Tons of new detail in the diff. So much so that I added a DetailLevel parameter throughout. (#6)

* Lots of detail/style about notes, and style about everything else.

* Don't create style info by looking at it.

* Better visualization of style differences (e.g. "changed note color").

* Add DetailLevel specification. Flesh out missing TextStyle stuff (e.g. alignVertical).

* Placement should only be diffed if DetailLevel >= AllObjectsWithStyle.

* Add new command line argument -d/--detail to specify DetailLevel of diff requested.

* More complete note style diffing. Update API docs.

* Fix up detailed annotation of notes.

* Major cleanup of style diffing, fixing a few bugs along the way.

Author: gregchapman-dev

musicdiff/__init__.py

@@ -20,6 +20,7 @@
 import music21 as m21
 
 from musicdiff.m21utils import M21Utils
+from musicdiff.m21utils import DetailLevel
 from musicdiff.annotation import AnnScore
 from musicdiff.comparison import Comparison
 from musicdiff.visualization import Visualization
@@ -48,6 +49,7 @@ def diff(score1: Union[str, Path, m21.stream.Score],
          out_path2:  Union[str, Path] = None,
          force_parse: bool = True,
          visualize_diffs: bool = True,
+         detail: DetailLevel = DetailLevel.Default
         ) -> int:
     '''
     Compare two musical scores and optionally save/display the differences as two marked-up
@@ -72,6 +74,9 @@ def diff(score1: Union[str, Path, m21.stream.Score],
         visualize_diffs (bool): Whether or not to render diffs as marked up PDFs. If False,
             the only result of the call will be the return value (the number of differences).
             (default is True)
+        detail (DetailLevel): What level of detail to use during the diff.  Can be
+            GeneralNotesOnly, AllObjects, AllObjectsWithStyle or Default (Default is
+            currently equivalent to AllObjects).
 
     Returns:
         int: The number of differences found (0 means the scores were identical, None means the diff failed)
@@ -137,8 +142,8 @@ def diff(score1: Union[str, Path, m21.stream.Score],
         return None
 
     # scan each score, producing an annotated wrapper
-    annotated_score1: AnnScore = AnnScore(score1)
-    annotated_score2: AnnScore = AnnScore(score2)
+    annotated_score1: AnnScore = AnnScore(score1, detail)
+    annotated_score2: AnnScore = AnnScore(score2, detail)
 
     diff_list: List = None
     _cost: int = None

musicdiff/__main__.py

@@ -16,10 +16,10 @@
 import argparse
 
 from musicdiff import diff
+from musicdiff import DetailLevel
 
 # To use the new Humdrum importer from converter21 in place of the one in music21:
-# git clone https://github.com/gregchapman-dev/converter21.git
-# pip install converter21   # or pip install -e converter21 if you want it "editable"
+# pip install converter21
 # Then uncomment all lines in this file marked "# c21"
 # import music21 as m21 # c21
 # from converter21 import HumdrumConverter # c21
@@ -43,11 +43,24 @@
                         help="first music score file to compare (any format music21 can parse)")
     parser.add_argument("file2",
                         help="second music score file to compare (any format music21 can parse)")
+    parser.add_argument("-d", "--detail", default="Default",
+                        choices=["GeneralNotesOnly", "AllObjects", "AllObjectsWithStyle", "Default"],
+                        help="set detail level")
     args = parser.parse_args()
 
+    detail: DetailLevel = DetailLevel.Default
+    if args.detail == "GeneralNotesOnly":
+        detail = DetailLevel.GeneralNotesOnly
+    elif args.detail == "AllObjects":
+        detail = DetailLevel.AllObjects
+    elif args.detail == "AllObjectsWithStyle":
+        detail = DetailLevel.AllObjectsWithStyle
+    elif args.detail == "Default":
+        detail = DetailLevel.Default
+
     # Note that diff() can take a music21 Score instead of a file, for either
     # or both arguments.
     # Note also that diff() can take str or pathlib.Path for files.
-    numDiffs: int = diff(args.file1, args.file2)
+    numDiffs: int = diff(args.file1, args.file2, detail=detail)
     if numDiffs is not None and numDiffs == 0:
         print(f'Scores in {args.file1} and {args.file2} are identical.', file=sys.stderr)

musicdiff/annotation.py

@@ -15,25 +15,45 @@
 __docformat__ = "google"
 
 from fractions import Fraction
+from typing import Optional
 
 import music21 as m21
 
 from musicdiff import M21Utils
-
+from musicdiff import DetailLevel
 
 class AnnNote:
-    def __init__(self, general_note, enhanced_beam_list, tuplet_list):
+    def __init__(self, general_note: m21.note.GeneralNote, enhanced_beam_list, tuplet_list, detail: DetailLevel = DetailLevel.Default):
         """
         Extend music21 GeneralNote with some precomputed, easily compared information about it.
 
