I am new to area of writing functions for using in R using Rcpp and Armadillo. I was looking trough the Tutorials and other questions which asked about this topic, but I was not able to find something suitable because most of them were out dated and did not work for me.
I would like to ask for any help with setting up the Eclipse for using Rcpp on Windows 10.
Best wishes.
The R Project documents all build and tests steps via commands starting with R with for example
R CMD build
R CMD COMPILE
R CMD SHLIB
and so on. Eclipse, as far as I know, prefers cmake so you need to write a CMakeLists.txt to do what you need. You will have to duplicate file location searches and all those features R provides for you. I suspect this to be a bit of work.
I did receive a contribution to RInside many moons ago so if you in its inst/examples/ directory:
edd#rob:~/git/rinside(master)$ find . -name CMakeLists.txt
./inst/examples/eigen/cmake/CMakeLists.txt
./inst/examples/wt/cmake/CMakeLists.txt
./inst/examples/qt/cmake/CMakeLists.txt
./inst/examples/mpi/cmake/CMakeLists.txt
./inst/examples/armadillo/cmake/CMakeLists.txt
./inst/examples/standard/cmake/CMakeLists.txt
edd#rob:~/git/rinside(master)$
I am not sure if these are still current and working but you could start there.
Related
I am a bit fresh to Julia (and stack overflow), so it might be a dummy question. I am using VS code in ubuntu 18.04.
Every time I start Julia REPL from VS code (Ctrl+shift+P, then "Julia: Start REPL"), the Julia REPL starts with a strange command "source /home/$user_name/bin/activate" and then an error (of course source is not defined in Julia) :
ERROR: UndefVarError: source not defined
Stacktrace:
[1] top-level scope
# REPL[1]:1
This problem was shown a couple of weeks ago but it does not affect my code at least for now. I have tried to search the problem but haven't found any similar questions. Any hints on what's going on? Many thanks!
Edit:
I think I might have used Conda somewhere.
Do "which Julia", I got:
/home/$user_name/julia-1.7.2/bin//julia
There is no "/home/$user_name/.julia/config/startup.jl" file, all the directories under "/home/$user_name/.julia/" are as follow:
artifacts compiled dev logs prefs scratchspaces
clones conda environments packages registries
Then I found a "startup.jl" file under "$JULIA_INSTALL_FOLDER\etc\julia\startup.jl", with following content (seems like not helping on my problem).
# This file should contain site-specific commands to be executed on Julia startup;
# Users may store their own personal commands in `~/.julia/config/startup.jl`.
Solved by following #SundarR's suggestion.
The problem was caused by the VSCode-Python extension installed.
Please see the comment above for the details and thanks again!
This was originally a Github Issue in the Dart-Code repository.
1. Context
I've been working on a package that has hundreds of tests, so an easy way of visualizing code coverage would be incredibly handy.
I would like to run my tests with, say, a .vscode configuration with an lcov.info output which would automatically be recognized by VS Code and highlighted on the respective editors with either red or green.
2. What I've Already Tried
I've tried many different solutions in the past few days — months actually — but none of them worked as the ideal one described above:
flutter test --coverage --coverage-path=lcov.info does work to generate the necessary file, but it's clunky to have to visualize it through a 3rd party program such as genhtml, all the more if you're on Windows.
And it does need Flutter in the end, which should not be necessary if you're working on pure Dart...
IntelliJ would supposedly work ideally, but I just can't seem to enable the Run with Coverage button on mine, even after installing the test_coverage package.
Though one person on Gitter told me he has it working on his IntelliJ.
Both the coverage and the test_coverage packages offer something close to what I described above, but their solutions are way clunkier — and on Windows they are tough to set up...
codecov.io is an alternative with a 3rd party, but it's annoying to have to handle this externally when the editor offers a much more flexible and faster experience.
And there is also the problem of ambiguous coverage, which is not clear with respect to codecov.io. For example, if one folder tests stuff that indirectly calls another folder, does that count as coverage for the indirectly called folder as well? That's almost always undesirable.
3. Other Resources
There's this old question on StackOverflow that was helpful initially.
You can take the genhtml.perl script here.
