Tools¶
This section provides some introduction to using third-party tools or libraries on which Code_TYMPAN relies. Those dependencies are introduced, with installation instructions in the section Installation. In this section we give some entry-points regarding use of the tools and libraries.
We also explain some not so common C++ idoms used in some portion of the code.
Accessing the source-code¶
The source code of the Code_TYMPAN project is versionned using the Git Version Control System. On GNU/Linux the simplest way to install is through your distribution package manager (sudo apt-get install git). The software is also available on Windows and a good alternative is to use the Tortoise application which integrates well with the Explorer (also available on GNU/Linux). Git_Extensions and SourceTree are interesting software allowing navigation and visualization of the repository as well as the visualization of the corresponding trees.
The following named branches are currently used:
- master
The default branch name in Git is master. As you start making commits, you’re given a master branch that points to the last commit you made.
- develop
Is the current branch used to keep in sync the feature branches. Indeed, it is critical to avoid any significant divergence regarding the build system for example. Moreover some changes (many bug fixes, documentation improvement) can be put in common without interfering.
- release/x.x.x
is the branch for the production version x.x.x Last version : 4.2.1
- issue_xxx
This type of branch allows the developer to work on different issues. Each branch corresponds to an issue which can be a bug fix, a function addition, an optimization etc…
Once the developer has completed his development and validated, he delivers on the branch before creating a merge request which will be reviewed by a second person. Once this review is done, the branch can be merged into develop.
- Example of operation:
Alice is working on issue 14 which corresponds to a bug reported by a user.
She places itself on the develop branch and does an update of it.
She creates the Issue_14 branch from develop and works on it locally.
Once satisfied with its development and this one having been validated by the TU and possibly manual tests. She decides to deliver her modifications and therefore the new branch.
Once the pipeline is validated, Alice creates a merge request corresponding to her issue and her branch and requests a review from TOM.
Tom does the code review and eventually retrieves the branch locally for testing (if necessary)
Tom can possibly provide feedback to Alice to improve the work. But once this is ok, Tom validates the merge request.
The Issue_14 branch can then be merged into develop and the corresponding issue can be closed (after demo to the rest of team)
- Other existing branch :
hotfix_branch : used for hotfixes, mainly on new releases
fix_ci_appveyor : used for pipeline maintenance and contains modification for pipeline debug
Of course other feature branches can be added by need.
CMake¶
CMake is an open-source build system. It is used to control the software compilation process using a simple platform and compiler independent configuration files. CMake generates native makefiles and workspaces that can be used in the compiler environment of your choice. On GNU/Linux, CMake generates Makefiles and on Windows, it can generate Visual Studio solutions.
Using CMake¶
First, create a directory where you’ll put the building of Code_TYMPAN.
Command line
Do:
cmake path/to/Code_TYMPAN
or:
ccmake path/to/Code_TYMPAN
cmake-gui path/to/Code_TYMPAN
to launch the ncurses or Qt graphical interfaces respectively. The command line is very helpful to get some info about a command, a variable, etc.:
# List of CMake commands
cmake --help-command-list
# List of CMake variables
cmake --help-variable-list
# Get doc for 'add_library' command
cmake --help-command add_library
# Get doc for 'CMAKE_BUILD_TYPE' variable
cmake --help-variable CMAKE_BUILD_TYPE
Note
A CMake ‘command’ is like a function. Documentation about CMake commands gives you indication about their different arguments etc.
GUI
You can use either ccmake
or cmake-gui
. Note that ccmake
is only
available on GNU/Linux. On Windows, you should use cmake-gui
(see the page
Running CMake on Windows).
Documentation
Getting help on CMake thanks to --help
command lines as above. Moreover,
there are other resources:
The reference book Mastering CMake
The online documentation with the cmake tutorial, a Wiki and a FAQ.
Add a library to the project¶
Suppose you create a new directory with C++ headers and compilation units. The
best way is to create a CMakeLists.txt
file in this current directory and write:
# The name of your new module (i.e. library)
set(MODULE_NAME new_tympan_module_name)
# Group all source files in a CMake list variable.
file(GLOB ${MODULE_NAME}_SRCS *.cpp *.h)
# Create the library as ('library_name' 'list of source files')
add_library(${MODULE_NAME} ${${MODULE_NAME}_SRCS})
You can further refer to your module with the variable
MODULE_NAME
.
