Refactor fallback logic into its own function

python/pymusiclooper/analysis.py

@@ -267,7 +267,57 @@ def _db_diff(power_db_f1: np.ndarray, power_db_f2: np.ndarray) -> float:
 def _norm(a: np.ndarray) -> float:
     return np.sqrt(np.sum(np.abs(a) ** 2, axis=0))
 
+
 @njit(cache=True)
+def _find_candidate_pairs_fallback(
+    chroma: np.ndarray,
+    power_db: np.ndarray,
+    beats: np.ndarray,
+    deviation: np.ndarray,
+    min_loop_duration: int,
+    max_loop_duration: int,
+    acceptable_loudness_difference: float,
+) -> List[Tuple[int, int, float, float]]:
+    """Python fallback function.
+    Generates a list of all valid candidate loop pairs using combinations of beat indices,
+    by comparing the notes using the chroma spectrogram and their loudness difference
+
+    Args:
+        chroma (np.ndarray): The chroma spectrogram
+        power_db (np.ndarray): The power spectrogram in dB
+        beats (np.ndarray): The frame indices of detected beats
+        min_loop_duration (int): Minimum loop duration (in frames)
+        max_loop_duration (int): Maximum loop duration (in frames)
+
+    Returns:
+        List[Tuple[int, int, float, float]]: A list of tuples containing each candidate loop pair data in the following format (loop_start, loop_end, note_distance, loudness_difference)
+    """
+    candidate_pairs = []
+
+    for idx, loop_end in enumerate(beats):
+        for loop_start in beats:
+            loop_length = loop_end - loop_start
+            if loop_length < min_loop_duration:
+                break
+            if loop_length > max_loop_duration:
+                continue
+            note_distance = _norm(chroma[..., loop_end] - chroma[..., loop_start])
+
+            if note_distance <= deviation[idx]:
+                loudness_difference = _db_diff(
+                    power_db[..., loop_end], power_db[..., loop_start]
+                )
+                if loudness_difference <= acceptable_loudness_difference:
+                    loop_pair = (
+                        int(loop_start),
+                        int(loop_end),
+                        note_distance,
+                        loudness_difference,
+                    )
+                    candidate_pairs.append(loop_pair)
+    return candidate_pairs
+
+
 def _find_candidate_pairs(
     chroma: np.ndarray,
     power_db: np.ndarray,
@@ -313,30 +363,15 @@ def _find_candidate_pairs(
         )
     except Exception:
         # Python fallback
-        candidate_pairs = []
-
-        for idx, loop_end in enumerate(beats):
-            for loop_start in beats:
-                loop_length = loop_end - loop_start
-                if loop_length < min_loop_duration:
-                    break
-                if loop_length > max_loop_duration:
-                    continue
-                note_distance = _norm(chroma[..., loop_end] - chroma[..., loop_start])
-
-                if note_distance <= deviation[idx]:
-                    loudness_difference = _db_diff(
-                        power_db[..., loop_end], power_db[..., loop_start]
-                    )
-                    loop_pair = (
-                        int(loop_start),
-                        int(loop_end),
-                        note_distance,
-                        loudness_difference,
-                    )
-                    if loudness_difference <= ACCEPTABLE_LOUDNESS_DIFFERENCE:
-                        candidate_pairs.append(loop_pair)
-        return candidate_pairs
+        return _find_candidate_pairs_fallback(
+            chroma,
+            power_db,
+            beats,
+            deviation,
+            min_loop_duration,
+            max_loop_duration,
+            ACCEPTABLE_LOUDNESS_DIFFERENCE,
+        )
 
 
 def _assess_and_filter_loop_pairs(