Example: minimal cost is not returned

[taylorpetty]

Consider the following scenario:

Inserting a C costs 3. Everything else costs 1. The function returns lev(' ','C') = 3, but in actuality the minimal cost should be 2, since it's cheaper to insert an A (+1) and then transform the A into a C (+1).

I haven't been able to think of a solution yet, but a workaround at the least is to say that the single-character costs entered in as weights must already be minimal.

Deletion has the same issue, but substitution does not.

Here's an example:

from weighted_levenshtein import lev
import numpy as np

icosts = np.ones(128, dtype=np.float64)
dcosts = np.ones(128, dtype=np.float64)
scosts = np.ones((128, 128), dtype=np.float64)

icosts[ord('C')] = 4

dcosts[ord('G')] = 10

scosts[ord('A'),ord('T')] = 7

print(lev('','C',icosts,dcosts,scosts)) # returns 4, should return 2

print(lev('','A',icosts,dcosts,scosts)+lev('A','C',icosts,dcosts,scosts)) # returns 2

print(lev('G','',icosts,dcosts,scosts)) # returns 10, should return 2

print(lev('G','T',icosts,dcosts,scosts)+lev('T','',icosts,dcosts,scosts)) # returns 2

print(lev('A','T',icosts,dcosts,scosts)) # returns 2
```

[pachewise]

Hi @tmpetty - thanks for reporting this issue. I was able to reproduce. If you do come up with a solution, please open a PR and we'll review and approve. (Right now, we do not have engineers who can focus on fixing this)