         Args:
             general_note (music21.note.GeneralNote): The music21 note/chord/rest to extend.
             enhanced_beam_list (list): A list of beaming information about this GeneralNote.
             tuplet_list (list): A list of tuplet info about this GeneralNote.
+            detail (DetailLevel): What level of detail to use during the diff.  Can be
+                GeneralNotesOnly, AllObjects, AllObjectsWithStyle or Default (Default is
+                currently equivalent to AllObjects).
+
         """
         self.general_note = general_note.id
         self.beamings = enhanced_beam_list
         self.tuplets = tuplet_list
+
+        self.stylestr: str = ''
+        self.styledict: dict = {}
+        if M21Utils.has_style(general_note):
+            self.styledict = M21Utils.obj_to_styledict(general_note, detail)
+        self.noteshape: str = 'normal'
+        self.noteheadFill: Optional[bool] = None
+        self.noteheadParenthesis: bool = False
+        self.stemDirection: str = 'unspecified'
+        if detail >= DetailLevel.AllObjectsWithStyle and isinstance(general_note, m21.note.NotRest):
+            self.noteshape = general_note.notehead
+            self.noteheadFill = general_note.noteheadFill
+            self.noteheadParenthesis = general_note.noteheadParenthesis
+            self.stemDirection = general_note.stemDirection
+
         # compute the representation of NoteNode as in the paper
         # pitches is a list  of elements, each one is (pitchposition, accidental, tie)
         if general_note.isRest:
@@ -101,7 +121,8 @@ def notation_size(self):
     def __repr__(self):
         # does consider the MEI id!
         return (f"{self.pitches},{self.note_head},{self.dots},{self.beamings}," +
-                f"{self.tuplets},{self.general_note},{self.articulations},{self.expressions}")
+                f"{self.tuplets},{self.general_note},{self.articulations},{self.expressions}" +
+                f"{self.styledict}")
 
     def __str__(self):
         """
@@ -151,6 +172,22 @@ def __str__(self):
         if len(self.expressions) > 0:  # add for articulations
             for e in self.expressions:
                 string += e
+
+        if self.noteshape != 'normal':
+            string += f"noteshape={self.noteshape}"
+        if self.noteheadFill is not None:
+            string += f"noteheadFill={self.noteheadFill}"
+        if self.noteheadParenthesis:
+            string += f"noteheadParenthesis={self.noteheadParenthesis}"
+        if self.stemDirection != 'unspecified':
+            string += f"stemDirection={self.stemDirection}"
+
+        # and then the style fields
+        for i, (k, v) in enumerate(self.styledict.items()):
+            if i > 0:
+                string += ","
+            string += f"{k}={v}"
+
         return string
 
     def get_note_ids(self):
@@ -188,7 +225,7 @@ def __eq__(self, other):
 
 
 class AnnExtra:
-    def __init__(self, extra: m21.base.Music21Object, measure: m21.stream.Measure, score: m21.stream.Score):
+    def __init__(self, extra: m21.base.Music21Object, measure: m21.stream.Measure, score: m21.stream.Score, detail: DetailLevel = DetailLevel.Default):
         """
         Extend music21 non-GeneralNote and non-Stream objects with some precomputed, easily compared information about it.
         Examples: TextExpression, Dynamic, Clef, Key, TimeSignature, MetronomeMark, etc.
@@ -197,6 +234,9 @@ def __init__(self, extra: m21.base.Music21Object, measure: m21.stream.Measure, s
             extra (music21.base.Music21Object): The music21 non-GeneralNote/non-Stream object to extend.
             measure (music21.stream.Measure): The music21 Measure the extra was found in.  If the extra
                 was found in a Voice, this is the Measure that the Voice was found in.
+            detail (DetailLevel): What level of detail to use during the diff.  Can be
+                GeneralNotesOnly, AllObjects, AllObjectsWithStyle or Default (Default is
+                currently equivalent to AllObjects).
         """
         self.extra = extra.id
         self.offset: float
@@ -213,6 +253,9 @@ def __init__(self, extra: m21.base.Music21Object, measure: m21.stream.Measure, s
             self.offset = float(extra.getOffsetInHierarchy(measure))
             self.duration = float(extra.duration.quarterLength)
         self.content: str = M21Utils.extra_to_string(extra)
+        self.styledict: str = {}
+        if M21Utils.has_style(extra):
+            self.styledict = M21Utils.obj_to_styledict(extra, detail) # includes extra.placement if present
         self._notation_size: int = 1 # so far, always 1, but maybe some extra will be bigger someday
 
