# -*- coding: utf-8 -*- # ------------------------------------------------------------------------------ # Name: search.serial.py # Purpose: music21 classes for serial searching # # Authors: Carl Lian # Michael Scott Asato Cuthbert # # Copyright: Copyright © 2009-2012, 2016 Michael Scott Asato Cuthbert # License: BSD, see license.txt # ------------------------------------------------- from __future__ import annotations from collections import Counter import copy from operator import attrgetter import unittest from music21 import base from music21 import common from music21 import environment from music21.serial import pcToToneRow, ToneRow from music21 import spanner from music21 import stream environLocal = environment.Environment() # ------- parsing functions for atonal music ------- class ContiguousSegmentOfNotes(base.Music21Object): ''' Class whose instantiations represent contiguous segments of notes and chords appearing within a :class:`~music21.stream.Stream`. Generally speaking, these objects are instantiated internally, though it is possible for the user to create them as well. >>> s = stream.Stream() >>> p = stream.Part() >>> n1 = note.Note('c4') >>> n2 = note.Note('d4') >>> p.append(n1) >>> p.append(n2) >>> p = p.makeMeasures() >>> s.insert(0, p) >>> cdContiguousSegment = search.serial.ContiguousSegmentOfNotes([n1, n2], s, 0) >>> cdContiguousSegment ''' _DOC_ATTR: dict[str, str] = { 'segment': 'The list of notes and chords in the contiguous segment.', 'containerStream': ''' The stream containing the contiguous segment - all contiguous segments must have a container stream.''', 'partNumber': ''' The part number in which the segment appears, or None (if the container stream has no parts). Note that this attribute is zero-indexed, so the top (e.g. soprano) part is labeled 0.''', 'activeSegment': ''' A list of pitch classes representing the way the contiguous segment of notes is being read as a sequence of single pitches. Set to None unless the container stream is being searched for segments or multisets (for example, using :meth:`~music21.search.serial.SegmentMatcher.find`), in which case the representation depends on the segments or multisets being searched for. If there are no chords in the segment, this attribute will simply give the pitch classes of the notes in the segment.''', 'matchedSegment': ''' A list of pitch classes representing the segment to which the contiguous segment of notes is matched when segments or multisets are searched for (for example, using :func:`~music21.search.serial.findSegments`); otherwise set to None. Note that the contiguous segment will only be matched to one of the segments or multisets being searched for.''' } _DOC_ORDER = ['startMeasureNumber', 'startOffset', 'zeroCenteredTransformationsFromMatched', 'originalCenteredTransformationsFromMatched'] def __init__(self, segment=None, containerStream=None, partNumber=0, **keywords): super().__init__(**keywords) self.segment = segment self.containerStream = containerStream self.partNumber = partNumber self.activeSegment = [] self.matchedSegment = None def _reprInternal(self): chordList = [] for ch in self.segment: chordPitches = ' '.join(str(p) for p in ch.pitches) chordList.append(chordPitches) return str(chordList) @property def startMeasureNumber(self): ''' The measure number on which the contiguous segment begins. ''' if self.segment: return self.segment[0].measureNumber else: return None @property def startOffset(self): ''' The offset of the beginning of the contiguous segment, with respect to the measure containing the first note. ''' if self.segment: return self.segment[0].offset else: return None def getActiveMatchedRows(self): ''' Returns two ToneRow objects, the activeSegment as ToneRow and the matchedSegment as ToneRow ''' if isinstance(self.activeSegment, ToneRow): activeRow = self.activeSegment else: activeRow = pcToToneRow(self.activeSegment) if isinstance(self.matchedSegment, ToneRow): matchedRow = self.matchedSegment else: matchedRow = pcToToneRow(self.matchedSegment) return (activeRow, matchedRow) @property def zeroCenteredTransformationsFromMatched(self): ''' The list of zero-centered transformations taking a segment being searched for to a found segment, for example, in :func:`~music21.search.serial.findTransformedSegments`. For an explanation of the zero-centered convention for serial transformations, see :meth:`music21.search.serial.ToneRow.zeroCenteredTransformation`. ''' (activeRow, matchedRow) = self.getActiveMatchedRows() return matchedRow.findZeroCenteredTransformations(activeRow) @property def originalCenteredTransformationsFromMatched(self): ''' The list of original-centered transformations taking a segment being searched for to a found segment, for example, in :func:`~music21.search.serial.findTransformedSegments`. For an explanation of the zero-centered convention for serial transformations, see :meth:`music21.search.serial.ToneRow.originalCenteredTransformation`. ''' (activeRow, matchedRow) = self.getActiveMatchedRows() return matchedRow.findOriginalCenteredTransformations(activeRow) def readPitchClassesFromBottom(self): ''' Returns the list of pitch classes in the segment, reading pitches within chords from bottom to top. >>> sc = stream.Score() >>> n1 = note.Note('d4') >>> n1.quarterLength = 1 >>> Cmaj = chord.Chord(['c5', 'e4', 'g4']) >>> Cmaj.quarterLength = 1 >>> sc.append(n1) >>> sc.append(Cmaj) >>> sc = sc.makeMeasures() >>> searcher = search.serial.ContiguousSegmentSearcher(sc) >>> segmentList = searcher.byLength(4) >>> csn = segmentList[0] >>> csn.readPitchClassesFromBottom() [2, 0, 4, 7] ''' seg = self.segment pitchClasses = [] for noteOrChord in seg: for p in noteOrChord.pitches: pitchClasses.append(p.pitchClass) return pitchClasses def getDistinctPitchClasses(self): ''' Returns a list of distinct pitch classes in the segment, in order of appearance, where pitches in a chord are read from bottom to top. Does not sort or order. >>> sc = stream.Score() >>> n1 = note.Note('d4') >>> n1.quarterLength = 1 >>> c = chord.Chord(['d4', 'e4', 'g4', 'd5']) >>> c.quarterLength = 1 >>> sc.append(n1) >>> sc.append(c) >>> sc = sc.makeMeasures() >>> searcher = search.serial.ContiguousSegmentSearcher(sc) >>> segmentList = searcher.byLength(4) >>> csn = segmentList[0] >>> csn.getDistinctPitchClasses() [2, 4, 7] ''' seg = self.segment pitchClasses = [] for noteOrChord in seg: for p in noteOrChord.pitches: if p.pitchClass not in pitchClasses: pitchClasses.append(p.pitchClass) return pitchClasses class ContiguousSegmentSearcher: ''' Class that when given a :class:`~music21.stream.Stream` and `.byLength()` is called, returns a :class:`~music21.search.serial.ContiguousSegmentOfNotes` objects where the desired number