If you have Git for Windows installed on your machine, you already have Perl installed, it should be here: <git-install-dir>\usr\bin\perl.exe
Replace backslash characters (\\) with slash characters (/) in all file path lines (prefixed with SF:) in the lcov.info file.
Run genhtml.perl script. For example — assumed current working directory is root directory of your project —:
<git-install-dir>\usr\bin\perl.exe \
C:\Scripts\genhtml.perl \
-o .\coverage\html .\coverage\lcov.info
Note. It may be useful also to add the --prefix option.
As a result of these actions, you should get generated HTML report in the .\coverage\html/ directory. Open .\coverage\html\index.html file in your browser to see the report.
I hope this helps — at least, it worked for me.
This is a long post, sorry.
I have been trying to start a project, using Rust, but ran into a problem: it just does not work correctly on Linux in VSCode/Atom.
Wasted 3 days, searching online, trying different tutorials/videos - nothing worked + most of the material is from 2017. I have tried Matrix chat, but no one knew what to do. Git too has no solution as people keep suggesting very random things, like "change this variable in toml file to something else, and back again"
Git: https://github.com/rust-lang/rls-vscode/issues/513
I installed (and re-installed rust many times in the last 3 days), and it works just fine from the terminal, but not in the Editor.
Two Issues:
Editors don't see any crates, so you can't run your code from the editor.
Autocomplete does not work (only works on std, not on extra crates you add).
What I did (out of many other things):
install Rust (on Manjaro and Debian computers): curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
Tried stable, beta, nightly (for racer)
Just did a clean install again:
stable installed - rustc 1.40.0 (73528e339 2019-12-16)
Rust is installed now. Great!
Installed Rust (rls) rust-lang.rust extension in VSCode (and rust in Atom)
Created a new project: cargo new test_proj and added new rand = "0.6" to [dependencies] and used cargo build. It did build rand
Added "rust-client.enableMultiProjectSetup": true to the settings file to avoid a warning: A Cargo.toml file must be at the root of the workspace in order to support all features. Alternatively set rust-client.enableMultiProjectSetup=true in settings. by Rust (rls)
I also install 'code runner' extension that I use with Python, C++, and Java, to run the code from within the editor.
So now I have just the main func and it runs just fine from the editor:
Now I add rand and it seems to work, and does SOME auto-completion...
But now it stops working:
OK, I will finish the code and try running it, And now we can't run it anymore as the crate is missing:
OK, let's try Ctrl + Shift + B and try cargo build:
For some odd reason, it is looking in /media/Work/Work/rust_code and not in /media/Work/Work/rust_code/test_proj/
One last thing: let's try running in the terminal:
So it does work just fine.
Sorry for the long post, but I have wasted 3 days now and it still can't get it up and running. Did anyone manage to set this up at all?
It has to be Atom or VSCode as I have all of the other languages/projects setup there + VSCode is listed on the official Rust website, so I presume it should work.
Basically, out of two editors (VSCode and Atom), that have Debugging capabilities, unlike Intellij Rust, both don't work for me and I just can't code in Rust as tools are literally broken/not mature enough for productive work. Please let me know if I am wrong and it is just a case of one little flag, that everyone forgets to mention, is missing in some config.
Short Version
Assuming I understand the issue correctly. Then the various build issues boil down to different ways you're trying to build the project, and attempting to do so in different directories.
TL;DR: The directory you have open in VSCode is the rust_code directory. Close it and instead open rust_code/test_proj. Now the Rust: cargo build (or Rust: cargo run) should work.
Long Version
Editors don't see any crates, so you can't run your code from the editor.
The Rust: cargo build task isn't working, because the directory you have open isn't a "Rust project" (Cargo package), it's a directory containing another directory, which is a Rust project.
Looking at your screenshot shows this, look at how the top line says rust_code/test_proj:
When you execute the task, Cargo is complaining that rust_code/Cargo.toml doesn't exist, which is true since it's located in rust_code/test_proj/Cargo.toml
If you look at the output of this screenshot, you can see that is the case:
Here you're executing cargo run manually. But the important difference is that you're inside the rust_code/test_proj directory.
