G+Smo  23.12.0
Geometry + Simulation Modules
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Macros Groups Pages
Tutorial 04: Writing code

In this tutorial we collect some tips and guidelines for writing code in G+Smo.

Information regarding code formatting are found in the wiki.

Configuration options

Upon configuration, G+Smo presents a set of basic configuration options, which are ON (enabled):

Configuration:
  CMAKE_BUILD_TYPE        RelWithDebInfo
  GISMO_COEFF_TYPE        double
  GISMO_BUILD_LIB         ON
  GISMO_BUILD_EXAMPLES    ON
  CMAKE_INSTALL_PREFIX    /usr/local

The full list of options can be accessed using the cmake-gui utility, or using the command-line version:

ccmake .

inside the build folder.

  • CMAKE_BUILD_TYPE

    Available values are the standard CMake build configurations: Debug, Release, RelWithDebInfo?, MinSizeRel?.

  • GISMO_COEFF_TYPE

    The arithmetic type to be used for all computations. Available options include double, long double, float. This is the type is alias as real_t.

  • GISMO_WITH_XDEBUG

    If set to ON additional debugging tools are enabled during compilation. These include checked iterators for GCC and MSVC compilers and call stack back-trace printout when a runtime exception occurs.

  • GISMO_BUILD_LIB

    If enabled a dynamic library is created using GISMO_COEFF_TYPE arithmetic. A target for a static library named gismo_static is also created but not compiled by default.

  • GISMO_BUILD_EXAMPLES

    If enabled the programs in the examples folder are compiled, and executables are created in build-folder/bin.

  • GISMO_PLUGIN_AXL

    If enabled the plugin for Axel modeler is compiled (requires Axel).

  • GISMO_WITH_PSOLID

    If enabled the extensions using functionalities of Parasolid geometric kernel are compiled.(requires Parasolid).

  • gsOpennurbs

    Extension for reading and writing of Rhinoceros' 3DM.

  • CMAKE_INSTALL_PREFIX

    The location for installation of the library, e.g. /usr/local on some Linux systems.

Debugging

There are 2 main tools for debugging C++ code:

  • The debugger, for instance GDB for GCC, and the Visual Studio debuggerfor Visual C++.
  • The memory checker, for instance valgrind on linux or the C Run-Time Libraries (CRT) on Visual studio.

If the CMake option GISMO_WITH_XDEBUG is set to ON, more checks are performed while running the code. In particular the Standard library containers are checked for invalid access, invalid iterators, and so on. Also upon failure of the program the call stack is printed on the screen.

Note
The setting GISMO_WITH_XDEBUG replaces the standard containers, for instance std::vector with debug types, eg std::debug_vector. This can cause problems when linking the resulting dynamic library with third-party codes which is not compiled with the same settings. Setting the option back to OFF restores the correct types.

CMake provides with a tool to execute programs and report the result status. Result status zero (main() function returns 0) means program succeded, return 1 means program failed. To run this check inside the build folder do:

ctest -j3

Problems or bugs might also be related to the platform or operating system. Nightly compilations and tests on several systems detects such problems. The CDash server:

https://cdash-ci.irisa.fr/index.php?project=Gismo

collects the result of compilation and testing from compilation nodes. To Submit your compilation and test log to the CDash server you can type:

make -j2 Experimental

inside the build folder. After building and executing all programs, a log file will be posted on the CDash server, under the section Experimental.

Gismo and Cxx Standards

Gismo should be able to compile with C++98 and on different machines with different datatypes and maybe architectures. To ensure this, and still be able to use some of the benefits of newer compilers like C++11, there are some helping functions inside gsUtils and gsMemory that should be used instead of native C++11 calls. It is helpful to keep in mind that real_t and size_t are definitions that will be exchanged by cmake and should be used instead of double and int.

Cxx(11) Gismo
std::move(...) give(...)
double real_t
int index_t
std::unique_ptr memory::unique_ptr or gs...::uPtr
std::shared_ptr memory::shared_ptr or gs...::Ptr
std::to_string util::to_string

Also keep in mind that C++98 std::ios doesn't like "strings". Therefore, just call .c_str() if you give a string to an fstream (like ofstream::open).

And please don't use auto and the fancy new foreach loop.

Quick Tips

  • Find intuitive names for your classes, prefix the names with "gs"
  • Prefer one header file for one class, with the same name. For several small objects, grouping in one file is an option.
  • Avoid statements like "using namespace std;" in the code. Prefer to just type eg "std::vector". The reason is that we do not want to pollute the gismo namespace with third-party functions.
  • The header file "gismo.h" contains all the headers of the library. It is intended to be used by clients using the library (such as the example programs). It should not be used as an included file in the source code of the library itself.
  • Avoid overloading the header files with a lot of included files. Prefer to put a header file in the .hpp (or .cpp) file rather than the .h file.
  • While coding, do frequent updates to your svn/git copy. This will keep your local copy up to date with the repository.
  • Avoid letting your program produce paraview files by default. Use a –plot switch to trigger output.
  • Try to fit your work in the general framework of Gismo, and derive your classes from the existing Core classes. For instance, if you implement a different function type, inherit from gsFunction and implement the evaluation interface.
  • Add Doxygen comments to all of your functions and classes. There are a lot of resources online in using Doxygen documentation. The basic usage is quite straightforward.
  • Add your name on the header of your files. Use the common header as follows:
       /** @file fitting_example.cpp
    
        @brief Demonstrates adaptive fitting of data samples
    
        This file is part of the G+Smo library.
    
        This Source Code Form is subject to the terms of the Mozilla Public
        License, v. 2.0. If a copy of the MPL was not distributed with this
        file, You can obtain one at http://mozilla.org/MPL/2.0/.
    
        Author(s): G. Kiss, A. Mantzaflaris
       */
  • Get in touch with us on Github wiki: https://github.com/gismo/gismo/wiki and issues: https://github.com/gismo/gismo/issues