of notes in the segment is specified. The inputStream is a Score or Part or Opus containing one score.. Furthermore, all notes must be contained within measures. The reps argument specifies how repeated pitch classes are dealt with. It may be set to 'skipConsecutive' (default), 'rowsOnly', 'includeAll', or 'ignoreAll'. These are explained in detail below. The includeChords argument specifies how chords are dealt with. When set to True (default), the pitches of all chords are read in order from bottom to top, and when set to False, all segments containing chords are ignored. The main subtleties of this function lie in how each reps setting works in conjunction with chords when `includeChords` is set to `True`, and how the lengths of the segments are measured. However, let us first examine what happens when includeChords is set to `False`, to get an idea of how the function works. To begin, we create a stream on which we will apply the function. >>> s = stream.Stream() >>> n1 = note.Note('e4') >>> n1.quarterLength = 6 >>> s.append(n1) >>> n2 = note.Note('f4') >>> n2.quarterLength = 1 >>> s.append(n2) >>> n3 = chord.Chord(['g4', 'b4']) >>> n3.quarterLength = 1 >>> s.append(n3) >>> n4 = note.Note('g4') >>> n4.quarterLength = 1 >>> s.repeatAppend(n4, 2) >>> n5 = note.Note('a4') >>> n5.quarterLength = 3 >>> s.repeatAppend(n5, 2) >>> n6 = note.Note('b4') >>> n6.quarterLength = 1 >>> s.append(n6) >>> n7 = note.Note('c5') >>> n7.quarterLength = 1 >>> s.append(n7) We can now try to apply this function: >>> searcher = search.serial.ContiguousSegmentSearcher(s, 'skipConsecutive', False) >>> contiguousList = searcher.byLength(3) >>> print(contiguousList) [, ] .. image:: images/serial-findTransposedSegments.png :width: 500 We now can apply the function, and in doing so we examine in detail each of the reps settings. 'skipConsecutive' means that whenever immediate repetitions of notes or chords occur, only the first instance of the note or chord is included in the segment. The durations of the repeated notes, do not have to be the same. >>> searcher = search.serial.ContiguousSegmentSearcher(s, includeChords=False) >>> searcher.reps = 'skipConsecutive' >>> skipConsecutiveList = searcher.byLength(3) >>> print(skipConsecutiveList) [, ] >>> [instance.segment for instance in skipConsecutiveList] [[, , ], [, , ]] In order to be considered repetition, the spellings of the notes in question must be exactly the same: enharmonic equivalents are not checked and notes with the same pitch in different octaves are considered different. To illustrate this, see the example below, in which all three notes, with pitch class 0, are considered separately. >>> new = stream.Stream() >>> n1 = note.Note('c4') >>> n2 = note.Note('c5') >>> n3 = note.Note('b#5') >>> new.append(n1) >>> new.append(n2) >>> new.append(n3) >>> new = new.makeMeasures() >>> searcherNew = search.serial.ContiguousSegmentSearcher(new, ... reps='skipConsecutive', includeChords=False) >>> foundSegments = searcherNew.byLength(3) >>> [seg.segment for seg in foundSegments] [[, , ]] 'rowsOnly' searches only for tone rows, in which all pitch classes in the segment must be distinct. Below, we are looking for sequences three consecutive notes within the stream `s`, all of which have different pitch classes. There is only one such set of notes, and by calling the :attr:`~music21.search.serial.ContiguousSegmentOfNotes` we can determine its location (the measure number of its first note). We'll return to our original searcher, but make sure that it has measures now: >>> searcher = search.serial.ContiguousSegmentSearcher(s.makeMeasures(), includeChords=False) >>> searcher.reps = 'rowsOnly' >>> rowsOnlyList = searcher.byLength(3) >>> [(instance.segment, instance.startMeasureNumber) for instance in rowsOnlyList] [([, , ], 4)] 'includeAll' disregards all repetitions, and simply gets all contiguous segments of the specified length (still subject to the includeChords setting). >>> searcher.reps = 'includeAll' >>> includeAllList = searcher.byLength(3) >>> for instance in includeAllList: ... print(instance.segment, instance.startMeasureNumber, instance.startOffset) [, , ] 3 0.0 [, , ] 3 1.0 [, , ] 3 2.0 [, , ] 4 1.0 Note that there only two total As appear in these segments, despite there being three :class:`~music21.note.Note` objects with the A4 as the pitch in the stream s; this is because only the first note of each set of tied notes is considered. This convention applies to this function and all parsing functions below. Also note that so far, neither of the first two notes n1, n2 nor the major third n3 in `s` have been included in any of the returned contiguous segments. This is because for each of these, any instance of three consecutive notes or chords contains the chord n3. This phenomenon also applies to the next example below. Finally, when includeChords is set to False, 'ignoreAll' finds all contiguous segments containing exactly three distinct pitch classes within it. It is unique in that unlike the previous three reps settings, the segments returned in fact have more than the number of notes specified (3). Rather, they each have 3 distinct pitch classes, and some pitch classes may be repeated. >>> searcher.reps = 'ignoreAll' >>> ignoreAllList = searcher.byLength(3) >>> [instance.segment for instance in ignoreAllList] [[, , , , ], [, , , ], [, , , ], [, , ]] Let us now examine what happens in the default chord setting, in which includeChords is set to True. There are two points to remember when considering chords: the first is that all chords are read as sequences of single notes, from bottom to top. The second is that 'length' always applies to the total number of single pitches or pitch classes found in the segment, including within chords, and not to the number of notes or chords. However, as we will see, when we search for contiguous segments of length 4, the returned segments may not have exactly 4 total notes (possibly existing as single notes or within chords), a natural point of confusion. Below is a new stream s0. >>> s0 = stream.Stream() >>> n1 = note.Note('d4') >>> maj2nd = chord.Chord(['f4', 'g4']) >>> bMaj1 = chord.Chord(['b4', 'd#5', 'f#5']) >>> bMaj2 = chord.Chord(['b4', 'd#5', 'f#5']) >>> n2 = note.Note('f#4') >>> n3 = note.Note('e4') >>> n4 = note.Note('a4') >>> s0.append([n1, maj2nd, bMaj1, bMaj2, n2, n3, n4]) >>> s0 = s0.makeMeasures() >>> #_DOCS_SHOW s0.show() .. image:: images/serial-getContiguousSegmentsOfLength2.png :width: 500 >>> searcher = search.serial.ContiguousSegmentSearcher(s0, 'skipConsecutive', True) >>> skipConsecutiveWithChords = searcher.byLength(4) >>> [seg.segment for seg in skipConsecutiveWithChords] [[, , ], [, ], [, ], [, , ], [, , , ]] Let us look closely