1e-exercises.md

Exercise: write a median function

Approaches (let's do at least 2)

• sort and pick
• quick-median (aka. lightweight sort&pick, as in quicksort)
• approximation-improving-median
  • this parallelizes well
  • at which point is this faster than quick-median?
  • did we lose any property?
    • what about median strings?

Homework 0

Make a package from the "trivial" median algorithm.

• Use julia's ]generate.
• Have a look at how unit tests are done
• Make sure this unittest works:

using MagicMedian

@testset "median funcitonality test" begin
	@test simplemedian([1,2,3,4]) >= 2
	@test simplemedian([1,2,3,4]) <= 3
	@test simplemedian([2,6,8]) == 6
end

• Zip the package (or .tar.gz it or whatever) and upload it to Moodle

How to run unit tests?

• place unittest code into test/runtests.jl in the package file
• run ]test MagicMedian

Exercise: let's play a maze game

• Read a maze from a file (let's have 0s for a corridor and 1s for a wall)
• Draw it to console using . and #, add a border.
• Make a function that annotates the whole maze based on how many (axis-aligned) steps a person (maze inhabitant) needs to take from some point
  • let's use the slow wave algorithm
• Use UnicodePlots to plot interesting stuff
  • shortest path length distribution
  • shortest-path-length heatmap of the maze
  • "most blocking walls" (how much time to reach the father side of the wall could be gained by removing the wall?)

Exercise: Tricky questions about the maze

What is the slowest part of the "wave" solution?

How can we make it faster?

Homework 1

Make a faster version of the maze distance computation.

• any definiton of "faster solution" works; keep it terse
• recommended way:
  • Use DataStructures.jl or any other Julia package to get a priority queue working
  • Implement some simple version of Dijkstra's algorithm
• alternatively, try to optimize our naive array algorithm
• measure your speedup with @time on mazes of at least 128*128 tiles
• data for testing will be available (probably on Moodle)

Submission:

• wrap your code into a package MagicMaze
• include a function that computes the distances, with a type signature such as maze_distance_map(maze::Matrix{Bool}, x::Int, y::Int)::Matrix{Int}
• write a unit test that demonstrates the result
• pack the package and upload it to Moodle