You can also use (if needed):
# If your module depends on other Tympan modules (i.e. other Tympan libraries).
add_dependencies(${MODULE_NAME} tympan_common tympan_lib)
# If you have to link your library with one or a few libraries such as Boost,
# Qt or OpenGL.
target_link_libraries(${MODULE_NAME} ${QT_LIBRARIES})
# If you would like to add a specific directory to the include path. A
# 'include_dir' is missing when you have a compilation error such as 'header.h
# no such file or directory'.
include_directory(<path/to/dir>)
Best Practices¶
Do not build a project in the source directory.
Do not add a manual path to look for a dependency library or to find a source file : instead edit the said path in the GUI (or code reliable auto-detection in case you are a seasoned CMakeLists’ developer).
CMake language is not case-sensitive for identifiers, but is for some optional arguments names within commands.
VARIABLE_NAME
orvariable_name
are the same variable butFILE(EXISTS ...)
will be ok whereasFILE(exists ...)
will not. Please try to follow the existed convention in reading someCmakeLists.txt
.Create a
CMakeLists.txt
file for each directory. Useadd_subdirectory
command to execute the CMake file in this directory.How to clean the CMake cache: remove the
CMakeCache.txt
file.Display the value of a variable — useful for debugging:
message (STATUS "Your message: " ${VARIABLE})
ccmake .
orcmake-gui .
in the building directory to display and edit all CMake variables of the current building project.cmake --build <building_dir>
to launch compilation, whatever the environment. Very useful when you would like to compile on Windows without launching Visual Studio.
Documentation¶
The C++ source code is annotated and API documentation can be
extracted with Doxygen. The main documentation is generated from
reStructuredText sources in the doc
directory thanks to
Sphinx.
Build the documentation¶
On Windows¶
Go to the root source documentation directory and:
make.bat html
See the results in the
_build/html/
directory and opening the fileindex.html
with your favorite Web browser.
Check the external links of the documentation with:
make.bat linkcheck
On Linux¶
Go to the root source documentation directory and:
make html
See the results in the
_build/html/
directory and opening the fileindex.html
with your favorite Web browser.
Check the external links of the documentation with:
make linkcheck
Note
The tools used to display maths equations properly is MathJax, a JavaScript library which prettily renders equations written in LaTeX.
Useful directives¶
Take a look at the Sphinx and reStructuredText websites.
.. note::
write a note... code-block:: c++
piece of code with the C++ syntax color.:ref:`section_name`
cross reference with a different location such as a section of an other file... todo::
add a todo task.. doxygenclass:: <class_name>
show the Doxygen documentation of the class<class_name>
... doxygenfunction:: <function_name>
show the Doxygen documentation of the class<function_name>
.Note
See other Doxygen directives in the official Doxygen website.
:math:`x^2 -2x +1=0`
or.. math::
to write LaTeX equations.Warning
A Web connection is needed to be able to see nice equations.
Formatting and Linting¶
We use formatters to automatically ensure consistency across the code base and thus reduce the amount of stylistic choices the developers must make. Linters are used to detect easily discoverable code smells. The following tools should be used to ensure that a minimal level of quality is preserved. On Gitlab, new commits trigger a pipeline which runs all these tools and fails when any of them reports an error. When creating a Merge Request, make sure that all these checks pass.
Linting python files with Flake8¶
Flake8 is a python linter which both checks for pep8 compliance and find code smells. If it is not already installed you can install it using pip:
$ pip install flake8
Assuming that flake8
is in your PATH
(which should be the case if you
installed it in your active virtualenv), it can be run from command line:
$ flake8 <some_python_file.py>
<some_python_file.py>:2:1: F401 'this' imported but unused
Formatting python files with black¶
black is used to format python files. It can be installed using pip:
$ pip install black
When calling black with a file or directory, all the passed files are formatted:
$ black config_gui.py
reformatted config_gui.py
All done!
