Programming

cmake offers the option to generate a compilation database in json format using the option CMAKE_EXPORT_COMPILE_COMMANDS=YES. The database holds essentially the command, paths and compilers flags used to compile each file of a project. The generated file (usually called compile_commands.json) can be used in conjunction with various code completion tools like YouCompleteMe and Visual Studio’s IntelliSense in order to offer more complete information, completion and errors about the code.

There are many resources, posts and articles about using cmake and writing modern cmake code. Unfortunately for a new user it is sometimes difficult to locate the proper resource that contains no deprecated options or any other non-recommended practices. One of the quickest and simplest resource I became aware recently is the official tutorial that lives in the documentation of cmake.

ccache is a utility that caches intermediate files during compilation offering significant speedups. It is especially useful if you are switching between two branches of the same project see below.

To use it, prepend it before the compiler i.e. Use ccache gcc instead of gcc ... . If you are using cmake with a version greater that 3.4 you can use the option -DCMAKE_CXX_COMPILER_LAUNCHER=ccache for older version you can modify the compiler settings or use the variable RULE_LAUNCH_COMPILE for more see here.

As you print more and more numbers in the screen it may become cumbersome to count digits just to see is if you are in the range of millions or 100’s of millions. Visually grouping digits can simmplify things. There is a pleathora of ways to do this in C++ but the one I find the best is the one provided by the standand C++ library through the std::numpunct object. Here is a small example that uses groups digits using commas.

Many times when trying to catch a bug , or to validate that your parallel code runs correctly you wish to run your code again and again and again until break on an assertion or even better until you persuade yourself that your code is good enough. Once more, stakoverflow has the answer for us:

set $n = 100
while $n-- > 0
  printf "starting program\n"
  run
  if $_siginfo
    printf "Received signal %d, stopping\n", $_siginfo.si_signo
    loop_break
  else
    printf "program exited\n"
  end
end

If you use it often save it in a file and launch it using

If you use ParaView often you may want to automate some of your workflows. One way to do this is writing a macro using the Python interpreter that ships with ParaView. Here, is an example of extracting all elements from a tetrahedral mesh that have a minimum dihedral angle less that 5 degrees.

from paraview.simple import *
#### disable automatic camera reset on 'Show'
paraview.simple._DisableFirstRenderCameraReset()
# get selection
G = GetActiveSource()
# create quality Data
# set min dihedraql angle as criterion
M = MeshQuality(Input=G, TetQualityMeasure="Minimum Dihedral Angle")
# select it
SetActiveSource(M)
# Create selector
S = SelectionQuerySource(QueryString = "Quality <=5")
# Create Extractor
E = ExtractSelection(Input = M,Selection = S)
Show()
Render()

#### uncomment the following to render all views
# RenderAllViews()
# alternatively, if you want to write images, you can use SaveScreenshot(...).

Another advantage of this approach is that you can run it in a headless environment using the pvpython interpreter that ships with ParaView.

In an effort to contribute back to the great matplotlib-cpp library I came across the need of having to create a mechanism that can select between different types in C++. This was needed in order to allow passing data from C++ to python without any copies using the numpy library. The only caveat in this case in that numpy requires to know the type you are passing. Since the matplotlib-cpp library is template-based there is a nice compile-time way to do it using template specialization. In simple terms it boils down to creating a base case for all the types you do not support and specialize for the types you care. Using a simple struct to hold the type completes the whole construct. The final result is :

Having the -Wall flag enabled by default can potentially save lots of time and headaches when programming in C/C++. Same other usefull flags are :

• -Wduplicated-cond : Warn about duplicated condition in if-else-if chains.
• -Wduplicated-branches : Warn when an if-else has identical branches.
• -Wlogical-op : Warn about use of logical operations where a bitwise operation probably was intended.
• -Wnull-dereference: Warn when the compiler detects paths that dereferences a null pointer.
• -Wjump-misses-init: Warn if a goto statement or a switch statement jumps forward across the initialization of a variable, or jumps backward to a label after the variable has been initialized.

source