This is a very superficial review, but you have your generic algorithm code mixed in with the problem you're applying it to. In a general sense, this should be avoided whenever possible. Having only loosely related code immediately beside each other is just asking for something bad to happen during a future change.

Whever I start on a learn a new language, I create a GA implementation for practice, and in case I ever actually need it.

It's a fairly easy concept to abstract out:

• The class contains information about the set of possible genes (if a closed set), the max population size, maybe percentage chances of gene crossovers and other chance events (provided you want this to be constant across generations).

• I keep mine simple and only expose a handful of methods. The main method is just a function that automates the processing of an entire generation. It takes the population, runs each genetic sequence through a fitness function (that the caller provides), then chops and repopulates.

By separating the GA code from the use code, you can safely make changes to either without risking breaking some almost unrelated, but coupled code. You also then have the benefit of using your independent GA implementation in any other projects you may need it for without needing to copy and paste select bits from your TSP code.

This is a very superficial review, but you have your generic algorithm code mixed in with the problem you're applying it to. In a general sense, this should be avoided whenever possible. Having only loosely related code immediately beside each other is just asking for something bad to happen during a future change.

Whenever I start on a learn a new language, I usually create a GA implementation for practice, and in case I ever actually need it.

It's a fairly easy concept to abstract out:

• The class contains information about the set of possible genes (if a closed set), the max population size, maybe percentage chances of gene crossovers and other chance events (provided you want this to be constant across generations).

• I keep mine simple and only expose a handful of methods. The main method is just a function that automates the processing of an entire generation. It takes the population, runs each genetic sequence through a fitness function (that the caller provides), then chops and repopulates.

By separating the GA code from the use code, you can safely make changes to either without risking breaking some almost unrelated, but coupled code. You also then have the benefit of using your independent GA implementation in any other projects you may need it for without needing to copy and paste select bits from your TSP code.