Update version. Replace WRatio by the much faster QRatio (#39)

* Replace WRatio by the much faster QRatio
i30817 · Nov 4, 2023 · f0d8776 · f0d8776
1 parent d6074d0
commit f0d8776
Show file tree

Hide file tree

Showing 3 changed files with 8 additions and 8 deletions.
diff --git a/libretrofuzz/__init__.py b/libretrofuzz/__init__.py
@@ -1 +1 @@
-__version__ = "3.6.1"
+__version__ = "3.6.2"
diff --git a/libretrofuzz/__main__.py b/libretrofuzz/__main__.py
@@ -334,7 +334,7 @@ def __call__(self, name, other, score_cutoff=None):
             return MAX_SCORE
         if not name_ns:
             return 0
-        # score is based on WRatio (a comprehensive 0-100 weighted heuristic from rapidfuzz)
+        # score is based on QRatio (a 0-100 heuristic from rapidfuzz, WRatio is better but unacceptably slow)
         # summed to some custom heuristics
         # Up to DEF_SCORE WRatio is used with 100 being exactly DEF_SCORE, after the heuristics
         # if -min is used, above DEF_SCORE, the heuristics will need better fit to win
@@ -354,33 +354,33 @@ def __call__(self, name, other, score_cutoff=None):
         # 97% of remaining score will be used for different heuristics
         heuristic = remaining * 0.97
         # used denormalized in all returns, just with different percentages
-        wratio = fuzz.WRatio(name, other) * 0.01
+        ratio = fuzz.QRatio(name, other) * 0.01
         # find a exact full name non digit match
         # (lots of dump or disc numbers as subtitles in some dumps)
         # on either a subtitle or a sequence of subtitles from the start
         # note that this doesn't include subtitles in 
         # 'name' matching a subtitle in 'other' and vice versa
         # because a subtitle match like this is a strong indicator of a match,
-        # give it the full default score slot, and wratio for the heuristic slot
+        # give it the full default score slot, and ratio for the heuristic slot
         if not name_ns.isdigit():
             sum_ns = ""
             for sub_ns in other_ns_subs:
                 if name_ns == sub_ns or name_ns == (sum_ns := sum_ns + sub_ns):
-                    rest_of_score += heuristic * wratio
+                    rest_of_score += heuristic * ratio
                     return DEF_SCORE + rest_of_score
         if not other_ns.isdigit():
             sum_ns = ""
             for sub_ns in name_ns_subs:
                 if other_ns == sub_ns or other_ns == (sum_ns := sum_ns + sub_ns):
-                    rest_of_score += heuristic * wratio
+                    rest_of_score += heuristic * ratio
                     return DEF_SCORE + rest_of_score
         # heuristic measures if the name is more completely at the start of other name
         common = len(os.path.commonprefix([name_ns, other_ns])) / len(name_ns)
         # heuristic measures how similar the lenghts of the names are
         parity = min(len(name_ns),len(other_ns))/max(len(name_ns),len(other_ns))
         rest_of_score += (heuristic * common * 0.80) + (heuristic * parity * 0.20)
         # remember that WRatio fills the DEF_SCORE slot
-        return rest_of_score + DEF_SCORE * wratio
+        return rest_of_score + DEF_SCORE * ratio
 
 # ---------------------------------------------------------------
 # Normalization functions, part of the functions that change both

diff --git a/pyproject.toml b/pyproject.toml
@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "libretrofuzz"
-version = "3.6.1"
+version = "3.6.2"
 description = "Fuzzy Retroarch thumbnail downloader"
 authors = ["i30817 <i30817@gmail.com>"]
 license = "MIT"