I'm trying the following example in RStudio

// [[Rcpp::depends(RcppProgress)]]
#include <progress.hpp>
// [[Rcpp::export]]
double long_computation3(int nb, bool display_progress=true) {
    double sum = 0;
    Progress p(nb*nb, display_progress);
    for (int i = 0; i < nb; ++i) {
        if (Progress::check_abort() )
            return -1.0;
        for (int j = 0; j < nb; ++j) {
            p.increment(); // update progress
        sum += R::dlnorm(i+j, 0.0, 1.0, 0);
    }
    }
    return sum + nb;
}

> Rcpp::sourceCpp('rcpp_progress.cpp')
> system.time(s  <- long_computation3(3000))
|----|----|----|----|----|----|----|----|----|----|
**************************************************|
|
   user  system elapsed 
   0.96    0.00    0.97 
> system.time(s  <- long_computation3(1000))
|----|----|----|----|----|----|----|----|----|----|
**************************************************|
|
   user  system elapsed 
   0.13    0.00    0.11 
> system.time(s  <- long_computation3(10000))
|----|----|----|----|----|----|----|----|----|----|
**************************************************|
********|
   user  system elapsed 
  10.75    0.00   10.79 
> system.time(s  <- long_computation3(10000))
|----|----|----|----|----|----|----|----|----|----|
**************************************************|
********|
   user  system elapsed 
  10.89    0.00   10.93 

For long time computation, extra * are always produced.

I was surprised to find that RcppProgress was removed from CRAN:

https://cran.r-project.org/web/packages/RcppProgress/index.html

Is there a reason for this, and will it be back?

e.g. RcppProgress:::test_sequential

Is there a way to have a Progress class member? That would allow me to define it only only place, rather than many times. Here is what I tried (this runs as a cpp file in Rstudio)

#include <Rcpp.h>
// [[Rcpp::plugins(cpp11)]]
// [[Rcpp::depends(RcppProgress)]]
#include <progress.hpp>
#include <progress_bar.hpp>

class Simple {
public:
  double sum;
  // I would like to do
  // Progress progressbar;
  // instead I tried
  Progress *progressbar;
  void start(int nb, bool display_progress){
    sum = 0;
    // I would like to do
    // progressbar->reset(nb*nb, display_progress);
    // or "InterruptableProgressMonitor._reset" ?
    
    // this compiles but does not work
    Progress p(nb*nb, display_progress);
    progressbar = &p;
  };
  
  bool do_something(int i, int j) {
    //the below causes a crash:
    //progressbar->increment();
    //if (Progress::check_abort() ) return false;
    sum += R::dlnorm(i+j, 0.0, 1.0, 0);
    return true;
  };
  void stop(){
    //?;
  };
  double compute(int nb, bool display_progress);   
};
  
  
double Simple::compute(int nb, bool display_progress) {
  start(nb*nb, display_progress);
  for (int i = 0; i < nb; ++i) {
    for (int j = 0; j < nb; ++j) {
      if (!do_something(i, j)) return -1; 
    }
  }
  stop();
  return sum + nb;
}

// [[Rcpp::export]]
double computecpp(int nb, bool display_progress) {
  Simple c;
  return c.compute(nb, display_progress);
}

/*** R
computecpp(300, TRUE)
*/

@nathan-russell and I are tracking an issue in Rcpp down, which appears to be linked to RcppProgress.h.

You seem to be doing

using namespace Rcpp;

here meaning that everybody using RcppProgress now has an open and unprotected Rcpp namespace.

That is probably not good design. Could we do without that?

I wrote a short gist based on the idea in this package that should work for RcppParallel and any other multithreading framework.
https://gist.github.com/traversc/8b26b9e689b23d3a174aee296c0503ca

I think this general approach would be nice to see in this package.

The changes you'd need to make are not too many. In summary:

• counter -> std::atomic
• increment -> counter.fetch_add(1)
• Check for main thread -> std::this_thread::get_id()

The testthat unit tests use the devtools::load_all function in order to make the contents of packages in inst/examples available to the testing suite. However, devtools itself depends on curl, which on unix systems requires a separate installation of libcurl. This is fine if you have root access, but otherwise no curl, so no devtools, so no RcppProgress.