at the found segments. First, because reps was set to 'skipConsecutive', the second B major chord (bMaj2) is never considered, as the chord right before it is the same. As was mentioned before, not all the segments found have exactly 4 notes total. This is because, for each segment, only a subset of the notes contained in the first and last elements are read. Given one of the found segments, it will always be possible to extract exactly four consecutive pitches from the notes and chords, reading in order, so that at least one pitch is taken from each of the first and last chords. In the first segment, there is one way to extract 4 consecutive pitches: we take the D in the first note, read the F and G (in that order) from the next chord, and finally, reading the last chord from bottom to top, the B from the B major chord. Note that no other reading of the segment is possible because the D from the first note must be used. The second segment in the returned list, on the other hand, can be read as a sequence of 4 consecutive pitches in two ways, both equally valid. We can either take the top note of the first chord, and all three notes, in order, of the second chord, or both notes of the first chord and the bottom two notes of the second chord. >>> searcher.reps = 'rowsOnly' >>> rowsOnlyChords = searcher.byLength(4) >>> rowsOnlyChords [, ] >>> [seg.segment for seg in rowsOnlyChords] [[, , ], [, ]] When reps is set to 'rowsOnly', the segments returned are those such that each may be read as a sequence of 4 pitches, in the same manner as explained above with the 'skipConsecutive' setting, such that the sequence of 4 pitches constitutes a four-note tone row. Above, the first segment corresponds to the row [2, 5, 7, 11], and the second may be read as either [5, 7, 11, 3] or [7, 11, 3, 6]. Note that, for example, we could not include both the B-major chord and the F# that comes right after it in the same segment, because there would have to be two consecutive instances of the pitch class 6 (corresponding to F#). Similarly, we could not include both instances of the B-major chord, as, again, we would have a pitch class repeated in any resulting four-note row. >>> searcher.reps = 'includeAll' >>> includeAll = searcher.byLength(4) >>> [seg.segment for seg in includeAll] [[, , ], [, ], [, ], [, ], [, , ], [, , , ]] Here, all segments from which sequences of four consecutive pitches can be extracted, again with at least one pitch coming from each of the first and last elements of the segments, are found. >>> searcher.reps = 'ignoreAll' >>> ignoreAll = searcher.byLength(4) >>> [seg.segment for seg in ignoreAll] [[, , ], [, ], [, , ], [, , , ], [, , , ], [, , ], [, , , ]] When reps is set to 'ignoreAll', the pitch classes from each segment are read by taking, in order, the pitch classes in the order in which they first appear, where chords are again read from bottom to top. For example, in the last segment, the first three pitch classes are those in the first chord, from bottom to top: 11, 3, and 6. Then, the next pitch class appearing is 6, which is disregarded because it has already appeared. Finally, the pitch classes 4 and 9 appear in that order. There are thus five pitch classes in this segment, in the order [11, 3, 6, 4, 9]. The segment can be read has having length 4 because four consecutive pitch classes, [3, 6, 4, 9], can be read from this sequence in such a way that the first pitch class of this subsequence is part of the first chord in the segment, and the last pitch class is that of the last note of the segment. More generally, in this setting the found segments are those which contain at least 4 distinct pitch classes, but the top note of the first chord (or note), the bottom note of the last chord (or note), and all pitches of all notes and chords other than the first and last contain at most 4 distinct pitch classes. OMIT_FROM_DOCS >>> import copy >>> sc = stream.Score(id='outerScore') >>> p = stream.Part(id='toBeCloned') >>> n1 = note.Note('f4') >>> n2 = note.Note('e4') >>> c1 = chord.Chord(['c5', 'd5']) >>> c2 = chord.Chord(['c4', 'd4']) >>> p.append([n1, n2, c1, c2]) >>> p = p.makeMeasures() >>> p1 = copy.deepcopy(p) >>> p1.id = 'clone1' >>> sc.insert(0.0, p1) >>> p2 = copy.deepcopy(p) >>> p2.id = 'clone2' >>> sc.insert(0.0, p2) >>> searcherNew = search.serial.ContiguousSegmentSearcher(sc, 'ignoreAll') >>> allSegments = searcherNew.byLength(3) >>> [seg.segment for seg in allSegments] [[, , ], [, ], [, , ], [, , ], [, ], [, , ]] ''' def __init__(self, inputStream=None, reps='skipConsecutive', includeChords=True): self.stream = inputStream self.reps = reps self.includeChords = includeChords self.searchLength = 1 self.currentNote = None self.partNumber = None self.chordList = [] # contains Chord or Note objects self.activeChordList = [] # can also be Note objects. self.totalLength = 0 self.listOfContiguousSegments = [] # for ignoreAll, this will reduce the number # of possibilities much faster if True self.trimToShortestLengthFast = False def getSearchBoundMethod(self): ''' Return a search method based on the setting of reps (how to classify repetitions), and the includeChord setting. ''' reps = self.reps if self.includeChords is False: if reps == 'skipConsecutive': return self.searchSkipConsecutiveExclude elif reps == 'rowsOnly': return self.searchRowsOnlyExclude elif reps == 'includeAll': return self.searchIncludeAllExclude elif reps == 'ignoreAll': return self.searchIgnoreAllExclude else: if reps == 'skipConsecutive': return self.searchSkipConsecutiveInclude elif reps == 'rowsOnly': return self.searchRowsOnlyInclude elif reps == 'includeAll': return self.searchIncludeAllInclude elif reps == 'ignoreAll': return self.searchIgnoreAllInclude def byLength(self, length): ''' Run the current setting for reps and includeChords to find all segments of length `length`. ''' self.searchLength = length self.listOfContiguousSegments = [] hasParts = True partList = self.stream[stream.Part] if not partList: partList = [self.stream] hasParts = False searchMethod = self.getSearchBoundMethod() self.listOfContiguousSegments = [] for partNumber, partObj in enumerate(partList): if hasParts is False: partNumber = None # self.chordList = [] self.totalLength = 0 # counts each pitch within a chord once for n in partObj.recurse().notes: if n.tie is not None and n.tie.type != 'start': continue searchMethod(n, partNumber) return self.listOfContiguousSegments def addActiveChords(self, partNumber): csn = ContiguousSegmentOfNotes(self.activeChordList, self.stream, partNumber) self.listOfContiguousSegments.append(csn) return csn def searchIncludeAllExclude(self, n, partNumber): if len(n.pitches) > 1: self.chordList = [] return False chordList = self.chordList chordList.append(n) self.totalLength = self.totalLength + len(n.pitches) if len(chordList) == self.searchLength + 1: chordList.pop(0) if len(chordList) == self.searchLength: self.activeChordList = chordList[:] self.addActiveChords(partNumber) return True return False def searchIncludeAllInclude(self, n, partNumber): ''' Returns the number added. ''' numCSNAdded = 0 chordList = self.chordList chordList.append(n) self.totalLength += len(n.pitches) lengthOfActive = self.totalLength numChordsToDelete = 0 for i in range(len(chordList)): activeChordList = chordList[i:] firstChordNumPitches = len(activeChordList[0].pitches) lastChordNumPitches = len(activeChordList[-1].pitches) if i: lengthOfActive -= len(chordList[i - 1].pitches) numPitchesMinusFirstLast = lengthOfActive - (firstChordNumPitches + lastChordNumPitches) if (lengthOfActive >= self.searchLength and numPitchesMinusFirstLast <= self.searchLength - 2): self.activeChordList = activeChordList self.addActiveChords(partNumber) numCSNAdded += 1 elif lengthOfActive >= self.searchLength: numChordsToDelete += 1 else: break for unused_counter in range(numChordsToDelete): removedChord = chordList.pop(0) self.totalLength -= len(removedChord.pitches) return numCSNAdded def searchSkipConsecutiveExclude(self, n, partNumber): chordList = self.chordList if chordList and chordList[-1].pitches == n.pitches: return False return self.searchIncludeAllExclude(n, partNumber) def searchSkipConsecutiveInclude(self, n, partNumber): chordList = self.chordList if chordList and chordList[-1].pitches == n.pitches: return False return self.searchIncludeAllInclude(n, partNumber) def searchIgnoreAllExclude(self, n, partNumber): if len(n.pitches) > 1: self.chordList = [] return False numCSNAdded = 0 numChordsToDelete = 0 chordList = self.chordList chordList.append(n) for i in range(len(chordList)): activeChordList = chordList[i:] activePitches = [] for thisChord in activeChordList: activePitches.extend(thisChord.pitches[:]) uniqueActivePitchClasses = {p.pitchClass for p in activePitches} numUniqueActivePitchClasses = len(uniqueActivePitchClasses) if numUniqueActivePitchClasses == self.searchLength: self.activeChordList = activeChordList self.addActiveChords(partNumber) if self.trimToShortestLengthFast: numChordsToDelete += 1 numCSNAdded += 1 elif numUniqueActivePitchClasses > self.searchLength: numChordsToDelete += 1 for unused_counter in range(numChordsToDelete): removedChord = chordList.pop(0) self.totalLength -= len(removedChord.pitches) return numCSNAdded def searchIgnoreAllInclude(self, n, partNumber): numCSNAdded = 0 numChordsToDelete = 0 chordList = self.chordList chordList.append(n) for i in range(len(chordList)): self.activeChordList = activeChordList = chordList[i:] csn = self.addActiveChords(partNumber) rowSuperset = set(csn.readPitchClassesFromBottom()) if len(rowSuperset) >= self.searchLength: middleSegment = ContiguousSegmentOfNotes(activeChordList[1:-1], None, None) middlePitchClassSet = set(middleSegment.readPitchClassesFromBottom()) setToCheck = middlePitchClassSet.union([activeChordList[0].pitches[-1].pitchClass, activeChordList[-1].pitches[0].pitchClass]) if (len(setToCheck)) > self.searchLength: self.listOfContiguousSegments.pop() numChordsToDelete += 1 elif self.trimToShortestLengthFast: numChordsToDelete += 1 else: self.listOfContiguousSegments.pop() break for unused_counter in range(numChordsToDelete): removedChord = chordList.pop(0) self.totalLength -= len(removedChord.pitches) return numCSNAdded def searchRowsOnlyExclude(self, n, partNumber): if len(n.pitches) > 1: self.chordList = [] return False chordList = self.chordList if len(chordList) == self.searchLength: chordList.pop(0) if n.pitch.pitchClass not in [oldN.pitch.pitchClass for oldN in chordList]: chordList.append(n) else: self.chordList = chordList = [n] # all unique if len(chordList) == self.searchLength: self.activeChordList = chordList[:] self.addActiveChords(partNumber) def searchRowsOnlyInclude(self, n, partNumber): chordList = self.chordList chordList.append(n) self.totalLength += len(n.pitches) lengthOfActive = self.totalLength numChordsToDelete = 0 for i in range(len(chordList)): activeChordList = chordList[i:] firstChordNumPitches = len(activeChordList[0].pitches) lastChordNumPitches = len(activeChordList[-1].pitches) if i: lengthOfActive -= len(chordList[i - 1].pitches) numPitchesMinusFirstLast = lengthOfActive - (firstChordNumPitches + lastChordNumPitches) if (lengthOfActive >= self.searchLength and numPitchesMinusFirstLast <= self.searchLength - 2): self.activeChordList = activeChordList csn = self.addActiveChords(partNumber) rowSuperset = csn.readPitchClassesFromBottom() lowerBound = max([0, len(rowSuperset) - self.searchLength - len(self.activeChordList[-1].pitches) + 1]) upperBound = min([len(self.activeChordList[0].pitches), len(rowSuperset) - self.searchLength + 1]) for j in range(lowerBound, upperBound): if len(set(rowSuperset[j:j + self.searchLength])) == self.searchLength: break else: # was not a match, should not have been added, # thus remove from listOfContiguousSegments self.listOfContiguousSegments.pop() elif lengthOfActive >= self.searchLength: numChordsToDelete += 1 else: break for unused_counter in range(numChordsToDelete): removedChord = chordList.pop(0) self.totalLength -= len(removedChord.pitches) class SegmentMatcher: ''' Matches all the ContiguousSegmentsOfNotes (found by ContiguousSegmentSearcher) within a :class:`~music21.stream.Stream` to one or more segments of pitch classes. The inputStream is a :class:`~music21.stream.Stream`; as in :class:`~music21.search.serial.ContiguousSegmentSearcher`, the inputStream can contain at most one :class:`~music21.stream.Score` and its notes must be contained in measures. The searchList is a list of contiguous segments to be searched for, each segment being given as a list of pitch classes. The reps and includeChords settings specify how repeated pitches and chords, respectively, are handled; the possible settings are the same as those in :class:`~music21.search.serial.ContiguousSegmentSearcher`. Returns a list of :class:`~music21.search.serial.ContiguousSegmentOfNotes` objects for which the :attr:`~music21.search.serial.ContiguousSegmentOfNotes.activeSegment` matches at least one of the elements of the searchList, subject to the settings specified in reps and includeChords. >>> sc = stream.Score() >>> part = stream.Part() >>> sig = meter.TimeSignature('2/4') >>> part.append(sig) >>> n1 = note.Note('e4') >>> n1.quarterLength = 6 >>> part.append(n1) >>> n2 = note.Note('f4') >>> n2.quarterLength = 1 >>> part.append(n2) >>> n3 = chord.Chord(['g4', 'b4']) >>> n3.quarterLength = 1 >>> part.append(n3) >>> n4 = note.Note('g4') >>> n4.quarterLength = 1 >>> part.repeatAppend(n4, 2) >>> n5 = note.Note('a4') >>> n5.quarterLength = 3 >>> part.repeatAppend(n5, 2) >>> n6 = note.Note('b4') >>> n6.quarterLength = 1 >>> part.append(n6) >>> n7 = note.Note('c5') >>> n7.quarterLength = 1 >>> part.append(n7) >>> newPart = part.makeMeasures() >>> newPart.makeTies(inPlace=True) >>> #_DOCS_SHOW newPart.show() .. image:: images/serial-findSegments.png :width: 500 >>> sc.insert(0, newPart) >>> matcher = search.serial.SegmentMatcher(sc, includeChords=False) >>> GABandABC = matcher.find([[7, 9, 11], [9, 11, 0]]) >>> print(GABandABC) [, ] >>> GABandABC[0].segment, GABandABC[1].segment ([, , ], [, , ]) >>> GABandABC[0].startMeasureNumber, GABandABC[1].startMeasureNumber (5, 6) In case it is not clear, we can use the :attr:`~music21.search.serial.ContiguousSegmentsOfNotes.matchedSegment` property to determine, to which element of the original searchList the found contiguous segments were matched. >>> GABandABC[0].matchedSegment [7, 9, 11] >>> GABandABC[1].matchedSegment [9, 11, 0] One can also search for segments of different lengths, simultaneously. Below, 'B' refers to the pitch class 11, which only coincidentally is the same as that of the note B. >>> x = (matcher.find([[7, 9, 11], ['B', 0]])) >>> x [, ] Below, we can see what happens when we include the chords and use ignoreAll >>> matcher = search.serial.SegmentMatcher(sc, reps='ignoreAll', includeChords=True) >>> [seg.segment for seg in matcher.find([[5, 7, 'B']])] [[, ]] As expected, the pitch classes found segment are read in the order 5, 7, 11 ('B'), as the pitches in the chord are read from bottom to top. Consider the following other example with chords, which is somewhat more complex: >>> sc0 = stream.Score() >>> p0 = stream.Part() >>> c1 = chord.Chord(['c4', 'd4']) >>> c2 = chord.Chord(['e4', 'f4']) >>> p0.append(c1) >>> p0.append(c2) >>> p0 = p0.makeMeasures() >>> sc0.insert(0, p0) >>> matcher = search.serial.SegmentMatcher(sc0) >>> foundSegments = matcher.find([[0, 2, 4]]) >>> len(foundSegments) 1 >>> seg = foundSegments[0] >>> seg.segment [, ] >>> seg.activeSegment [0, 2, 4] If we are just searching for a single term, the list-of-lists can be given just as a list: >>> foundSegments = matcher.find([2, 4, 5]) >>> len(foundSegments) 1 >>> seg = foundSegments[0] >>> seg.segment [, ] >>> seg.activeSegment [2, 4, 5] In the two function calls, despite the fact that two different segments of pitch classes were searched for, the same :class:`~music21.search.serial.ContiguousSegmentOfNotes` object was found for each. This is because the found object can be read in two ways as a sequence of three pitch classes: either as [0, 2, 4], by taking the two notes of the first chord in order and the bottom note of the second, or as [2, 4, 5], by taking the top note of the first chord and the two notes of the second chord in order. Both times, the chords are read from bottom to top. >>> matcher = search.serial.SegmentMatcher(sc, includeChords=False) >>> foundSegments = matcher.find([[7, -3, 11], [9, 11, 0]]) >>> for a in foundSegments: ... print(a.matchedSegment) [7, -3, 11] [9, 11, 0] >>> len(foundSegments) 2 >>> s = stream.Stream() >>> n1 = note.Note('e4') >>> n2 = note.Note('f4') >>> n3 = note.Note('g4') >>> c = chord.Chord(['b4', 'g5', 'a5']) >>> s.append(n1) >>> s.append(n2) >>> s.append(n3) >>> s.append(c) >>> s = s.makeMeasures() >>> matcher = search.serial.SegmentMatcher(s, 'ignoreAll') >>> foundSegments = matcher.find([4, -7, 7]) >>> [seg.segment for seg in foundSegments] [[, , ]] >>> foundSegments = matcher.find([7, 'B', 9]) >>> [seg.segment for seg in foundSegments] [[, ]] ''' includeMultisetDuplicates = False def __init__(self, inputStream, reps='skipConsecutive', includeChords=True): self.stream = inputStream self._reps = reps self._includeChords = includeChords self.searchedAlready = [] self.matchedSegments = [] self.currentSearchSegmentLength = 0 self._contiguousSegmentsByLength = {} @property def reps(self): return self._reps @reps.setter def reps(self, newReps): self._reps = newReps self._contiguousSegmentsByLength = {} @property def includeChords(self): ''' Returns or sets bool on whether chords should be included. Clears the segment cache when it is changed. ''' return self._includeChords @includeChords.setter def includeChords(self, newChords): self._includeChords = newChords self._contiguousSegmentsByLength = {} def getContiguousSegmentsByLength(self, searchSegmentLength): ''' Creates a ContiguousSegmentSearcher and finds all segments in self.stream based on the .reps and .includeChords settings and the searchSegmentLength If we are searching on lots of segments to match, then we could end up running this expensive routine multiple times, so we'll cache the answer for each length. If .reps or .includeChords is changed, then this will be obsolete, so cleared. ''' if searchSegmentLength in self._contiguousSegmentsByLength: theseSegments = self._contiguousSegmentsByLength[searchSegmentLength] else: searcher = ContiguousSegmentSearcher(self.stream, self.reps, self.includeChords) theseSegments = searcher.byLength(searchSegmentLength) self._contiguousSegmentsByLength[searchSegmentLength] = theseSegments return theseSegments def find(self, searchList): if not searchList: return [] elif not (common.isIterable(searchList[0])): # a single list instead of a list of lists: searchList = [searchList] self.matchedSegments = [] self.searchedAlready = [] self._contiguousSegmentsByLength = {} for unNormalizedCurrentSearchSegment in searchList: # normalize and check. if self.checkSearchedAlready(unNormalizedCurrentSearchSegment): continue currentSearchSegment = self.normalize(unNormalizedCurrentSearchSegment) searchSegmentLength = len(unNormalizedCurrentSearchSegment) theseSegments = self.getContiguousSegmentsByLength(searchSegmentLength) self.currentSearchSegmentLength = searchSegmentLength for thisSegment in theseSegments: if self.reps == 'ignoreAll': self.findOneIgnoreAll(thisSegment, currentSearchSegment, unNormalizedCurrentSearchSegment) else: self.findOneOtherReps(thisSegment, currentSearchSegment, unNormalizedCurrentSearchSegment) return self.matchedSegments def findOneIgnoreAll(self, thisSegment, searchSegment, unNormalizedCurrentSearchSegment=None): ''' Checks whether thisSegment is a match for the searchSegment if 'ignoreAll' is the search term. If so adds it to self.matchedSegments. Only matches once per segment ''' if unNormalizedCurrentSearchSegment is None: unNormalizedCurrentSearchSegment = searchSegment length = self.currentSearchSegmentLength thisChordList = thisSegment.segment pitchClassList = thisSegment.getDistinctPitchClasses() lastStartPosition = len(pitchClassList) - len(searchSegment) + 1 for i in