         # precomputed representations for faster comparison
@@ -235,20 +278,27 @@ def __str__(self):
         Returns:
             str: the compared representation of the AnnExtra. Does not consider music21 id.
         """
-        return f'[{self.content},off={self.offset},dur={self.duration}]'
+        string = f'{self.content},off={self.offset},dur={self.duration}'
+        # and then any style fields
+        for k, v in self.styledict.items():
+            string += f",{k}={v}"
+        return string
 
     def __eq__(self, other):
         # equality does not consider the MEI id!
         return self.precomputed_str == other.precomputed_str
 
 
 class AnnVoice:
-    def __init__(self, voice):
+    def __init__(self, voice: m21.stream.Voice, detail: DetailLevel = DetailLevel.Default):
         """
         Extend music21 Voice with some precomputed, easily compared information about it.
 
         Args:
             voice (music21.stream.Voice): The music21 voice to extend.
+            detail (DetailLevel): What level of detail to use during the diff.  Can be
+                GeneralNotesOnly, AllObjects, AllObjectsWithStyle or Default (Default is
+                currently equivalent to AllObjects).
         """
         self.voice = voice.id
         note_list = M21Utils.get_notes(voice)
@@ -269,7 +319,7 @@ def __init__(self, voice):
             self.annot_notes = []
             for i, n in enumerate(note_list):
                 self.annot_notes.append(
-                    AnnNote(n, self.en_beam_list[i], self.tuplet_list[i])
+                    AnnNote(n, self.en_beam_list[i], self.tuplet_list[i], detail)
                 )
 
         self.n_of_notes = len(self.annot_notes)
@@ -323,35 +373,44 @@ def get_note_ids(self):
 
 
 class AnnMeasure:
-    def __init__(self, measure, score, spannerBundle):
+    def __init__(self, measure: m21.stream.Measure,
+                       score: m21.stream.Score,
+                       spannerBundle: m21.spanner.SpannerBundle,
+                       detail: DetailLevel = DetailLevel.Default):
         """
         Extend music21 Measure with some precomputed, easily compared information about it.
 
         Args:
             measure (music21.stream.Measure): The music21 measure to extend.
+            score (music21.stream.Score): the enclosing music21 Score.
+            spannerBundle (music21.spanner.SpannerBundle): a bundle of all the spanners in the score.
+            detail (DetailLevel): What level of detail to use during the diff.  Can be
+                GeneralNotesOnly, AllObjects, AllObjectsWithStyle or Default (Default is
+                currently equivalent to AllObjects).
         """
         self.measure = measure.id
         self.voices_list = []
         if (
             len(measure.voices) == 0
         ):  # there is a single AnnVoice ( == for the library there are no voices)
-            ann_voice = AnnVoice(measure)
+            ann_voice = AnnVoice(measure, detail)
             if ann_voice.n_of_notes > 0:
                 self.voices_list.append(ann_voice)
         else:  # there are multiple voices (or an array with just one voice)
             for voice in measure.voices:
-                ann_voice = AnnVoice(voice)
+                ann_voice = AnnVoice(voice, detail)
                 if ann_voice.n_of_notes > 0:
                     self.voices_list.append(ann_voice)
         self.n_of_voices = len(self.voices_list)
 
         self.extras_list = []
-        for extra in M21Utils.get_extras(measure, spannerBundle):
-            self.extras_list.append(AnnExtra(extra, measure, score))
+        if detail >= DetailLevel.AllObjects:
+            for extra in M21Utils.get_extras(measure, spannerBundle):
+                self.extras_list.append(AnnExtra(extra, measure, score, detail))
 
-        # For correct comparison, sort the extras_list, so that any list slices
-        # that all have the same offset are sorted alphabetically.
-        self.extras_list.sort(key=lambda e: ( e.offset, str(e) ))
+            # For correct comparison, sort the extras_list, so that any list slices
+            # that all have the same offset are sorted alphabetically.
+            self.extras_list.sort(key=lambda e: ( e.offset, str(e) ))
 
         # precomputed values to speed up the computation. As they start to be long, they are hashed
         self.precomputed_str = hash(self.__str__())
@@ -400,17 +459,25 @@ def get_note_ids(self):
 
 
 class AnnPart:
-    def __init__(self, part, score, spannerBundle):
+    def __init__(self, part: m21.stream.Part,
+                       score: m21.stream.Score,
+                       spannerBundle: m21.spanner.SpannerBundle,
+                       detail: DetailLevel = DetailLevel.Default):
         """
         Extend music21 Part/PartStaff with some precomputed, easily compared information about it.
 