I've looked at your unit tests and I believe they can be rewritten without the use of devtools::load_all:

1. Most tests can be combined into single cpp files, which in turn can be compiled on-the-fly with Rcpp::sourceCpp.
2. The test in inst/examples/RcppProgressExample checks for multiple definitions in two cpp's. Since Rcpp::sourceCpp is not intended for compiling multi-file projects, I suggest to put these test files in src and thus have them compiled with RcppProgress itself.

If you're OK with the suggestion I'm happy to get started on a pull request.

Is it possible to implement a prediction of the time a loop monitored by rcpp_progress will take to finish? Maybe a simple extrapolation of the time needed so far?

I would love to have that feature!

Sometimes I have to kill a function which includes an RcppProgress bar, and then when I run it again, I get

ERROR: there is already an InterruptableProgressMonitor instance defined

It would be nice for there to be a convenience R function such as RcppProgress::clear_pbs() that would remove any InterruptableProgressBarMonitor (or similar objects) associated with the current session, so one could continue without restarting.

I'm seeing warnings from R-devel with my package that R:

Found ‘abort’, possibly from ‘abort’ (C)

in two objects. The two objects don't call stop or abort or any of that, but they do use RcppProgress, which has a function called abort(). Is it possible the new warning is related to RcppProgress?

Thanks.

I added

// [[Rcpp::depends(RcppProgress)]]
#include <progress.hpp>
#include <progress_bar.hpp>

to my code, and now I warnings during compilation:

C:/---------/R/win-library/3.4/RcppProgress/include/interruptable_progress_monitor.hpp:101:0: warning: ignoring #pragma omp critical [-Wunknown-pragmas]
 #pragma omp critical
 ^
C:/---------/R/win-library/3.4/RcppProgress/include/interruptable_progress_monitor.hpp:143:0: warning: ignoring #pragma omp atomic [-Wunknown-pragmas]
 #pragma omp atomic
 ^
C:/---------/R/win-library/3.4/RcppProgress/include/interruptable_progress_monitor.hpp:149:0: warning: ignoring #pragma omp critical [-Wunknown-pragmas]
 #pragma omp critical

http://gallery.rcpp.org/articles/using-rcppprogress/ says that RcppProgress is compatible with multithreaded code, is there reason to believe it should work with RcppParallel? I realize this may turn out to be a question for RcppParallel, but thought I would ask here.

I had some success on Windows with the following skeleton:

struct MyWorker : public RcppParallel::Worker {
  Progress progress;
  tthread::mutex mutex;

  MyWorker(Progress& progress) : progress(progress) {  }

  void operator()(std::size_t begin, std::size_t end) {
    for (std::size_t i = begin; i < end; i++) {
      // real work here
      {
        tthread::lock_guard<tthread::mutex> guard(mutex);
        progress.increment();
        if (Progress::check_abort()) {
          return;
        }
      }
    } 
  }
};

but on Linux (admittedly only tried under a VM), I get messages along the lines of Error: C stack usage 23365264492 is too close to the limit if I increase the number of threads above 1, even with the simple parallel matrix square root example from: http://gallery.rcpp.org/articles/parallel-matrix-transform/

Is this approach doomed to failure?

Thank you for providing this useful package. We have been using it to develop a new package (see here) and so far it is working very well.

I have a feature request of sorts. I'm wondering if it would be possible to provide an example class (inheriting from ProgressBar) that mimics the display of the default progress bar provided by the progress package. e.g.,

[======================================================>----------------]  77%

Or perhaps this already exists somewhere?

I suspect I am doing something silly, but I installed RcppProgress on R 4.2.1. and tried to run the test

library("RcppProgress")
RcppProgress:::test_sequential(max, nb, display_progress)
Error: object 'test_sequential' not found

Similarly,
test_sequential <- function(max=100, nb=1000, display_progress=TRUE) {
.Call("test_sequential_wrapper2", max, nb, display_progress, PACKAGE = "RcppProgressExample")
invisible()
}

did not work.

It seems I have failed in installing the package, using install.packages("RcppProgress").

When I look in

/Library/Frameworks/R.framework/Versions/4.2/Resources/library/Rcpp/include

the header files, for example, are missing.

I am trying to put a progress bar into some cpp code. Thank you for any help you can give me.

Elizabeth