range(lastStartPosition): pitchClassSubset = pitchClassList[i:i + length] subsetToCheck = self.normalize(pitchClassSubset) if not self.equalSubset(searchSegment, subsetToCheck): continue if pitchClassSubset[0] not in [p.pitchClass for p in thisChordList[0].pitches]: continue startSeg = thisChordList[:-1] startSegPitchClasses = ContiguousSegmentOfNotes(startSeg).readPitchClassesFromBottom() if self.includeMultisetDuplicates or pitchClassSubset[-1] not in startSegPitchClasses: thisSegment.activeSegment = subsetToCheck thisSegment.matchedSegment = list(unNormalizedCurrentSearchSegment) self.matchedSegments.append(thisSegment) break def findOneOtherReps(self, thisSegment, searchSegment, unNormalizedCurrentSearchSegment): ''' Checks whether thisSegment is a match for the searchSegment if 'ignoreAll' is NOT the search term. If so adds it to self.matchedSegments. Only matches once per segment ''' if unNormalizedCurrentSearchSegment is None: unNormalizedCurrentSearchSegment = searchSegment length = self.currentSearchSegmentLength thisChordList = thisSegment.segment pitchClassList = thisSegment.readPitchClassesFromBottom() lowerBound = max([0, len(pitchClassList) - length - len(thisChordList[-1].pitches) + 1]) upperBound = min([len(thisChordList[0].pitches), len(pitchClassList) + 1 - length]) for i in range(lowerBound, upperBound): subsetToCheck = self.normalize(pitchClassList[i:i + length]) if not self.equalSubset(searchSegment, subsetToCheck): continue thisSegment.activeSegment = subsetToCheck thisSegment.matchedSegment = list(unNormalizedCurrentSearchSegment) self.matchedSegments.append(thisSegment) break def checkSearchedAlready(self, unNormalizedSearchSegment): ''' Check to see if we have searched this segment already. Called out to be subclassible by Transformed searchers. If yes, return True. If not, add to searchedAlready and return False >>> matcher = search.serial.SegmentMatcher(None) >>> matcher.checkSearchedAlready([4, 5, 6]) False >>> matcher.checkSearchedAlready([1, 2, 3]) False >>> matcher.checkSearchedAlready([4, 5, 6]) True ''' currentSearchSegment = self.normalize(unNormalizedSearchSegment) if currentSearchSegment in self.searchedAlready: return True self.searchedAlready.append(currentSearchSegment) return False @staticmethod def normalize(segment): ''' Normalize an input segment for searching. This class just changes letters to numbers, etc. Staticmethod: >>> search.serial.SegmentMatcher.normalize([3, 4, 5]) [3, 4, 5] >>> search.serial.SegmentMatcher(None).normalize(['B', -24, '1']) [11, 0, 1] ''' return pcToToneRow(segment).pitchClasses() def equalSubset(self, searchSegment, subsetToCheck): ''' Returns True if these are equal in some way. Here's it is simple -- are they equal? But it will be harder for other classes. ''' return bool(searchSegment == subsetToCheck) class TransposedSegmentMatcher(SegmentMatcher): ''' Finds all instances of given contiguous segments of pitch classes, with transpositions, within a :class:`~music21.stream.Stream`. The inputStream is a :class:`~music21.stream.Stream`; as in :class:`~music21.search.serial.ContiguousSegmentSearcher`, the inputStream can contain at most one :class:`~music21.stream.Score` and its notes must be contained in measures. The searchList is a list of contiguous segments to be searched for, each segment being given as a list of pitch classes. The reps and includeChords settings specify how repeated pitches and chords, respectively, are handled; the possible settings are the same as those in :class:`~music21.search.serial.ContiguousSegmentSearcher` Returns a list of :class:`~music21.search.serial.ContiguousSegmentOfNotes` objects for which some transposition of the :attr:`~music21.search.serial.ContiguousSegmentOfNotes.activeSegment` matches at least one of the elements of the searchList, subject to the settings specified in reps and includeChords. >>> part = stream.Part() >>> n1 = note.Note('e4') >>> n1.quarterLength = 6 >>> part.append(n1) >>> n2 = note.Note('f4') >>> n2.quarterLength = 1 >>> part.append(n2) >>> n3 = chord.Chord(['g4', 'b4']) >>> n3.quarterLength = 1 >>> part.append(n3) >>> n4 = note.Note('g4') >>> n4.quarterLength = 1 >>> part.repeatAppend(n4, 2) >>> n5 = note.Note('a4') >>> n5.quarterLength = 3 >>> part.repeatAppend(n5, 2) >>> n6 = note.Note('b4') >>> n6.quarterLength = 1 >>> part.append(n6) >>> n7 = note.Note('c5') >>> n7.quarterLength = 1 >>> part.append(n7) >>> newPart = part.makeMeasures() >>> newPart.makeTies(inPlace=True) >>> #_DOCS_SHOW newPart.show() .. image:: images/serial-findTransposedSegments.png :width: 500 First, note that it is impossible, using the 'ignoreAll' setting, to find segments, transposed or not, with repeated pitch classes. >>> matcher = search.serial.TransposedSegmentMatcher(newPart, 'ignoreAll') >>> matcher.find([0, 0]) [] A somewhat more interesting example is below. >>> matcher = search.serial.TransposedSegmentMatcher(newPart, 'rowsOnly', ... includeChords=False) >>> halfStepList = matcher.find([0, 1]) >>> L = [step.segment for step in halfStepList] >>> print(L) [[, ], [, ]] >>> [step.startMeasureNumber for step in halfStepList] [1, 5] In addition to calling the :attr:`~music21.search.serial.ContiguousSegmentOfNotes.startMeasureNumber` property to return the measure numbers on which the half steps start, one may also call the :attr:`~music21.base.Music21Object.measureNumber` property of the first :class:`~music21.note.Note` of each segment. >>> s = stream.Stream() >>> s.repeatAppend(newPart, 2) # s has two parts, each of which is a copy of newPart. >>> sMatcher = search.serial.TransposedSegmentMatcher(s, includeChords=False) >>> wholeStepList = sMatcher.find([12, 2]) >>> [(step.segment, step.startMeasureNumber, step.partNumber) for step in wholeStepList] [([, ], 3, 0), ([, ], 3, 0), ([, ], 3, 1), ([, ], 3, 1)] Including chords works similarly as in :class:`~music21.search.serial.findSegments`. >>> sMatcher = search.serial.TransposedSegmentMatcher(newPart, includeChords=True) >>> foundSegments = sMatcher.find([4, 6, 'A']) >>> [seg.segment for seg in foundSegments] [[, ]] OMIT_FROM_DOCS >>> sMatcher = search.serial.TransposedSegmentMatcher(newPart, 'skipConsecutive', ... includeChords=False) >>> testSameSeg = sMatcher.find([(12, 13), (0, 1)]) # duplicates >>> len(testSameSeg) 2 >>> testSameSeg[0].matchedSegment [12, 13] >>> sMatcher = search.serial.TransposedSegmentMatcher(newPart, 'rowsOnly', ... includeChords=False) >>> sMatcher.find([9, 'A', 'B']) [] >>> s = stream.Stream() >>> n1 = note.Note('e4') >>> n2 = note.Note('f4') >>> n3 = note.Note('g4') >>> c = chord.Chord(['b4', 'g5', 'a5']) >>> s.append(n1) >>> s.append(n2) >>> s.append(n3) >>> s.append(c) >>> s = s.makeMeasures() >>> sMatcher = search.serial.TransposedSegmentMatcher(s, 