MiniMIP

MiniMIP is an open source, machine learning oriented Mixed-Integer Programming (MIP) solver. We provide a range of interfaces for all aspects of solving MIPs (e.g. heuristics, cut generators, LP solvers), supplying users with a constant view of the internal state and allowing them to propose modifications that are integrated into the global state internally.

This project is still under heavy development and is not ready for practical application.

Build system

Bazel

The bazel build environment composes of:

• The file WORKSPACE that tells from where to fetch all the dependencies
• The file .bazelrc that defines commonly used options while building (e.g., in MiniMIP we always compile with a flag --cxxopt=-std=c++2a)
• The build rules for all targets, which are stated explicitly in each (sub-)directory's BUILD.bazel file.

If you don't have Bazel installed already, you can download it from https://bazel.build/.

To build a single target in minimip/examples run:

bazel build //minimip/examples:example

To build all targets in minimip/examples run:

bazel build //minimip/examples:all

Bazel puts all artifacts generated during compilation in the top directory in directories that start with bazel-. The most useful is bazel-bin -- this is where the executables are kept (including the executables for unit tests).

Since you have built your binaries already, you can now run the test:

bazel-bin/minimip/examples/example_test

Or the main executable:

bazel-bin/minimip/examples/example_main --logtostderr

The --logtostderr flag tells the logging system to output logs to stderr (by default is streamed to a file).

You could achieve the same thing by running from bazel directly:

bazel run //minimip/examples:example_main -- -logtostderr

The -- -logtostderr is needed to pass the run-time argument to example_main. All command-line arguments after "--" are passed to the launched binary.

Similarly, you can also run the tests directly from bazel:

bazel test //minimip/examples:example_test

Or, if you have many tests to run (we don't have here, so the effect will be identical as above):

bazel test //minimip/examples:all

Bazel by default builds in "fast-build" mode. To maximize binary performance (e.g., for benchmarking purposes) compile in "opt" mode:

bazel build -c opt //minimip/example:example_main

To include symbols allowing for debugging (with gdb) compile in dbg mode:

bazel build -c dbg //minimip/example:example_main

See the bazel doc for more details.

CMake

We also provide a CMakeLists.txt file for building the project with CMake. To build the project with CMake, you can run the following commands:

cmake -S. -Bbuild
cmake --build build -v -j 12
(cd build && ctest)

Dependencies

See WORKSPACE or CMakeLists.txt file for a list of dependencies and extra information. These dependencies are automatically pulled by bazel.

Important Boost still needs to be installed on the machine for SoPlex to read LP files and be tested. On Debian machines, the package libboost-all-dev covers it.

Style guide

Google C++ style guide is available at: https://google.github.io/styleguide/cppguide.html

To automatically format the syntax you can run: clang-format -i PATH_TO_FILES/*.h PATH_TO_FILES/*.cc which will pick up the style specified in .clang-format.

Be careful not run clang-format on BUILD files. To format bazel's BUILD files automatically, run buildifier (see https://github.com/bazelbuild/buildtools).

VSCode setup

There is an official VSCode plugin for bazel called vscode-bazel. While it allows context highlights when editing Bazel files (BUILD, WORKSPACE, etc.) it doesn't configure IntelliSense for code files compiled using Bazel. Instead, use the following steps to get code suggestions using clangd. Tested on Linux Mint 20 (Ubuntu 20.04 derivative) on 2022-10-04.

1. Make sure clangd is installed on your system

sudo apt install clangd

2. Download the Bazel Compilation Database tool.

3. Go to MiniMIP's root directory (where the WORKSPACE file is located) and run the generate.py script you just downloaded. When the script has completed (may take several minutes), you should see a compile_commands.json file.

4. Install the vscode-clangd extension for VSCode.

   Note: The clangd extension is incompatible with Microsoft's IntelliSense, and will ask you to disable it. If you want to control this manually, the setting is found under C/C++: Intelli Sense Engine.

5. In your VSCode settings, find Clangd: Arguments and add --compile-commands-dir=${workspaceFolder}/ and make sure Clangd: Check Updates is enabled.

6. Restart VSCode.

Your MiniMIP files should now give you contextual highlighting, symbol lookup, clang-tidy warnings and more. You may need to occasionally repeat step 3 above if the information in compile_commands.json becomes stale, e.g. if you add a new target or introduce new dependencies.