         Args:
             part (music21.stream.Part, music21.stream.PartStaff): The music21 Part/PartStaff to extend.
+            score (music21.stream.Score): the enclosing music21 Score.
+            spannerBundle (music21.spanner.SpannerBundle): a bundle of all the spanners in the score.
+            detail (DetailLevel): What level of detail to use during the diff.  Can be
+                GeneralNotesOnly, AllObjects, AllObjectsWithStyle or Default (Default is
+                currently equivalent to AllObjects).
         """
         self.part = part.id
         self.bar_list = []
         for measure in part.getElementsByClass("Measure"):
-            ann_bar = AnnMeasure(measure, score, spannerBundle)  # create the bar objects
+            ann_bar = AnnMeasure(measure, score, spannerBundle, detail)  # create the bar objects
             if ann_bar.n_of_voices > 0:
                 self.bar_list.append(ann_bar)
         self.n_of_bars = len(self.bar_list)
@@ -456,19 +523,22 @@ def get_note_ids(self):
 
 
 class AnnScore:
-    def __init__(self, score):
+    def __init__(self, score: m21.stream.Score, detail: DetailLevel = DetailLevel.Default):
         """
         Take a music21 score and store it as a sequence of Full Trees.
         The hierarchy is "score -> parts -> measures -> voices -> notes"
         Args:
             score (music21.stream.Score): The music21 score
+            detail (DetailLevel): What level of detail to use during the diff.  Can be
+                GeneralNotesOnly, AllObjects, AllObjectsWithStyle or Default (Default is
+                currently equivalent to AllObjects).
         """
         self.score = score.id
         self.part_list = []
-        spannerBundle = score.spannerBundle
+        spannerBundle: m21.spanner.SpannerBundle = score.spannerBundle
         for part in score.parts.stream():
             # create and add the AnnPart object to part_list
-            ann_part = AnnPart(part, score, spannerBundle)
+            ann_part = AnnPart(part, score, spannerBundle, detail)
             if ann_part.n_of_bars > 0:
                 self.part_list.append(ann_part)
         self.n_of_parts = len(self.part_list)

musicdiff/comparison.py

@@ -504,6 +504,11 @@ def _annotated_extra_diff(annExtra1: AnnExtra, annExtra2: AnnExtra):
             cost += duration_cost
             op_list.append(("extradurationedit", annExtra1, annExtra2, duration_cost))
 
+        # add for the style
+        if annExtra1.styledict != annExtra2.styledict:
+            cost += 1
+            op_list.append(("extrastyleedit", annExtra1, annExtra2, 1))
+
         return op_list, cost
 
     @staticmethod
@@ -632,6 +637,26 @@ def _annotated_note_diff(annNote1: AnnNote, annNote2: AnnNote):
             )
             op_list.extend(expr_op_list)
             cost += expr_cost
+        # add for noteshape
+        if annNote1.noteshape != annNote2.noteshape:
+            cost += 1
+            op_list.append(("editnoteshape", annNote1, annNote2, 1))
+        # add for noteheadFill
+        if annNote1.noteheadFill != annNote2.noteheadFill:
+            cost += 1
+            op_list.append(("editnoteheadfill", annNote1, annNote2, 1))
+        # add for noteheadParenthesis
+        if annNote1.noteheadParenthesis != annNote2.noteheadParenthesis:
+            cost += 1
+            op_list.append(("editnoteheadparenthesis", annNote1, annNote2, 1))
+        # add for stemDirection
+        if annNote1.stemDirection != annNote2.stemDirection:
+            cost += 1
+            op_list.append(("editstemdirection", annNote1, annNote2, 1))
+        # add for the styledict
+        if annNote1.styledict != annNote2.styledict:
+            cost += 1
+            op_list.append(("editstyle", annNote1, annNote2, 1))
 
         return op_list, cost