Lastly simply executing rustc main.rs is failing since you aren't passing the arguments needed. So rustc doesn't know anything about your dependencies.
Try executing cargo build -v then you can see all the arguments that Cargo is passing to rustc.
Autocomplete does not work (only works on std, not on extra crates you add).
Try and open a directory that contains a Cargo.toml, then code completion should work for dependencies. If I don't, then I get the following notification, and code completion only works for the standard library as you said. That being said, RLS is weird sometimes.
If I had to guess, then I think RLS compiles the code, and at some stage extracts the needed information. Thus if the code doesn't compile, then code completion could be affected. But this is 100% an educated guess.
Alternatively, I have heard praise about using IntelliJ + the Rust plugin, but I haven't used it myself yet.
I have figured out one part: problems with running your code from within the VSCode. I had to modify default code-runner command for rust:
Original command:
"rust": "cd $dir && rustc $fileName && $dir$fileNameWithoutExt",
Changed to:
"rust": "if [ $(basename $dir) = 'examples' ]; then cargo run --example $fileNameWithoutExt; else cargo run; fi",
And now it works, and I can run my code quickly.
Part 2: Autocomplete still is very very bad, unfortunately. I am using RLS. there is this thing racer. Is it a better choice?
Alternatively, I have heard praise about using IntelliJ + the Rust
plugin, but I haven't used it myself yet.
As I mentioned in my previous post, IntelliJ, free edition does not have a debugger from what I can see. I need it.
I am wondering how people get a good autocomplete from Rust?
I have started to try to use StatET and Texlipse with a view to producing SWEAVE reports. When starting the R console in the StatET plug-in for Eclipse (OS X 10.6.4), I get the message:
[INFO] The R package 'rj' is not available, R-StatET tools cannot be initialized.
Information on http://www.walware.de/goto/statet states:
"The package only works in the default R Console (RJ) in StatET. Therefore you have to install and update it using the command line":
R CMD INSTALL --no-test-load rj_*.tar.gz
What does this mean exactly?
In Eclipse under "Install new Software" and the site
WalWare - http://download.walware.de/eclipse-3.6
it states that
RJ Core Library 0.5.0.b201008271600sw
is already installed.
Any helpful guidance gratefully received.
[rant] I also lack the reputation to put my response where it belongs. And there's no button there indicating this. Just a big screen full of nowhere desired to click upon. Negative discoverability sucks. It sucks in Eclipse and it sucks here, too. [/rant]
I'm finding rj frustrating under Ubuntu. On my system, with each major Ubuntu upgrade with a new version of R, I have to go through this procedure over again.
One time I tried to use a tilde (~) character in a path name within some Eclipse configuration field. This failed silently.
My current configuration is as follows. In my ~/.Renviron file:
R_LIBS_USER=~/etc/R/amd64/2.11
Once rJava works, there is an rJava directory here. Just so I don't drive myself insane, in ~/etc/R/README I put a note that my current R configuration is controlled by ~/.Renviron
Somehow at one point I ended up with permission problems where
R CMD javareconf
was failing. Once in frustration I foolishly ran this under sudo to work around this problem, and ended up with permission problems on the generated files. Stupid! The correct fix was to make all the files in /etc/R world readable. Maybe javareconf is only important if you mess with your default Java. Previously, I was running the Sun's JRE. Since 10.10 I'm running icedtea6 and it seems to be working OK. One time it was just the thing to make my configuration work again.
Once I have the right Java binding, from command line R:
install.packages("rJava")
I've had file permission problems running this in the past. I think it was fixed by making all files in /etc/R world readable.
The following is a useful diagnostic to see if this worked, from within any R console:
> system.file("jri",package="rJava")
[1] "/home/allan/etc/R/amd64/2.11/rJava/jri"
If it doesn't look right under Eclipse, there is also some startup debugging available, though it took forever before I noticed this in the docs at StatET.
I created an R Console run configuration for R 2.11 debug In the JRE tab add -Dde.walware.rj.verbose=true. In the Common tab, click Allocate additional Error Log Consoles.
