Feature/parangonar

matchmaker/__init__.py

@@ -6,7 +6,7 @@
 
 import pkg_resources
 
-from . import dp, features, io, prob, utils
+from . import dp, features, io, prob, utils, external
 from .matchmaker import *
 
 __all__ = ["dp", "features", "io", "prob", "utils"]

matchmaker/external/__init__.py

@@ -0,0 +1,84 @@
+#!/usr/bin/python
+# -*- coding: utf-8 -*-
+"""
+Top module for alignment methods imported from the parangonar library:
+https://github.com/sildater/parangonar
+"""
+import parangonar as pa
+# from matchmaker.base import OnlineAlignment
+from typing import Callable, Dict, Generator
+from numpy.typing import NDArray
+
+QUEUE_SENTINEL = object()
+
+class OnlineParangonarAlignment():#(OnlineAlignment):
+    def __init__(self, 
+                 queue,
+                 score_note_array,
+                 parangonar_tracker_type: str = "SLT_OLTW"):
+        # an instance of 
+        self.queue = queue
+        if parangonar_tracker_type == "SLT_OLTW":
+            self.parangonar_tracker = pa.TOLTWMatcher(score_note_array, 
+                                                      tracker_type=parangonar_tracker_type)
+        elif parangonar_tracker_type == "SL_OLTW":
+            self.parangonar_tracker = pa.OLTWMatcher(score_note_array, 
+                                                    tracker_type=parangonar_tracker_type)
+        elif parangonar_tracker_type == "OPTM":
+            self.parangonar_tracker = pa.OnlinePureTransformerMatcher(score_note_array)
+        elif parangonar_tracker_type == "OTM":
+            self.parangonar_tracker = pa.OnlineTransformerMatcher(score_note_array)
+
+    def __call__(self, performance_note):
+        # process
+        score_position = self.parangonar_tracker(performance_note)
+        return score_position
+
+    def run(self) -> Generator[int, None, float]:
+        while self.parangonar_tracker.is_still_following():
+            input_feature = self.queue.get(block=True)
+            if input_feature is QUEUE_SENTINEL:
+                print("empty queue")
+                return None
+            else:
+                current_state = self(input_feature)
+                yield current_state
+
+        return None
+
+    def run_offline(self):
+        self.queue.put(QUEUE_SENTINEL)
+        for position in self.run():
+            print(position, self.parangonar_tracker.unique_onsets[position])
+
+        return self.parangonar_tracker.warping_path
+
+if __name__ == "__main__":
+    import partitura as pt
+    from queue import Queue
+    # load the example match file included in the library
+    perf_match, groundtruth_alignment, score_match = pt.load_match(
+        filename= pa.EXAMPLE, # 
+        create_score=True
+    )
+
+    # compute note arrays from the loaded score and performance
+    pna_match = perf_match[0].note_array()
+    sna_match = score_match[0].note_array(include_grace_notes=True)
+
+    # create queue
+    input_queue = Queue()
+    for note_row in pna_match:
+        input_queue.put(note_row)
+
+    # create matchmaker follower
+    score_follower = OnlineParangonarAlignment(
+        input_queue, sna_match, "OTM")
+
+    # run the follower offline
+    warping_path = score_follower.run_offline()
+
+
+
+
+

pyproject.toml

@@ -39,6 +39,7 @@ dependencies = [
     "pyaudio>=0.2.14",
     "pyfluidsynth>=1.3.3",
     "matplotlib>=3.9.4",
+    "parangonar",
 ]
 
 [project.optional-dependencies]