'ignoreAll', includeChords=True) >>> sMatcher.find([3, 4, 6]) [] >>> sMatcher.find([4, 8, 6]) [] ''' @staticmethod def normalize(segment): ''' Normalize an input segment for searching. For this class changes to intervals Staticmethod: >>> search.serial.TransposedSegmentMatcher.normalize([3, 4, 5]) '11' >>> search.serial.TransposedSegmentMatcher.normalize(['12', 'B', 7]) 'E8' ''' return pcToToneRow(segment).getIntervalsAsString() class TransformedSegmentMatcher(SegmentMatcher): ''' Finds all instances of given contiguous segments of pitch classes, with serial transformations, within a :class:`~music21.stream.Stream`. The inputStream is :class:`~music21.stream.Stream`; as in :class:`~music21.search.serial.ContiguousSegmentSearcher`, the inputStream can contain at most one :class:`~music21.stream.Score` and its notes must be contained in measures. The searchList is a list of contiguous segments to be searched for, each segment being given as a list of pitch classes. The reps and includeChords settings specify how repeated pitches and chords, respectively, are handled; the possible settings are the same as those in :class:`~music21.search.serial.ContiguousSegmentSearcher`. The convention for serial transformations must be specified to either 'zero' or 'original', as described in :meth:`~music21.search.serial.ToneRow.zeroCenteredTransformation` and :func:`~music21.search.serial.ToneRow.originalCenteredTransformation` - the default setting is 'original', as to relate found segments directly to the given segments, without first transposing the given segment to begin on the pitch class 0. Returns a list of :class:`~music21.search.serial.ContiguousSegmentOfNotes` objects for which some transformation of the :attr:`~music21.search.serial.ContiguousSegmentOfNotes.activeSegment` matches at least one of the elements of the searchList, subject to the settings specified in reps and includeChords. >>> n1 = note.Note('c#4') >>> n2 = note.Note('e4') >>> n3 = note.Note('d#4') >>> n4 = note.Note('f4') >>> n5 = note.Note('e4') >>> n6 = note.Note('g4') >>> noteList = [n1, n2, n3, n4, n5, n6] >>> part = stream.Part() >>> part.append(noteList) >>> part = part.makeMeasures() >>> #_DOCS_SHOW part.show() .. image:: images/serial-findTransformedSegments.png :width: 150 >>> tsMatcher = search.serial.TransformedSegmentMatcher(part, 'rowsOnly', ... includeChords=False) >>> rowInstances = tsMatcher.find([2, 5, 4]) >>> row = [2, 5, 4] >>> len(rowInstances) 2 >>> firstInstance = rowInstances[0] >>> firstInstance >>> firstInstance.activeSegment, firstInstance.startMeasureNumber (, 1) >>> firstInstance.activeSegment.pitchClasses() [1, 4, 3] >>> firstInstance.originalCenteredTransformationsFromMatched [('T', 11)] We have thus found that the first instance of the row [2, 5, 4] within our stream appears as a transposition down a semitone, beginning in measure 1. We can do a similar analysis on the other instance of the row. >>> secondInstance = rowInstances[1] >>> secondInstance.activeSegment.pitchClasses(), secondInstance.startMeasureNumber ([5, 4, 7], 1) >>> secondInstance.zeroCenteredTransformationsFromMatched [('RI', 7)] Let us give an example of this function used with chords included and reps set to 'ignoreAll'. >>> s = stream.Stream() >>> n1 = note.Note('e4') >>> n2 = note.Note('f4') >>> n3 = note.Note('g4') >>> c = chord.Chord(['b4', 'g5', 'a5']) >>> s.append(n1) >>> s.append(n2) >>> s.append(n3) >>> s.append(c) >>> s = s.makeMeasures() >>> tsMatcher = search.serial.TransformedSegmentMatcher(s, 'ignoreAll', ... includeChords=True) >>> found643 = tsMatcher.find([6, 4, 3]) >>> [seg.segment for seg in found643] [[, , ]] >>> found684 = tsMatcher.find([6, 8, 4]) >>> for seg in found684: ... print(seg.segment) [, ] [] >>> [seg.activeSegment.pitchClasses() for seg in found684] [[7, 11, 9], [11, 7, 9]] >>> [seg.originalCenteredTransformationsFromMatched for seg in found684] [[('R', 3)], [('RI', 3)]] Pitch classes are extracted from segments in order of appearance, with pitches in chords being read from bottom to top. However, only the first instance of each pitch class is considered, as seen in the :attr:`~music21.search.serial.ContiguousSegmentOfNotes.activeSegment` calls. As long as the first and last pitch classes in the active segment first appear in the first and last elements of the found segment, respectively, the segment will be matched to the segment being searched for. To make this more clear, consider the following example in the same stream s: >>> tsMatcher = search.serial.TransformedSegmentMatcher(s, 'includeAll') >>> found = tsMatcher.find([4, 0, 4]) >>> [(seg.segment, seg.activeSegment.pitchClasses()) for seg in found] [([, ], [7, 11, 7])] Above, the pitch classes of the found segment are read in the order 7, 11, 7, 9. Because a subsequence of this, [7, 11, 7], is an inversion of the search segment, [4, 0, 4], and furthermore, the first 7 is part of the first note of the segment (G), and the last 7 is part of the last chord of the segment, the found segment is matched to the segment being searched for. OMIT_FROM_DOCS >>> s = stream.Stream() >>> s.repeatAppend(part, 2) >>> tsMatcher = search.serial.TransformedSegmentMatcher(s, includeChords=False) >>> testNegativePitchClass = tsMatcher.find([2, -7, 4]) >>> len(testNegativePitchClass) 4 >>> testNegativePitchClass[0].matchedSegment [2, -7, 4] ''' def checkSearchedAlready(self, unNormalizedSearchSegment): ''' Here a segment is returned as searchedAlready if it is a transformation of a previous search segment. >>> transMatcher = search.serial.TransformedSegmentMatcher(None) >>> transMatcher.checkSearchedAlready([0, 1, 2]) False >>> transMatcher.checkSearchedAlready([0, 1, 2]) True >>> transMatcher.checkSearchedAlready([0, 1, 3]) False >>> transMatcher.checkSearchedAlready([3, 5, 6]) # RI of 0, 1, 3 True ''' segmentRow = self.normalize(unNormalizedSearchSegment) for usedRow in self.searchedAlready: if self.getTransformations(segmentRow, usedRow): return True self.searchedAlready.append(segmentRow) return False @staticmethod def normalize(segment): ''' Normalize an input segment for searching. For this class changes to intervals Staticmethod: >>> search.serial.TransformedSegmentMatcher.normalize([3, 4, 5]) >>> search.serial.TransformedSegmentMatcher.normalize(['12', 'B', 7]) ''' return pcToToneRow(segment) @staticmethod def getTransformations(row1, row2): ''' Returns a list of transformations that transform row1 into row2 Staticmethod: >>> TSM = search.serial.TransformedSegmentMatcher >>> TSM.getTransformations(TSM.normalize([3, 4, 5]), TSM.normalize([4, 5, 6])) [('P', 3), ('RI', 5)] >>> TSM.getTransformations(TSM.normalize(['12', 'B', 7]), TSM.normalize([0, 1, 5])) [('I', 0)] ''' return