The main configuration options are as follows. Main tab, Launch Type: RJ. R_Config tab, you need to select a Configured R Installation, or create/edit one if you don't have one that works. Mine is Name=R, Location=/usr/lib64/R In the Environment tab I have R_LIBS_USER=/home/USER/etc/R/amd64/2.11 according to how I set things up above. Do not use the tilde (~) character here!
If you end up mucking about with this problem as much as I did, you end up killing a lot of dead consoles. In the Main tab I added the program option --no-save. I don't think this does anything, since I can't end my console by pressing CTRL-D. In the R_Console tab, I added the following R snippet:
q <- function(save = "no", status = 0, runLast = TRUE){
.Internal(quit(save, status, runLast))
#<environment: namespace:base>
}
This gets rid of the annoying "save workspace" prompt when killing an R console with the red square (is that symbolic?) My current workspace also includes a C project, so my R console constantly comes up underneath a blank C-Build console. Annoying as all hell! Haven't solved that yet, but I did discover that from within a StatET code windows the keybinding CTRL-R C pulls the R console to the top.
With my R 2.11 debug configuration I get an extra console titled /usr/lib/jvm/java-6-openjdk/bin/java. You'll have to spelunk the console drop-down to find it. The following is a pertinent snippet of a successful initialization:
CONFIG: JR library path:
/home/allan/etc/R/amd64/2.11
/usr/local/lib64/R/site-library
/usr/lib64/R/library
10-Dec-2010 7:56:47 AM org.rosuda.rj.JRClassLoader addClassPath
FINE: Added '/home/allan/etc/R/amd64/2.11/rJava/java' to classpath of URL loader
A lot more spoo, then finally:
10-Dec-2010 7:56:48 AM de.walware.rj.server.jriImpl.RosudaJRIServer start
INFO: R engine started successfully. New Client-State: 'Connected'.
Looks good, right? But lo and behold in the R console:
[INFO] The R package 'rj' is not available, R-StatET tools cannot be initialized.
Ignore the lying bastard! Turns out my rj is working just fine. No idea why this failure message persists.
Every time around the block with this problem some new amazing piece of gravel ends up in my sneaker.
I don't know whether I've covered everything, but I've written enough for now. Maybe if there are comments I'll come back and fill in the blanks.
Note: found an Eclipse plugin for console switching by keyboard here: Eclipse: Keyboard shortcut for switching between consoles?
It looks like you need to install the RJ library which can be downloaded from the StatET site here: http://download.walware.de/rj/rj_0.5.0-5.tar.gz (more recent versions of RJ can be found at http://www.walware.de/it/downloads/rj.mframe)
If you save that file on your desktop, you can then load it into R with the command they provided R CMD INSTALL --no-test-load rj_*.tar.gz after navigating to your desktop or wherever you saved the library file. Alternatively, you MAY be able to use the GUI interface and select Install packages from local ZIP files under Packages in R, though I haven't used R on a Mac so the layout may be slightly different.
After installing rj with in the R console using
install.packages(c("rj", "rj.gd"), repos="http://download.walware.de/rj-1.1")
(See http://www.walware.de/?page=/it/statet/troubleshooting.mframe for the correct rj version),
do not forget to update the R_LIBS variable in Eclipse (Windows -> Preferences -> StatET -> Run/Debug -> R Environments -> Edit). To find out the value just type in the R console:
packageDescription("rj").
In my windows 8 installation this returns C:/Users/Fibonacci.PYTHAGORAS/Documents/R/win-library/3.0/rj/Meta/package.rds. The value of R_LIBS I set to
C:/Users/Fibonacci.PYTHAGORAS/Documents/R/win-library/3.0
which is the path where all libraries are installed.
As stated in http://www.walware.de/?page=/it/statet/troubleshooting.mframe newer versions (Versions > 0.5.5) do not need the R CMD INSTALL etc... command.
Here the quite similar I had and the solution I found.
I originally installed Statet using the Eclipse built-in tool "Install new software ...."
I seems it downloaded Statet + rj and that the installation process went well.
After this first installation, I tried to install both packages again. Eclipse refused reading that they already are installed. And, both packages/programs were actually listed in the window of Eclipse.
