Intro

In neutron scattering, we shoot neutrons (tiny elementary particles) onto a substance and want to know if they change direction. Therefore, a common type of detector is, like a computer screen, divided into many pixels. When it is placed behind the substance, it can measure when and where a neutron hits one of the pixels.

This task simulates performing data acquisition: receiving data from a detector over the network, with a format given by the manufacturer, parsing it and making a quick visualization of the data.

In addition to this specification, the repository contains two .data files that represent data streams from the detector, and for each one a .png image that shows the desired final result.

Your task

• Write a very simple server in C that accepts a TCP connection on port 50000 and simply streams the contents of a dump file to a client.

  It does not need to handle multiple clients, no threading is required.

• Write a client in Python that connects to the server, receives the data and parses it according to the format documented below.

  It should create a histogram of the data using numpy (see below for an explanation) and after reaching the end of the data, display the histogram as an image using matplotlib.

All code must run on a Linux host running e.g. Ubuntu 20.04. It should handle unexpected errors (like incomplete packets or a broken TCP connection) reasonably. Don't use any libraries except for the C and Python standard libraries, numpy and matplotlib. The code does not need to be particularly optimized for speed.

⚠️ The finished code should be sent in a few days before the interview. During the interview, you should be able to show the code using screen sharing in an editor/IDE, and run it. ⚠️

Data format specification

The file content is a sequence of packets. Each packet contains a header and optionally a number of events.

The header is 16 bytes long and has the following structure:

+---------+---------+-------------------+
| length  | pktype  | timestamp         |
| 32 bits | 32 bits | 64 bits           |
+---------+---------+-------------------+

All integers are in little-endian format:

• The length gives the length of the whole packet in bytes.

• If pktype is anything other than 0x1000, the data does not contain useful events and must be discarded.

• The timestamp can be ignored for your task.

If pktype is 0x1000, the rest of the packet contains a number of events, each of which is a 32-bit little-endian integer:

MSB                               LSB
+---------+---------+---------------+
| X coord | Y coord | timestamp     |
| 7 bits  | 7 bits  | 18 bits       |
+---------+---------+---------------+

X and Y are the pixel coordinates used for the histogram, the timestamp can again be ignored.

X/Y values and histogramming

The client should create an X-Y histogram of the events, which means the following:

• At the start, create a new 2-dimensional numpy array filled with zeros and the dimensions 128x128 (enough for 7-bit coordinates).

• Then, when you process each event, add 1 to the array element at the event's X and Y coordinate.

• In the end, if you display the array as an image using matplotlib, you should get something that looks like the .png file in this repository.