`filter` score about alfred-workflow

[deanishe]

That makes a lot of sense. The ranking maths in filter definitely needs improvement.

Could you rebase your fork first, however? Should make a merge easier/possible.

[fractaledmind]

I'm working on rebasing. Also, testing some other elements of filter. Will pull after.

[fractaledmind]

I've found another bug. It relates to the scoring, so instead of starting a new Issue, I'm posting here.

Essentially, filter will chop the results list when multiple items have the same score and same value:

if score > 0:
    results[(100.0 / score, value.lower())] = (item, score, rule)

With ZotQuery this is actually a common occurrence. For example, whenever the scope is simply creators' last names, any items written only by "Smith" will have the value.lower() = "smith". If the query was "smith", then each result will have the same score and same value, so whenever the filter sorts on keys:

# sort on keys, then discard the keys
keys = sorted(results.keys(), reverse=ascending)
results = [results.get(k) for k in keys]

only one item will remain.

Proposed Solution
Use enumeration to make sort keys item unique. So, for example, at the opening of the for block:

for i, item in enumerate(items):
    rule = None
    score = 0
    value = key(item)

This would allow for change in the key function like so:

if score > 0:
    results[(100.0 / score, value.lower(), i)] = (item, score, rule)

[deanishe]

The solution was included in your other pull request.