change build system to bazel (#78)

* change build system to bazel

Using bazel allows for using travis-docker infrastructure without
sudo privileges.

* adapt README file for the new changes

* remove sudo

.gitignore

@@ -48,3 +48,21 @@ CppNumericalSolvers.config
 *.mexw32
 *.mexmaci64
 *.mexglx
+
+# Ignore backup files.
+*~
+# Ignore Vim swap files.
+.*.swp
+# Ignore files generated by IDEs.
+/.classpath
+/.factorypath
+/.idea/
+/.project
+/.settings
+/WORKSPACE.user.bzl
+/bazel.iml
+# Ignore all bazel-* symlinks. There is no full list since this can change
+# based on the name of the directory bazel is cloned into.
+/bazel-*
+# Ignore outputs generated during Bazel bootstrapping.
+/output/

.travis.yml

@@ -1,65 +1,35 @@
-sudo: required
-dist: precise
-language: cpp
+dist: trusty
+
+addons:
+  apt:
+    sources:
+      - ubuntu-toolchain-r-test
+    packages:
+      - wget
+      - pkg-config
+      - g++
+      - g++-4.9
+      - g++-5
+      - pkg-config
+      - zip
+      - zlib1g-dev
+      - unzip
 
-matrix:
-  include:
-    - compiler: gcc
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-4.9
-      env: COMPILER=g++-4.9
-    - compiler: gcc
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-5
-      env: COMPILER=g++-5
 
 before_install:
-  - sudo apt-get update -qq
-  - chmod u+x get_dependencies.sh
-  - ./get_dependencies.sh
-  - git submodule update --init --recursive
+  - wget https://github.com/bazelbuild/bazel/releases/download/0.4.4/bazel-0.4.4-installer-linux-x86_64.sh
+  - wget https://github.com/bazelbuild/bazel/releases/download/0.4.4/bazel-0.4.4-installer-linux-x86_64.sh.sha256
+  - sha256sum -c bazel-0.4.4-installer-linux-x86_64.sh.sha256
+  - chmod +x bazel-0.4.4-installer-linux-x86_64.sh
+  - ./bazel-0.4.4-installer-linux-x86_64.sh --user
+  - export PATH="$PATH:$HOME/bin"
 
-install:
-  ############################################################################
-  # Install a recent CMake
-  ############################################################################
-  - |
-    if [[ "${TRAVIS_OS_NAME}" == "linux" ]]; then
-      CMAKE_DIR="install_cmake"
-      CMAKE_URL="http://www.cmake.org/files/v3.3/cmake-3.3.2-Linux-x86_64.tar.gz"
-      mkdir $CMAKE_DIR && travis_retry wget --no-check-certificate --quiet -O - ${CMAKE_URL} | tar --strip-components=1 -xz -C $CMAKE_DIR
-      export PATH=${TRAVIS_BUILD_DIR}/${CMAKE_DIR}/bin:${PATH}
-    else
-      brew install cmake
-    fi
 
 script:
-  - mkdir build
-  - cd build
-  - cmake -DEIGEN3_INCLUDE_DIR=eigen -DCMAKE_CXX_COMPILER=$COMPILER -DCMAKE_BUILD_TYPE=Release ..
-  - make all -j
-  - make lint
-  - ./bin/verify
+  - bazel run verify
+  - python lint.py
 
-addons:
-  apt:
-    sources:
-    - ubuntu-toolchain-r-test
-    packages:
-    - gcc-4.8
-    - g++-4.8
 
 notifications:
   email: false
 
-branches:
-  except:
-    -gtest

BUILD

@@ -0,0 +1,21 @@
+package(default_visibility = ["//visibility:public"])
+
+load("//:generator.bzl", "build_example", "build_test")
+
+build_example("linearregression")
+build_example("logisticregression")
+build_example("rosenbrock")
+build_example("rosenbrock_float")
+build_example("simple")
+build_example("simple_withoptions")
+build_example("neldermead")
+build_example("neldermead-customized")
+
+build_test("verify")
+
+
+py_binary(
+    name = "lint",
+    args = ["src/test/verify.cpp"],
+    srcs = ["lint.py"],
+)