But, the R console still refused to launch, the dialog box reading that I had to check whether rj was available and that the file paths were correct. Another strange thing was that window explorer had been unable to locate rj on my disk.
???
To solve my whole problem, I simply downloaded rj from http://www.walware.de/ and installed this package via the r console and using the basic install.packages() command.
It now works perfectly .
But, I still don't understand everything ..... :-)
I'd make a comment but I lack the necessary reputation. For Windows, the procedure is the same. There's no zip file. Use the tar.gz. Worked for me. The command, working from the same folder as the downloaded file:
$ G:\R-2.12.0\bin\i386\R.exe CMD INSTALL --no-test-load rj_*.tar.gz
The project is now hosted on GitLab, see
https://gitlab.com/walware/de.walware.rj-server.gr/-/wikis/Installation.
To install:
install.packages(c("rj", "rj.gd"), repos="https://download.walware.de/rj-4.0")
Note that binary packages are available for Windows.
As explained in this message from Stephan Wahlbrink on 2022-02-03 announcing StatET 4.5 release, this version works with R 3.6 to 4.1, and RJ-4.0.
For current information on StatET, check the home page and especially the news.
I have been used to working with VS2005 and 2008 - using msbuild, etc, but I have inherited a set of projects that have to remain in vc6.0 for now. I don't like opening each project in the developer studio and building. I prefer to build form command line (I am automating the builds). Is this possible?
I have tried the nmake utility, but I still need to open up the projects and save/export the make file. This is tedious if the project changes - each time I have to save the make file. nmake seems to work, but it had a problem when I changed the location of the project in my directory tree (I checked out of svn into a new clean dir to try the build). It seemed to have hard coded paths in it, but I will have to check on that - it might have been a different problem.
Any alternatives out there?
Eventually I will migrate these to 2008, but for now that is not an option.
I'm recording the most important part of your link here - Microsoft has a bad habit of moving stuff around and leaving dead links. Not only is this the most important bit, but it gives enough information to do a search if/when they move it.
Building a Project from the Command Line
You can build a Visual C++ project from the command line without first exporting a makefile (MAKEFILE, or filename.mak) and using the NMAKE utility.
The basic command syntax is
msdev FileName [/MAKE "ProjectName – ConfigName | ALL"] [/REBUILD /CLEAN /NORECURSE /OUT LogFile /USEENV]
where FileName is the name of your project (.dsp) or workspace (.dsw) file.
I think I found m answer here:
http://msdn.microsoft.com/en-us/library/aa699274.aspx
thanks all
of course you can automate. I haven't used vc in years, but I think the compiler is called c8.exe or wow, I can't remember silly little tidbits like that anymore, but look in your vc\bin directory at all the exes and it will be obvious by name.
you can write a batch file worst case. But I also remember the UI having a "create makefile" function. So you do that once, and then just run make from the command line and voila. or maybe it's nmake. Again, been a long long time.
Microsoft provides a command line driver for building Visual Studio projects. In VC6 it's called "msdev" (do msdev /? for a list of options).
At some point (probably VS.NET/VS 2002) they started calling the command line build driver "devenv" for some reason. It has a somewhat different syntax, but for driving builds the options are the same or similar.
We use automated builds at my work place. Essentially just a batch file i fire off from the command line. Let me make sure i am allowed to post some sample code before i go ahead and post it. But yes, it IS possible to automate the build.
Sample Code:
:::::::: CompileSolution :::::::::::::::::::::::::
call X:\BuildTools\bin\BuildVbProj.bat
%COMPONENTNAME% %SOLUTIONDIR%
%PROJFILE% %BUILDOUTPUTFILE%
%PREBUILDFILE% if %ERRORLEVEL% NEQ 0
goto BuildErrors
goto Cleanup
EDIT: The BuildVbProj.bat file ultimately calls VB6.exe in the Program Files\MS Visual Studio\VB98\ folder. Try calling it with "VB6.exe /?" or "VB6.exe -?" and it will show you a list of options. You can basically automate your process using those options.
There should be a similar exe for VC in the VC98 folder as well.
Another option which is less labor intensive is
Pulldown Menu (BUILD)
Select (BATCHBUILD)
Push Button (REBUILDALL)