Merge pull request #5579 from Textualize/fuzzy-fix

Fix for fuzzy matcher

Author: willmcgugan

CHANGELOG.md

@@ -5,6 +5,12 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](http://keepachangelog.com/)
 and this project adheres to [Semantic Versioning](http://semver.org/).
 
+## [2.1.2] - 2025-02-26
+
+### Fixed
+
+- Fixed command palette fuzzy search bailing too early https://github.com/Textualize/textual/pull/5579
+
 ## [2.1.1] - 2025-02-22
 
 ### Fixed
@@ -2773,6 +2779,7 @@ https://textual.textualize.io/blog/2022/11/08/version-040/#version-040
 - New handler system for messages that doesn't require inheritance
 - Improved traceback handling
 
+[2.1.2]: https://github.com/Textualize/textual/compare/v2.1.1...v2.1.2
 [2.1.1]: https://github.com/Textualize/textual/compare/v2.1.0...v2.1.1
 [2.1.0]: https://github.com/Textualize/textual/compare/v2.0.4...v2.1.0
 [2.0.4]: https://github.com/Textualize/textual/compare/v2.0.3...v2.0.4

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "textual"
-version = "2.1.1"
+version = "2.1.2"
 homepage = "https://github.com/Textualize/textual"
 repository = "https://github.com/Textualize/textual"
 documentation = "https://textual.textualize.io/"

src/textual/command.py

@@ -949,7 +949,6 @@ async def _search_for(
             )
             for provider in self._providers
         ]
-
         # Set up a delay for showing that we're busy.
         self._start_busy_countdown()
 

src/textual/fuzzy.py

@@ -13,6 +13,7 @@
 
 import rich.repr
 
+from textual.cache import LRUCache
 from textual.content import Content
 from textual.visual import Style
 
@@ -43,8 +44,8 @@ def branch(self, offset: int) -> tuple[_Search, _Search]:
     def groups(self) -> int:
         """Number of groups in offsets."""
         groups = 1
-        last_offset = self.offsets[0]
-        for offset in self.offsets[1:]:
+        last_offset, *offsets = self.offsets
+        for offset in offsets:
             if offset != last_offset + 1:
                 groups += 1
             last_offset = offset
@@ -57,13 +58,17 @@ class FuzzySearch:
     Unlike a regex solution, this will finds all possible matches.
     """
 
+    cache: LRUCache[tuple[str, str, bool], tuple[float, tuple[int, ...]]] = LRUCache(
+        1024 * 4
+    )
+
     def __init__(self, case_sensitive: bool = False) -> None:
         """Initialize fuzzy search.
 
         Args:
             case_sensitive: Is the match case sensitive?
         """
-        self.cache: dict[tuple[str, str, bool], tuple[float, tuple[int, ...]]] = {}
+
         self.case_sensitive = case_sensitive
 
     def match(self, query: str, candidate: str) -> tuple[float, tuple[int, ...]]:
@@ -76,7 +81,6 @@ def match(self, query: str, candidate: str) -> tuple[float, tuple[int, ...]]:
         Returns:
             A pair of (score, tuple of offsets). `(0, ())` for no result.
         """
-
         query_regex = ".*?".join(f"({escape(character)})" for character in query)
         if not search(
             query_regex, candidate, flags=0 if self.case_sensitive else IGNORECASE
@@ -124,13 +128,13 @@ def score(search: _Search) -> float:
             """
             # This is a heuristic, and can be tweaked for better results
             # Boost first letter matches
-            score: float = sum(
-                (2.0 if offset in first_letters else 1.0) for offset in search.offsets
+            offset_count = len(search.offsets)
+            score: float = offset_count + len(
+                first_letters.intersection(search.offsets)
             )
             # Boost to favor less groups
-            offset_count = len(search.offsets)
             normalized_groups = (offset_count - (search.groups - 1)) / offset_count
-            score *= 1 + (normalized_groups**2)
+            score *= 1 + (normalized_groups * normalized_groups)
             return score
 
         stack: list[_Search] = [_Search()]
@@ -139,20 +143,24 @@ def score(search: _Search) -> float:
         query_size = len(query)
         find = candidate.find
         # Limit the number of loops out of an abundance of caution.
-        # This would be hard to reach without contrived data.
-        remaining_loops = 200
-
+        # This should be hard to reach without contrived data.
+        remaining_loops = 10_000
         while stack and (remaining_loops := remaining_loops - 1):
             search = pop()
             offset = find(query[search.query_offset], search.candidate_offset)
             if offset != -1:
+                if not set(candidate[search.candidate_offset :]).issuperset(
+                    query[search.query_offset :]
+                ):
+                    # Early out if there is not change of a match
+                    continue
                 advance_branch, branch = search.branch(offset)
                 if advance_branch.query_offset == query_size:
                     yield score(advance_branch), advance_branch.offsets
                     push(branch)
                 else:
-                    push(advance_branch)
                     push(branch)
+                    push(advance_branch)
 
 
 @rich.repr.auto