2_1/R.framework/Resources/etc/Renviron
TEXI2DVI=/usr/local/bin/texi2dvi ## Added by Ken
R_TEXI2DVICMD=${R_TEXI2DVICMD-${TEXI2DVI-'/usr/local/opt/texinfo/bin/texi2dvi'}}
I don't like it much, because I'll lose those settings next time I upgrade R. And R's documentation specifically says "do not change ‘R_HOME/etc/Renviron’ itself". But at least it works.
I am new to the R programming language and am having basic issues with it. I want to untar a file, but it has not been able to work for me.
Here is the code that I enter:
untar("CD_data.tar", exdir="data")
It then returns the following error message:
/bin/sh: /usr/bin/gnutar: No such file or directory
Warning message:
In untar("CD_data.tar", exdir = "data") :
‘/usr/bin/gnutar -xf 'CD_data.tar' -C 'data'’ returned error code 127
Please help! Thanks!
R on OS X 10.9 (Mavericks) seems to set a wrong TAR environment variable.
You can fix this by adding the following to your .Rprofile (or executing it manually):
Sys.setenv(TAR = '/usr/bin/tar')
Alternatively, you can provide the tar path as an argument when calling untar.
My 2 cents is that you are using a mac and have not installed tar. You are getting value 127 because the command is not found within your $PATH and it's not a built-in command (which is usually the case if you were in unix...
In other words you need to install tar.
Or run it in linux.