If you like Boost.Histogram, please star the project on Github! We want Boost.Histogram to be the best histogram library out there. If you give it a star, it becomes more visible and will gain more users. More users mean more user feedback to make the library even better.
Feel free to ask questions on https://gitter.im/boostorg/histogram.
We value your feedback about issues you encounter. The more information you provide the easier it is for developers to resolve the problem.
Issues should be reported to the issue tracker.
Issues can also be used to submit feature requests.
Don't be shy: if you are friendly, we are friendly!
Please report any tests failures to the issue tracker along with the test output and information on your system:
- platform (Linux, Windows, OSX, ...)
- compiler and version
Fork the main repository. Base your changes on the develop branch. Make a new branch for any feature or bug-fix in your fork. Start developing. You can start a pull request when you feel the change is ready for review.
Please rebase your branch to the original develop branch before submitting (which may have diverged from your fork in the meantime).
You can build and test Boost Histogram with cmake, but this is experimental. It is very convenient, however, as it does not require setting up the Boost superproject, you just need a checkout of Boost Histogram, but when you need to recompile many times, it is slower than using b2 (see next section). The following steps compile the tests with cmake
mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=Debug ..
cmake --build . --target check -j3Increase -j3 to the number of jobs that you want to run in parallel.
For general advice on how to set up the Boost project for development, see the Getting Started section on https://github.com/boostorg/wiki/wiki.
Boost comes with a build system called b2, which is the most efficient way to run the develop-test cycle. It takes a few extra steps and some reading to set up, but the payoff is worth it in the long run. After following the general advice from the Boost wiki, you should be all set up to run the tests from the Boost Histogram project folder with the command
../../b2 cxxstd=latest warnings-as-errors=on testYou can also test the examples by executing
../../b2 cxxstd=latest examplesTo make the tests complete faster, you can use the option -j4 (or another number) to build in parallel.
The output of b2 is verbose and there is no highlighting, so sometimes it is difficult to spot a problem. You can give b2filt are try, which you can install with pip install b2filt. Just replace ../../b2 with b2filt, for example:
b2filt cxxstd=latest warnings-as-errors=on testNote: b2filt is not an official part of Boost and has never been tested extensively on Windows. Please leave an issue on GitHub if you have trouble with b2filt.
Boost.Histogram maintains 100% line coverage. Coverage is automatically checked by CI. To generate a report locally, you need to build the code with gcc-8 or gcc-12 (these version are known to work, others might) and coverage instrumentation enabled, do b2 toolset=gcc-8 cxxstd=latest coverage=on test.
To generate the coverage report, run tools/cov.py from the project root directory of Boost.Histogram. This script has not been tested on Windows, so probably will not work on Windows. If you have several versions of gcc installed, you may need to specify the gcov version to use, run for example GCOV=gcov-8 tools/cov.py.
This will generate a new folder coverage-report with a HTML report. Open coverage-report/index.html in a browser.
Notes: Generating coverage data is very fickle. You need to use gcc-8 or gcc-12 and a matching version of gcov. Other gcc versions were known to be broken or were not supported by lcov at the time of this writing, which is used to process the raw coverage data. Generating coverage data with clang and llvm-cov is not supported by lcov at the ime of this writing.
To build the documentation, you need to install a few extra things, see Quickbook documentation.
Follow the Boost Library Requirements and Guidelines and the established style in Boost.Histogram.
All names are written with small letters and _. Template parameters are capitalized and in camel-case. Do not worry about whitespace, which is automatically formattet with clang-format.
To automatically format whitespace in your commits, use pre-commit. You can install it with pip install pre-commit. Then do pre-commit install in the project directory. From now on, pre-commit will run clang-format automatically whenever you run git commit. If it complains, simply add the changes it made to the files and run git commit again.
Doxygen comments should be added for all user-facing functions and methods. Implementation details are not documented (everything in the boost::histogram::detail namespace is an implementation detail that can change at any time).