1 file reformatted.
Linting cython files with cython-lint¶
cython-lint is a cython linter which both checks for pep8 compliance and find code smells. If it is not already installed you can install it using pip:: It is an equivalent to the flake8 library for cython files.
$ pip install cython-lint
Assuming that cython-lint
is in your PATH
(which should be the case if you
installed it in your active virtualenv), it can be run from command line:
$ cython-lint --max-line-length 120 <some_cython_file.pyx> <some_cython_file.pxd
<some_cython_file.pyx>:2:1: F401 'this' imported but unused
Formatting C++ files with clang-format¶
For C++ source and header files, the formatting is done using clang-format. On Debian based linux distributions it can be installed with apt:
$ sudo apt install clang-format-15
On Windows clang-format is distributed as part of LLVM. You can download the installer here: https://github.com/llvm/llvm-project/releases
The current version to install is 15.0.6.
When running installation, make sure that one of Add LLVM to the system PATH for all
users
and Add LLVM to the system PATH for current user
is checked.
Once the installation is finished, Check that it is correctly installed by
running clang-format --version
. It should output something like
clang-format version 15.0.6
.
Warning
Some formatting rules need a version of clang-format greater than 11. On
some linux distributions, the clang-format
command is bound to an old
version of clang-format (version 7 on debian buster). You have to ensure
that your editor can find a version greater than or equal to 11 of clang-format.
You can easily check that the whole C++ code base is correctly formatted by
running the script tools/check_format.py
.
Experimenting clang-tidy¶
clang-tidy can be used to enforce coding rules like C++ core guidelines https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines.
It is included in the LLVM distribution installed before.
A first experimentation is done on Code_TYMPAN by implementing the rule
cppcoreguidelines-init-variables
.
The .clang-tidy
configuration file is used at the root of the source tree.
In order to run clang-tidy, we must generate a compilation database following this guide : https://clang.llvm.org/docs/HowToSetupToolingForLLVM.html#setup-clang-tooling-using-cmake-on-windows
To run clang-tidy, we can use cmder and run this command :
find ..\code_tympan\Tympan\core -iname "*.hpp" -o -iname "*.h" -o -iname "*.cpp" -type f -print0 | xargs -0 clang-tidy -p C:\projects\code_tympan_build_ninja_d\compile_commands.json --fix
Launch linters from cmake¶
You can launch all the linters with cmake:
$ cmake --build . --target lint
Integration with IDEs¶
Usage with Vim¶
In Vim, all these linters and formatters can be automatically used by installing ALE. Assuming you are using version 7 or greater of Vim, setup is made as follow (on linux or using git-bash):
$ mkdir -p ~/.vim/pack/lint/start
$ git clone --depth 1 https://github.com/dense-analysis/ale.git ~/.vim/pack/lint/start/ale
Lines with warnings should now be highlighted. The warnings are displayed at
the bottom of the screen when the cursor is on these lines. black and
clang-format can be activated by adding the following lines to your
~/.vimrc
file:
let g:ale_fixers = {
\ 'python': ['black'],
\ 'cpp': ['clang-format'],
\}
" Set this variable to 1 to fix files when you save them.
let g:ale_fix_on_save = 1
Running clang-format on Visual Studio¶
clang-format is already available in Visual Studio. When opening the project,
the clang configuration file should be automatically detected and Visual Studio
should ask you if you want to enable clang-format support.
If that’s not the case you can enable it by checking Tools->Options->Text
Editor->C/C++->Formatting->General->Enable ClangFormat support
.
For the moment being, Visual Studio Community 2019 integrates the version 12.0.0
of clang-format.
You must specify the version of clang-format you wish to use (15.y.z) by selecting
clang-format exe in Tools->Options->Text
Editor->C/C++->Code Style->Formatting->General->Use custom clang-format exe file
.
You should now be able to format your code using the Ctrl-K, Ctrl-D
keyboard shortcut. Install the extension named Format on Save
if you want
to run clang-format on save.
