Add symlinks to C++ source and a default .Rproj (grf-labs#440)

Instead of copying files over, symlinks in `grf/src` now point to the
Rcpp bindings (`grf/bindings`) and core C++ (`core/src`). This allows
RStudio's Build menu (accompanying .Rproj file) to work, allowing for
quicker conditional compilation when working on both R and C++ changes.

.gitignore

@@ -1 +1,2 @@
 *.DS_Store
+.Rproj.user

DEVELOPING.md

@@ -2,9 +2,31 @@
 
 In addition to providing out-of-the-box forests for quantile regression and instrumental variables, grf provides a framework for creating forests tailored to new statistical tasks. Certain components around splitting and prediction can be swapped out, within the general infrastructure for growing and predicting on trees.
 
-### Working with the code
+## Contributing
 
-The core forest implementation is written in C++, with an R interface powered by Rcpp. We recommend using a full-powered C++ IDE such as CLion, Xcode, or Visual Studio when working with the core code. To build the R package from source, cd into `r-package` and run the `build_package.R` script. Code style consistency is checked with the [lintr](https://github.com/jimhester/lintr) package, and in RStudio you can make sure the code is consistent with the R standard by clicking _Addins->Style_.
+This repository follows the standard open source protocol and setup with git where there is an abundance of existing resources to get up to speed (see for example the
+contributing guidelines for well known packages in other languages, like Scikit-learn, Scipy, and pandas)
+
+Condensed greatly, the workflow is to fork this repository, check out a branch, commit your changes (forming an ideally legible commit history),
+then submitting a pull request explaining your contribution, ideally referring to the issue you created, or the issue you chose to work on.
+
+## Working with the code
+
+The core forest implementation is written in C++, with an R interface powered by Rcpp. We recommend using a full-powered C++ IDE such as CLion, Xcode, or Visual Studio when working with the core code.
+
+## R package
+
+To build the R package from source, cd into `r-package` and run `build_package.R`. Required development dependencies are listed there
+(note: it is recommended to install the latest [lintr](https://github.com/jimhester/lintr) package with `devtools::install_github("jimhester/lintr")`). This mimics the tests run when submitting a pull request.
+
+An alternative development workflow is to use the accompanying grf.Rproj and build and test the package with RStudio's build menu, which can be convenient
+for quickly iterating C++/R code changes.
+
+### Note for Windows users:
+
+Symlinks in the src directory point to the core C++ and R bindings. On Windows one has to clone this repository with symlinks enabled: `git clone -c core.symlinks=true https://github.com/grf-labs/grf.git` (this command needs to be run as an administrator: right click _Command Prompt -> Run as administrator_). Caveat: the above RStudio workflow is not tested on Windows.
+
+## Core C++
 
 ### Code structure
 

r-package/.gitignore

@@ -23,8 +23,8 @@
 # clion files
 .idea/
 
-# copied src files
-grf/src/*
+# Generated files
+grf/src/RcppExports.cpp
 
 # documentation by roxygen
 *.Rd

r-package/build_package.R

@@ -14,7 +14,6 @@ library(roxygen2)
 library(lintr)
 
 package.name <- "grf"
-package.src <- "grf/src"
 
 # Check code style consistency
 linters <- with_defaults(
@@ -36,21 +35,13 @@ if (!is.na(args[1]) && args[1] == "--as-cran") {
   write_union("grf/.Rbuildignore", "^tests/testthat/test_((?!cran).).*")
 }
 
-# Copy Rcpp bindings and C++ source into the package src directory. Note that we
-# don't copy in third_party/Eigen, because for the R package build we provide
-# access to the library through RcppEigen.
-unlink(package.src, recursive = TRUE)
-dir.create(package.src)
-
-binding.files <- list.files("grf/bindings", full.names = TRUE)
-file.copy(binding.files, package.src, recursive = FALSE)
-file.copy("../core/src", package.src, recursive = TRUE)
-file.copy("../core/third_party/optional", package.src, recursive = TRUE)
-
 # Auto-generate documentation files
 roxygen2::roxygenise(package.name)
 
 # Run Rcpp and build the package.
+# Symlinks in `grf/src` point to the Rcpp bindings (`grf/bindings`) and core C++ (`core/src`).
+# Note: we don't link in third_party/Eigen, because for the R package build we provide
+# access to the library through RcppEigen.
 compileAttributes(package.name)
 clean_dll(package.name)
 build(package.name)

r-package/grf/grf.Rproj

@@ -0,0 +1,21 @@
+Version: 1.0
+
+RestoreWorkspace: Default
+SaveWorkspace: Default
+AlwaysSaveHistory: Default
+
+EnableCodeIndexing: Yes
+UseSpacesForTab: Yes
+NumSpacesForTab: 2
+Encoding: UTF-8
+
+RnwWeave: Sweave
+LaTeX: pdfLaTeX
+
+AutoAppendNewline: Yes
+StripTrailingWhitespace: Yes
+
+BuildType: Package
+PackageUseDevtools: Yes
+PackageInstallArgs: --no-multiarch --with-keep.source
+PackageRoxygenize: rd,collate,namespace

r-package/grf/src/AnalysisToolsBindings.cpp

@@ -0,0 +1 @@
+../bindings/AnalysisToolsBindings.cpp

r-package/grf/src/CausalForestBindings.cpp

@@ -0,0 +1 @@
+../bindings/CausalForestBindings.cpp

r-package/grf/src/CustomForestBindings.cpp

@@ -0,0 +1 @@
+../bindings/CustomForestBindings.cpp

r-package/grf/src/InstrumentalForestBindings.cpp

@@ -0,0 +1 @@
+../bindings/InstrumentalForestBindings.cpp

r-package/grf/src/Makevars

@@ -0,0 +1 @@
+../bindings/Makevars

r-package/grf/src/QuantileForestBindings.cpp

@@ -0,0 +1 @@
+../bindings/QuantileForestBindings.cpp

r-package/grf/src/RcppUtilities.cpp

@@ -0,0 +1 @@
+../bindings/RcppUtilities.cpp

r-package/grf/src/RcppUtilities.h

@@ -0,0 +1 @@
+../bindings/RcppUtilities.h

r-package/grf/src/RegressionForestBindings.cpp

@@ -0,0 +1 @@
+../bindings/RegressionForestBindings.cpp

r-package/grf/src/optional

@@ -0,0 +1 @@
+../../../core/third_party/optional/

r-package/grf/src/src

@@ -0,0 +1 @@
+../../../core/src/