Plotting with C++

C++ is an extremely useful language, especially when it comes to working with hardware. One of the limitations with using C++ is it's inflexibility when it comes to plotting data. This repository explores different ways to plot with C++.

This repository uses the below three different apporaches, and each may be useful in different scenarios.

1. Matplotlib wrapper
2. gtkmm
3. Local server-client connection

Matplotlib Wrapper

This method uses the matplotlib-cpp written by Benno Evers. This library basically gives you a Python like functions to be used directly inside C++ code.

Dependencies

• Matplotlib Wrapper

  sudo apt-get install python-tk python-numpy python2.7-dev python-dev3
  pip install matplotlib numpy
  

  If you already don't have one, create a file named matplotlibrc in ~/.matplotlib and add this line to the file.

  backend: TkAgg
  

Running the Code

• Compiling the code

  cd 01_matplotlib-wrapper 
  mkdir build && cd build
  cmake ../
  make
  

• Running the program:
  • Draw a line plot with static data: ./plot_line
  • Draw a plot with changing data: ./animate

gtkmm

This method uses gtkmm-plplot by Tom Schoonjans. This library uses PLplot, and brings it's functions to gtkmm. This has been successfully tested in both Ubuntu 18.04 and macOS Mojave.

Dependencies

• Linux

  sudo apt-get install libgtkmm-3.0-dev
  sudo apt-get install libltdl3-dev libqhull-dev libpango1.0-0 libplplot-dev
  

• MacOS

  brew install gtkmm gtkmm3 glade plplot
  brew install tschoonj/tap/gtkmm-plplot
  

Running the Code

• Compiling the code

  cd 02_gtkmm 
  mkdir build && cd build
  cmake ../
  make
  

• Running the program:
  • Hello world (just an application screen with a button): ./hello_world
  • A copy of a PLplot example which includes some interactive plots: ./hello_world_plot
  • A plot example with minimum requirements: ./minimal_working_example
  • A plot which utlizes multithreading: ./updating_plot

Known Issues

• At least in macOS Mojave, the main GUI window needs to called in the main thread. So, using multithreading requires the threads to not to be joined. Users need to stick to other measures to make sure that the program does not prematurely end once a single thread exits.
• The app crashes randomly when the mouse is being moved over the drawing canvas area.

Quick Troubleshooting

• OSX:
  • Package 'libffi', required by 'gobject-2.0', not found: export PKG_CONFIG_PATH="/usr/local/opt/libffi/lib/pkgconfig"
  • Package 'icu4c' not found: export PKG_CONFIG_PATH="/usr/local/opt/icu4c/lib/pkgconfig

Local Server-Client Connection

All the methods discussed so far use different libraries developed in C++. In contrast, this method uses a server-client communication between C++ code and Python code so that all the plots are drawn using Python's Matplotlib. This does not suffer from issues related to multithreading as discussed above, and also can be used for plotting real-time data while talking to hardware.

This method has the potential to explore all the Matplotlib functions even after the updates, without waiting for some third party developers (most of whome develop them as a service to others in spite of working on real jobs) to update their libraries.

This method works as follows:

• C++ code: This is the server connection. C++ code does some calculations and send those data which requires poltting to the Python code.
• Python code: This is the client connection. Python code receives the data from the C++ code and plots them.

Both of these codes utilize multithreading so that the data transmission and plotting work independently.

C++ code

cd 03_server-client 
mkdir build && cd build
cmake ../
make
./server

NOTE: Server (C++ code) must be run first.

Python code

Below instructions assume that you have installed Anaconda (or atleast Python with numpy and matplotlib) in your system.

cd 03_server-client/scripts
python client.py

NOTE: Client (Python code) must be run while the server (C++ code) is running.