SequenceMatcher should have `autojunk` set to `false` for `partial_ratio`

[wesbarnett]

Here SequenceMatcher is being called with the default autojunk parameter being inferred as true:

https://github.com/nol13/fuzzball.js/blob/4e7393ca7af47f71abefe6071b3b7d3d82ece4ef/fuzzball.js#L932

This does not match the Python implementation of fuzzywuzzy / rapidfuzz as documented here:

blocks = SequenceMatcher(None, needle, longer, False).get_matching_blocks()
score = 0
for block in blocks:
    long_start = block[1] - block[0] if (block[1] - block[0]) > 0 else 0
    long_end = long_start + len(shorter)
    long_substr = longer[long_start:long_end]
    score = max(score, fuzz.ratio(needle, long_substr))

Alternatively, it could be an option the user could pass.

[nol13]

Ya this could definitely at the very least be added as an option. If that's what their current implementation is doing I can probably keep this in sync and make it default. I think their implementation has probably diverged since this was written when they split from the old codebase? (https://github.com/seatgeek/fuzzywuzzy/blob/master/fuzzywuzzy/fuzz.py#L61)

Edit: Nm I see where they explain this difference here now https://github.com/rapidfuzz/RapidFuzz/blob/main/api_differences.md#partial_ratio-implementation

[nol13]

Do you happen to have any good test cases for this? Will add the option, but I wanted to better understand the behavior before making it the default. I'm not sure if the difflib port I'm using's autojunk option works exactly like the current Python implementation either.

[nol13]

Tbh I couldn't even find a good test case, but no harm in adding it as an option. Will leave the default to be true for now to not change behavior.

[nol13]

2.2.3 is released so closing but if there was an alternative javascript implementation that was better that could work to