Skip to content

Latest commit

 

History

History
59 lines (46 loc) · 3.6 KB

Readme.md

File metadata and controls

59 lines (46 loc) · 3.6 KB
Raw

Random Number Server in C++ with Python-Bokeh Gui Client

Here is my version of a Client-Server according to the requirements of the coding challenge:

  • Write a C++ program to generate a random number once per sec.
  • Each new random number will be transmitted via websocket to a Python script
  • The Python script uses the Bokeh Python GUI framework to create a very simple display (just a screen, with the number shown, continuously updating).

Environment setup/build

  1. I wrote this for Linux - so start there. I'm running Manjaro (Arch-based), so my versions of things might be annoyingly recent, but I bet it will work fine on any modern Linux with slightly older packages too.

  2. First, clone the repository with submodules (required for uWebSockets library and dependencies):

    mkdir <repo dir name>; cd <repo dir name>
git clone --recurse-submodules <this repo's url> .

  3. Python setup:

    • To be sure, use Python 3.12, and an updated pip. It should work down to Python version 3.10 (this version of Bokeh's limit), but I did not test other versions.

    • I recommend installing/setting up a virtual environment for python. If you're already using something like conda or pyenv, just use that. If not, you can use the built in venv before installing dependencies:

      cd <path/to/repo>
python3 -m venv .venv
source .venv/bin/activate

    • Install websockets and Bokeh libraries, and the dependencies those bring:

      pip install -r requirements.txt

  4. C++ setup/build:

    • Ensure you have a C and C++ compiler installed with modern C++ support (C++17 or newer). For the record I used gcc 14.2.1 and GNU make 4.4.1.
    • Run ./build.sh from the root of the repository (it's written for bash). This just calls gcc in a couple of one-liners, and outputs the executable "server" to the same directory.

Running

Here's how to run the example once you've followed the previous steps:

  1. Launch the C++ server: ./server
  2. Launch the Python GUI: python gui.py

Both C++ server and python scripts log the random number to stdout for easy verification they are in sync. The gui should launch automatically in a browser window.

Implementation Notes

I hard coded port number and localhost for both the C++ and python. No config file or input params.

It uses non-secure websocket connections (no TLS).

Python

It's a little hokey but gui.py is using a separate websocket connection with the C++ server to grab the raw random numbers. It runs a Bokeh server (with its own, independent websocket protocol that I didn't feel like implementing in C++) to display the number on a plot in a browser window.

C++

The random number is generated each second from whatever the system has available (std::random_device). I don't use a pseudo-random number generator because I figure the system RNG can handle 1 Hz (tho who knows if the system RNG isn't pseudorandom).

I send the number as a 32-bit unsigned int. Would've been simpler to limit this to uchar, but as it is I have the python script hard-coded to interpret the number as little-endian (which is how the C++ sends it on my machine).

I used the uWebSockets library because I was avoiding pulling in boost, and it seemed lightweight. I use its special timer mechanism instead of launching my own periodically firing thread... it seems to schedule things right. The design of this library feels like a consultation honeypot rather than easy to use. They go wild with templates and a builder pattern that make it hard to debug. IXWebSocket API looks easier to use without needing to know the internals.