Result is inappropriate

[lugingf]

fmt.Println("1:", hungarian.SolveMin([][]float64{
	{3.16, 0, 0, 0, 0},
	{2.58, 0, 0, 0, 0},
	{2.35, 0, 0, 0, 0},
	{1.76, 0, 0, 0, 0},
	{70.9, 0, 0, 0, 0},
}))

First try:
map[0:map[1:0] 1:map[2:0] 2:map[3:0] 3:map[0:1.76] 4:map[4:0]]
Another try:
map[0:map[3:0] 1:map[2:0] 2:map[4:0] 3:map[1:0] 4:map[0:70.9]]

I believe it should be the same map[0:1.76] at least in every run. But currently if I run algo with zeros in whole column it give me diff sum result

[arthurkushman]

Hi @lugingf, yes, it definitely gives diff results right now, but it works pretty well for standard cases (though volatile).
Actually I have no time at this moment, probably this bug is related to the recursive call, where if the same value found it has been replaced by extra traverse:

go-hungarian/hungarian.go

Line 202 in 09ee229

b.checkAndReplace()

[carsonfeng]

Hello arthurkushman, is this bug fixed?

[carsonfeng]

fmt.Println("1:", hungarian.SolveMin([][]float64{
	{3.16, 0, 0, 0, 0},
	{2.58, 0, 0, 0, 0},
	{2.35, 0, 0, 0, 0},
	{1.76, 0, 0, 0, 0},
	{70.9, 0, 0, 0, 0},
}))

First try: map[0:map[1:0] 1:map[2:0] 2:map[3:0] 3:map[0:1.76] 4:map[4:0]] Another try: map[0:map[3:0] 1:map[2:0] 2:map[4:0] 3:map[1:0] 4:map[0:70.9]]

I believe it should be the same map[0:1.76] at least in every run. But currently if I run algo with zeros in whole column it give me diff sum result

After analysis, I found that the instability occurred because the key order was not guaranteed during the map iteration process.