how-it-works.md

How it works

This document explains how that solution was developed and how it can be explored.

Visual demonstration

Preview generated by Kibana

Algorithms

Controlling the maximum cargos for each truck

• The default max cargos for truck is 1

make dev

• We can try a max cargo option for truck = 2

make dev/maxcargo

Calculating distances

• The default distance calculation use the Harvesine formula

make dev

• More accurate, you can use the Google Routes remote provider (you must provide GOOGLE_API_KEY env var):

make run/remote

Designating trucks for cargos

Five different algorithms were developed, with 2 remaining:

• Iteration with sorted dictionary The classic cargos x trucks iteration (sorted dictionary)

make run

• Iteration only for cargos + O log(n) kd-tree (optimized when the trucks are loaded): While the trucks are loaded, for each truck appended, we use a kd-tree to manage the locations.

make run/kdtree

How to dev

• Watch for modifications, run lint and tests

make dev/watch

• Run dev mode (with performance check)

make dev

• Apply lint corrections (90% effective), based on PEP 8

make lint/fix

References

I'm not Python senior, but I really appreciate Python's language.

• I followed The Zen of Python
• Read Luciano Ramalho's "Fluent Python"
• Beyond the pythonic way, or in case of doubts, my guide is the Uncle Bob's "Clean Code"

What I can do with more time

• Add a hash function for the locations: to improve even more the checking for the closest trucks
• Add parallelization processing
• Add algorithm to decide which cargo is best in case of an already designated truck
• Add Docker: there's some code, but it's not totally functional
• Plot maps for the inputs and the final solution

More options

• For more options check the Makefile