README.md

@@ -6,22 +6,15 @@ CppOptimizationLibrary
 Get this Library
 -----------
 
-    git clone --recursive https://github.com/PatWie/CppNumericalSolvers.git
-
-or a complete copy and paste version
 ````bash
-    git clone --recursive https://github.com/PatWie/CppNumericalSolvers.git
+    git clone https://github.com/PatWie/CppNumericalSolvers.git
     cd CppNumericalSolvers/
-    ./get_dependencies.sh
-    mkdir build
-    cd build
-    cmake ..
-    make all
-    ./bin/verify
+    bazel run verify
 ````
+
 About
 -----------
-Have you ever googled for a c++ version of *fminsearch*, which is easy to use without adding tons of dependencies and without editing many setting-structs? This project exactly addresses this issue by providing a *header-only* library without dependencies. All solvers are written from scratch, which means they do not represent the current state-of-the-art implementation with all tricky optimizations (at least for now). But they are very easy to use. Want a full example?
+Have you ever looked for a c++ version of MATLAB's *fminsearch*, which is easy to use without adding tons of dependencies and without editing many setting-structs? This project exactly addresses this issue by providing a *header-only* library without dependencies. All solvers are written from scratch, which means they do not represent the current state-of-the-art implementation with all tricky optimizations (at least for now). But they are very easy to use. Want a full example?
 
 ````cpp
     class Rosenbrock : public Problem<double> {
@@ -62,24 +55,9 @@ For checking your gradient this library uses high-order central difference. Stud
 Install
 -----------
 
-Before compiling you need to adjust one path to the [Eigen3][eigen3]-Library (header only), which can by downloaded by a single bashscript, if you haven't Eigen already.
-````bash
-    # download Eigen
-    ./get_dependencies.sh
-````
-To compile the examples and the unit test just do
-````bash
-    # create a new directory
-    mkdir build && cd build   
-    cmake ..
-    # build tests and demo  
-    make all    
-    # run all tests                
-    ./bin/verify  
-    # run an example
-    ./bin/examples/linearregression    
-````
-For using the MATLAB-binding open Matlab and run `make.m` inside the MATLAB folder once.
+Note, this library is header-only, so you just need to add `include/cppoptlib` to your project without compiling anything and without adding further dependencies. We ship some examples for demonstration purposes and use [bazel](https://bazel.build/) to compile these examples and unittests. The latest commit using CMake is da314c6581d076e0dbadacdd263aefe4d06a2397.
+
+When using the MATLAB-binding, you need to compile the mex-file. Hereby, open Matlab and run `make.m` inside the MATLAB folder once.
 
 Extensive Introduction
 -----------
@@ -192,7 +170,8 @@ Currently, not all solvers are equally good at all objective functions. The file
 
 # Contribute
 
-Make sure that `make lint` does not display any errors and check if travis is happy. Do not forget to `chmod +x lint.py`. It would be great, if some of the Forks-Owner are willing to make pull-request.
+Make sure that `python lint.py` does not display any errors and check if travis is happy. It would be great, if some of the Forks-Owner are willing to make pull-request.
 
 [eigen3]: http://eigen.tuxfamily.org/
+[bazel]: https://bazel.build/
 [matlab]: http://www.mathworks.de/products/matlab/

WORKSPACE

@@ -0,0 +1,15 @@
+new_http_archive(
+  name = "eigen_archive",
+  url = "https://bitbucket.org/eigen/eigen/get/3.2.10.tar.gz",
+  build_file = "eigen.BUILD",
+  strip_prefix = "eigen-eigen-b9cd8366d4e8",
+)
+
+
+new_http_archive(
+    name = "gtest",
+    url = "https://github.com/google/googletest/archive/release-1.7.0.zip",
+    sha256 = "b58cb7547a28b2c718d1e38aee18a3659c9e3ff52440297e965f5edffe34b6d0",
+    build_file = "gtest.BUILD",
+    strip_prefix = "googletest-release-1.7.0",
+)

bazel.rc

@@ -0,0 +1,19 @@
+# This is from Bazel's former travis setup, to avoid blowing up the RAM usage.
+startup --host_jvm_args=-Xmx2500m
+startup --host_jvm_args=-Xms2500m
+startup --batch
+test --ram_utilization_factor=10
+
+# This is so we understand failures better
+build --verbose_failures
+
+# This is so we don't use sandboxed execution. Sandboxed execution
+# runs stuff in a container, and since Travis already runs its script
+# in a container (unless you require sudo in your .travis.yml) this
+# fails to run tests.
+build --spawn_strategy=standalone --genrule_strategy=standalone
+test --test_strategy=standalone
+
+# Below this line, .travis.yml will cat the default bazelrc.
+# This is needed so Bazel starts with the base workspace in its
+# package path.