Running flake8 in PyCharm¶
Flake8 is not natively supported by PyCharm. You can add it as an external
tool. Open Settings
(Ctrl-Alt-S) and go to Tools->External Tools
. Then
click the +
button and fill in the form this way:
You can then run this external tool from Tools->linters->Flake8
.
Running black in PyCharm¶
Black can be configured the same way as flake8 as an external tool:
The BlackConnect plugin uses a black server to avoid restarting a python
interpreter each time the files are formatted. You can also configure it so
that black run automatically when files are saved. You can install it from
Settings->Plugins
and configure it from Settings->Tools->BlackConnect
.
You will have to install blackd (pip install black[d]
) and run blackd
in order for this plugin to work.
C++ idioms¶
C++ is a very complex language, full of subtleties and pitfalls… but some good practices, nice (template) libraries and well identified idioms can improve development efficiency and code base maintainability.
In this section, we highlight some of such idioms and tools which are used (or to be used) in Code_TYMPAN. Anyhow we strongly recommend C++ developers to have a look at some the following books:
- Practical and motivated idioms
Scott Meyers, Addison-Wesley, Effective C++ and Effective STL
- Complete introduction of the language
Bjarne Stroustrup, Addison-Wesley, C++, the language
Some powerful techniques are exposed to the developer through relatively simple interfaces by Boost or the STL, such as the standard smart pointers or the ranges concept and the foreach loops for example.
Note
Further resources
Here are some useful inline resources regarding C++
Introductory online C++ course OpenClassRoomsCpp with corresponding printed book, in french
cplusplus.com tutorial cplusplus_tutorial or reference cplusplus_reference
Google hands-on, practical online course https://developers.google.com/edu/c++/
Smart Pointers¶
- https://www.dreamincode.net/forums/topic/86846-tr1-intro-smart-pointers/
A very good, nice and accessible, introduction to smart pointers and modern C++ memory management.
- https://www.boost.org/doc/libs/1_53_0/libs/smart_ptr/smart_ptr.htm
Official documentation for Boost smart pointers, among which
shared_ptr
andweak_ptr
have been included in the TR1 and C++’11.
BOOST_FOREACH¶
Before C++’11 the language lacked a foreach loop construct to iterate other data structure in a convenient and readable way. Boost tries to workaround this limitation with the BOOST_FOREACH macro (which is some kind of black magic we should get rid of as soon as C++’11 is widespread and fully supported).
For example, say c is some standard STL container of type C<T>
(i.e. with a begin()
and end()
method providing iterators to
elements of type T
) then you can write:
BOOST_FOREACH( T& e, c )
{
// do something with the element e
}
instead of the usual and horribly verbose when C
and/or T
are
complex:
for( C<T>::iterator it = c.begin() ;
it != c.end() ; ++it )
{
T& e = *it;
// do something with the element e
}
- https://www.boost.org/doc/libs/1_53_0/doc/html/foreach.html
Official documentation for Boost foreach loops.
The range concept¶
To be right to BOOST_FOREACH
, this construct is far more powerful
than just iterating in a simple way through a container. It relies
on the Range concept.
The Range concept is a generalisation of the input sequences introduced by Stroustrup in C++, the language 18.3.1 : basically a range is something that can be seen as a pair of begin/end iterators.
The whole point of ranges resides in the fact that they do not have to be STL container, or any data-structure, they can pretty-well represent a simple counter as well as an on-fly processing over an other range ! For example :
/* ... */
struct double_int
{
typedef int result_type;
int operator()(int x) const { return x * 2; }
};
int main(int argc, const char* argv[])
{
std::vector<int> input; // An STL container is a range
/* Fill `input` with 1, 2, 3, ... */
BOOST_FOREACH( int i, input | transformed(double_int()) )
{
cout << i << ", "; // Will display 2, 4, 6, ...
}
return 0;
}
- https://www.boost.org/doc/libs/1_70_0/libs/range/doc/html/index.html
Official documentation for Boost range concept which goes along so well with foreach loops.
Beware : Ranges are a very powerful but quite advanced - and thus pitfall prone - C++ technique. Please resort to it only if you know what you are doing, otherwise just transforming the data by-hand and storing the result in some container is likely to be a far easier and cleaner way…