Anyone got an idea why
system2("bash", "ls")
would result in (both in R-Gui and RStudio Console)
/usr/bin/ls: /usr/bin/ls: cannot execute binary file
while other shells work w/o issues (see example below) and the "bash" in the RStudio Terminal also works w/o problems. AND if there is a remedy for that?
Working system2 example:
system2("powershell", "dir")
Running R-3.6.3 on Windows 10 here ... with "Git bash" executing as "bash".
PS: just in case someone wonders - I also tried the full path ie /usr/bin/ls -> same result
Ok - this does the trick
system2("bash", input = "ls")
so instead of args = ... as with the powershell command one needs (or more interestingly sometimes 'can') use the input = ... parameter with bash
OR as mentioned in the comments by #oguzismail this will also work
system2("bash", "-c ls")
as well as pointed out by #Christoph
system2("ls")
also works sometimes (ie in some RStudio[?] setups it will not return results in some it will).
But I have come to notice that different combinations of R-versions with different RStudio versions [or more probably Locales] will not always behave consistently of course also depending on the availability of bash commands like ls (more generally /usr/bin) being on the PATH or not.
So choose what suits u best!
Related
I have the following code that should open the Recode.xlsx inside the subf folder but doesn't
write_xlsx(mtcars, "subf/Recode.xlsx")
shell("subf/Recode.xlsx", wait=FALSE)
The following code works, so if anyone has an idea on why it doesn't work it would help me
write_xlsx(mtcars, "Recode.xlsx")
shell("Recode.xlsx", wait=FALSE)
Here’s what the documentation of the shell function says (emphasis mine):
shell is a more user-friendly wrapper for system. To make use of Windows file associations, use shell.exec.
And here is another bit of relevant information from the shell.exec documentation:
To be interpreted as relative, the path also needs to use backslashes as separators (at least in Windows 10).
So the following is the correct usage:
shell.exec("subf\\Recode.xlsx")
I have a daily use case where I need to work with projects on different version of Java (8, 11, ...).
I would like to have it displayed in the right side prompt in my shell (ZSH with Oh-My-Zsh). I know of a dummy way (computationally expensive) to do it (just java --version to var and display it). I would like it to have it cached until I don't source a file (which is a specific project file that sets the new env vars for different java versions).
Do you have any ideas how to do this efficiently?
Br,
Stjepan
The PROMPT and RPROMPT variables can have both static and dynamic parts, so you can set the version there when you source the project file, and it will only be calculated one time. The trick is to get the quoting right.
This line goes in the project file that sets the env variables, somewhere after setting PATH to include the preferred java executable:
RPROMPT="${${=$(java --version)}[1,3]}"
The pieces:
RPROMPT= - variable for the right-side prompt.
"..." - the critical part. Variables in double quotes will be expanded then and there, so the commands within this will only be executed when the project file is sourced.
${...[1,3]} - selects the first three words of the enclosed expression. On my system, java --version returns three lines of data, which is way too big for a prompt; this reduces it to something manageable.
${=...} - splits the enclosed value into words.
$(java --version) - jre version info.
I am trying to set the dictionary option (to allow autocompletion of certain words of my choosing) using wildcards in a filename glob, as follows:
:set dict+=$VIM/dict/dict*.lst
The hope is that, with this line in the initially sourced .vimrc (or, in my case of Windows 10, _vimrc), I can add different dictionary files to my $VIM/dict directory later, and each new invocation of Vim will use those dictionary files, without me needing to modify my .vimrc settings.
However, an error message says that there is no such file. When I give a specific filename (as in :set dict+=$VIM/dict/dict01.lst ), then it works.
The thing is, I could swear that this used to work. I had this setting in my .vimrc files since I started using Vim 7.1, and I don't recall any such error message until recently. Now it shows up on my Linux laptop as well as my Windows 7 and Windows 10 laptops. I can't remember exactly when this started happening.
Yes, I tried using backslashes (as in :set dict+=$VIM\dict\dict*.lst ) in case it was a Windows compatibility issue, but that still doesn't work. (Also this is happening on my Linux laptop, too, and that doesn't use backslashes for filepaths.)
Am I going senile? Or is there some other mysterious force going on?
Assuming for now that it is a change in the latest version of Vim, is there some way to specify "use all the dictionary files that fit this glob"?
-- Edited 2021-02-14 06:17:07
I also checked to see if it was due to having more than one file that fits the wildcard glob. (I thought that if I had more than one file that fit the wildcard, the glob would turn into two filenames, equivalent to saying dict+=$VIM/dict/dict01.lst dict02.lst which would not be syntactically valid.) But it still did not working after removing extra files so that only one file fit my pathname of $VIM/dict/dict*.lst . (I had previously put another Addendum here happily explaining that this was how I solved my problem, but it turned out to be premature.)
