Developer tools for the NIMBLE project

First install this package:

devtools::install_github('nimble-dev/nimble-devtools')

Next restart RStudio and verify that you see menu items Tools > Addins > Hot Reload Code and Tools > Addins > Open Genrated C++ Files.

To use RStudio addins, you may need to upgrade to a more recent RStudio.

RStudio Addin: Reload NIMBLE code without reinstalling NIMBLE

If you modify an internal NIMBLE function in-place and want to reload it quickly (without a lengthy R CMD INSTALL process), you can use our Hot Reload addin for RStudio. (This is just a smart wrapper around assignInNamespace.)

1. (optional) Set a keyboard shortcut for the hot reload addin using Tools > Addins > Browse Addins.... For example I use CTRL-R.

2. In an RStudio editor window, position the cursor on the first line of the definition name <- function(...){...} of your modified NIMBLE code.

3. Either click on the Tools > Addins > Hot Reload Code menu item, or use your new keyboard shortcut.

RStudio Addin: Open the latest generated C++ files in tempdir()

1. (optional) Set a keyboard shortcut for the Open Generated C++ addin using Tools > Addins > Browse Addins.... For example I use CTRL-G.

2. Generate some C++ files with compileNimble()

3. Either click on the Tools > Addins > Open Generated C++ Files menu item, or use your new keyboard shortcut.

Speed up package installation

These steps attempt to speed up R CMD build nimble && R CMD INSTALL nimble_*.tar.gz:

1. (Savings: 50%) On Linux or OS X, use ccache. Following Dirk Eddelbuettel's advice, you can enable ccache globally on your system by creating symbolic links to ccache for each of the compilers you use:

   cd /usr/local/bin
   sudo ln -s /usr/bin/ccache gcc
   sudo ln -s /usr/bin/ccache g++
   sudo ln -s /usr/bin/ccache clang
   sudo ln -s /usr/bin/ccache clang++
   

2. (Savings: 25%) Add tons of --no-<expensive operation> flags to your R CMD INSTALL invocation. Fritz wrapped this in a script nim-install.sh in this repo. TODO add a .bat file or something for Windows users.

3. Further research is needed. Fritz started to profile package installation with profile_install_nimble.R

Run a single test file

Perry suggested the following test wrapper to run tests in a single file:

library(nimble)
library(testthat)
source(system.file(file.path('tests', 'test_utils.R'), package = 'nimble'))
if (0) options(error = recover) else options(error = NULL)  # Toggle 0/1 to enable/disable recover().

# This tests the file 'tests/test-optim.R'.
test_package('nimble', 'optim', reporter = 'tap')  # tap has prettier error reporting.

Quickly open the latest NIMBLE-generated C++ files

To quickly open the latest generated C++ files

1. Install clang-format so that the files can be automatically formatted. For example

   brew install clang-format          # OS X
   sudo apt-get install clang-format  # linux
   

2. Set your EDITOR environment variable to your favorite editor (vim, emacs, sublime, etc.)

3. Configure nimble to compile to the directory ~/tmp

   compileNimble(myFunction, dirName = '~/tmp')
   

4. Run nim-show.py. I've added this to my PATH so I can type nim-sh<TAB><ENTER> anywhere to quickly see the latest

Navigate Source Code

If you're an Emacs or Vim user, you can make use of the rtags() function to generate tags files to more quickly navigate source code. Perry recommends the command R CMD rtags. This lets you jump-to-defintion of a symbol.