You must expand wildcards before setting an option. Multiple file names must be separated by commas. For example,
let &dictionary = tr(expand("$VIM/dict/dict*.lst"), "\n", ",")
If adding a value to a non-empty option, don't forget to add comma too (let is more universal than set, so it's less forgiving):
let &dictionary .= "," . tr(expand(...)...)
I'm trying to run the code below from BioSeqClass package, however I get an error message:
Error in system(command, intern = TRUE) : '“C:\Program' not found
selectWeka(data, evaluator="CfsSubsetEval", search="BestFirst", n)
This is a problem with how BioSeqClass is calling java: it is leaving the file names unprotected/unquoted, and R's system and system2 commands are horrible by not forcing quoting. (If you ever think of using these commands directly yourself, I strongly recommend something like processx.)
One should create an issue or bug-report, but I don't know how to do that with Bioconductor, and their mirror on github (https://github.com/Bioconductor-mirror) is defunct, so I'm at a loss there. Hopefully somebody with more info can weigh in on this.
Workarounds
It is not obvious if the problem is due to where weka.jar is located or perhaps one of the other arguments. You can find out where the problem is by debugging the selectWeka function and inspecting the value of command before the system call. Look for the Program Files component of a path.
If the problem is with weka.jar, then this suggests that you are installing packages somewhere under C:\Program Files\, which is in my experience bad practice on two counts:
For many problems I cannot recall (but this one re-ignites the discussion), I never install R in the default location under C:\Program Files\...; instead, I install it under a new directory, C:\R\R-3.5.3 (version-based) and go from there. You may not have control over this if on a university/company system.
Since this is not in a base-R package, this suggests that either you are not using a personal library location (collection of packages), or have placed your personal library for some reason under C:\Program Files. If the former, I strongly suggest you never install new packages inside the base-R installation directory, instead using your own. See ?.libPaths and many other tutorials/discussions on the topic online. Using packrat or checkpoint might also mitigate this problem.
If the problem is with trainFile (if I'm reading the source correctly), then your "permanent" fix is to change where Windows puts temporary files, as this trainFile is a temporary file created specifically for this function-run. If this is your problem, I'll leave it up to you to fix.
Regardless, you may not have time or need to make a more permanent solution, you just want to run this once or twice and then move on. For that fix:
Again, debug(selectWeka), and once command is defined (the next command to be executed is tmp <- system(command,intern = TRUE)), run this code to fix the value of command:
if(search=="Ranker"){
command = paste("java -cp ", shQuote(file.path(.path.package("BioSeqClass"), "scripts", "weka.jar")),
" weka.attributeSelection.", evaluator, " -i ", shQuote(trainFile),
" -s \"weka.attributeSelection.", search, " -N ", n, "\"", sep="" )
}else{
command = paste("java -cp ", shQuote(file.path(.path.package("BioSeqClass"), "scripts", "weka.jar")),
" weka.attributeSelection.", evaluator, " -i ", shQuote(trainFile),
" -s weka.attributeSelection.", search, sep="" )
}
(For the record, all I changed was adding shQuote twice to each paste.) Confirm that command now has quotes around things, something like
Browse[2]> command
[1] "java -cp \"C:\\Program Files\\...\\weka.jar" weka.attributeSel... -i \"c:\\path\\to\\some\\tempfile\" ...
Then you can continue-out of the debugger and let it run its course.
(I hope you don't have hundreds of calls to selectWeka.)
Caveat: I am not a use of BioSeqClass, so I'm saying all of this from speculation and inference. I might have mis-located the source of the error. And since I don't know what I'm doing with it, I have not tested the modified command assignment within selectWeka. I believe shQuote(...) is the right way to go, but you might need to use sQuote or dQuote instead, I'm not sure how your system is setup.
I recently installed Zsh in hope of a better life and brighter mornings. However, I quickly realized Zsh introduces various issues in conjunction with tmux.
The first issue was some weird stuff happening at the end of the prompt, before my commands, but this was resolved by supplying tmux with the -u flag for unicode-support. However, I am stuck with one final issue that needs resolution before I can use Zsh with tmux:
Usecase: Autocomplete a command which contains multiple suggestions
Issue: Autocompletion shifts suggestion one character to the right, while leaving the original character behind (visual bug, it is not included in the command)
Example 1.
Then I hit TAB..
Example 2.
Then I hit TAB..
Note 1: This does NOT occur when using the Bash-shell.
Note 2: I am using "oh-my-zsh". This issue only occurs when using the provided themes. This narrows it down to an "oh-my-zsh"-theme issue, not native zsh/tmux.
In case some people still get a similar issue, see also the top-voted answer of Remnant characters when tab completing with ZSH. The plugin you were using may have had the same non-printable characters issue, that would explain the symptoms.