row2.findZeroCenteredTransformations(row1) def equalSubset(self, searchSegment, subsetToCheck): ''' Returns True if these are equal in some way. If is case two rows are equal if they have a zeroCenteredTransformation of the other. >>> TSM = search.serial.TransformedSegmentMatcher(None) >>> TSM.equalSubset(TSM.normalize([3, 4, 5]), TSM.normalize([4, 5, 6])) True >>> TSM.equalSubset(TSM.normalize([0, 1, 2]), TSM.normalize([0, 1, 3])) False ''' return bool(self.getTransformations(searchSegment, subsetToCheck)) class MultisetSegmentMatcher(SegmentMatcher): ''' Finds all instances of given multisets of pitch classes within a :class:`~music21.stream.Stream`. A multiset is a generalization of a set, in which the order of the elements in the multiset does not matter, but multiple instances of the same thing (in this case, same pitch class) are treated as distinct elements. Thus, two multisets of pitch classes are considered to be equal if and only if the number of times any given pitch class appears in one multiset is the same as the number of times the pitch class appears in the other multiset. The inputStream is :class:`~music21.stream.Stream`; as in :class:`~music21.search.serial.ContiguousSegmentSearcher`, the inputStream can contain at most one :class:`~music21.stream.Score` its notes must be contained in measures. However, the inputStream may have multiple parts. The searchList is a list of multisets to be searched for, each multiset being given as a list of pitch classes. Note that the order of pitch classes given in a multiset does not matter. The reps and includeChords settings specify how repeated pitches and chords, respectively, are handled; the possible settings are the same as those in :class:`~music21.search.serial.ContiguousSegmentSearcher` Returns a list of :class:`~music21.search.serial.ContiguousSegmentOfNotes` objects for the :attr:`~music21.search.serial.ContiguousSegmentOfNotes.activeSegment`, interpreted as a multiset, matches at least one of the elements of the searchList, subject to the settings specified in reps and includeChords. >>> part = stream.Part() >>> n1 = note.Note('e4') >>> n1.quarterLength = 4 >>> n2 = note.Note('e4') >>> n2.quarterLength = 4 >>> n3 = note.Note('f4') >>> n3.quarterLength = 4 >>> n4 = note.Note('e4') >>> n4.quarterLength = 4 >>> n5 = note.Note('g4') >>> n5.quarterLength = 4 >>> part.append(n1) >>> part.append(n2) >>> part.append(n3) >>> part.append(n4) >>> part.append(n5) >>> part.makeMeasures(inPlace=True) >>> part.show('text') {0.0} {0.0} {0.0} {0.0} {4.0} {0.0} {8.0} {0.0} {12.0} {0.0} {16.0} {0.0} {4.0} >>> #_DOCS_SHOW part.show() .. image:: images/serial-findMultisets.png :width: 150 Find all instances of the multiset [5, 4, 4] in the part >>> MSS = search.serial.MultisetSegmentMatcher(part, 'includeAll', includeChords=False) >>> EEF = MSS.find([5, 4, 4]) >>> EEF [, ] >>> [(seg.activeSegment, seg.startMeasureNumber) for seg in EEF] [([4, 4, 5], 1), ([4, 5, 4], 2)] >>> MSS = search.serial.MultisetSegmentMatcher(part, 'ignoreAll') >>> EF = MSS.find([5, 4]) >>> EF [, , , , ] Consider the following examples, with chords. >>> sc0 = stream.Score() >>> part0 = stream.Part() >>> part0.append(note.Note('c4')) >>> part0.append(note.Note('d4')) >>> part0.append(note.Note('e4')) >>> part0.append(chord.Chord(['f4', 'e5'])) >>> part0 = part0.makeMeasures() >>> sc0.insert(0, part0) >>> MSS = search.serial.MultisetSegmentMatcher(sc0, 'ignoreAll') >>> CDE = MSS.find([0, 4, 2]) >>> [seg.segment for seg in CDE] [[, , ]] Also: >>> sc1 = stream.Score() >>> part1 = stream.Part() >>> part1.append(note.Note('c4')) >>> part1.append(note.Note('d4')) >>> part1.append(chord.Chord(['e4', 'f4'])) >>> part1 = part1.makeMeasures() >>> sc1.insert(0, part1) >>> searcher = search.serial.ContiguousSegmentSearcher(sc1) >>> segmentList = searcher.byLength(3) >>> [seg.getDistinctPitchClasses() for seg in segmentList] [[0, 2, 4, 5], [2, 4, 5]] >>> MSS = search.serial.MultisetSegmentMatcher(sc1) >>> CDF = MSS.find([0, 2, 5]) >>> CDF [] OMIT_FROM_DOCS >>> import copy >>> s = stream.Score() >>> s.insert(0, copy.deepcopy(part)) >>> s.insert(0, copy.deepcopy(part)) >>> MSS = search.serial.MultisetSegmentMatcher(part, 'rowsOnly') >>> MSS.find([5, 4, 4]) [] >>> MSS = search.serial.MultisetSegmentMatcher(s, 'includeAll', includeChords=False) >>> testMultiple = MSS.find([[-7, 16, 4], [5, 4, 4]]) # test 5 4 4 twice >>> len(testMultiple) 4 >>> testMultiple[0].matchedSegment [-7, 16, 4] >>> sc = stream.Score() >>> part = stream.Part() >>> part.append(note.Note('c4')) >>> part.append(note.Note('d4')) >>> part.append(note.Note('e4')) >>> part.append(chord.Chord(['f4', 'e5'])) >>> part = part.makeMeasures() >>> sc.insert(0, part) >>> MSS = search.serial.MultisetSegmentMatcher(sc, 'ignoreAll') >>> CDE = MSS.find([0, 4, 2]) >>> CDE [] ''' includeMultisetDuplicates = True def equalSubset(self, searchSegment, subsetToCheck): ''' Returns True if there are the same number of each pitchClass in searchSegment as in subsetToCheck ''' searchSegmentCounter = Counter(searchSegment) subsetCounter = Counter(subsetToCheck) return bool(searchSegmentCounter == subsetCounter) class TransposedMultisetMatcher(SegmentMatcher): ''' Finds all instances of given multisets of pitch classes, with transpositions, within a :class:`~music21.stream.Stream`. A multiset is a generalization of a set, as described in :class:`~music21.search.serial.MultisetSegmentMatcher`. The inputStream is :class:`~music21.stream.Stream`; as in :class:`~music21.search.serial.ContiguousSegmentSearcher`, the inputStream can contain at most one :class:`~music21.stream.Score` and its notes must be contained in measures. The searchList is a list of multisets to be searched for, each multiset being given as a list of pitch classes. Note that the order of pitch classes given in a multiset does not matter. The reps and includeChords settings specify how repeated pitches and chords, respectively, are handled; the possible settings are the same as those in :class:`~music21.search.serial.ContiguousSegmentSearcher`. Returns a list of :class:`~music21.search.serial.ContiguousSegmentOfNotes` objects for some transposition of the :attr:`~music21.search.serial.ContiguousSegmentOfNotes.activeSegment`, interpreted as a multiset, matches at least one of the elements of the searchList, subject to the settings specified in reps and includeChords. >>> part = stream.Part() >>> n1 = note.Note('c4') >>> n2 = note.Note('c#4') >>> n3 = note.Note('d4') >>> n4 = note.Note('e4') >>> n5 = note.Note('e-4') >>> n6 = note.Note('e4') >>> n7 = note.Note('d4') >>> for n in [n1, n2, n3, n4, n5, n6, n7]: ... n.quarterLength = 2 ... part.repeatAppend(n, 2) >>> part = part.makeMeasures() >>> #_DOCS_SHOW part.show() .. image:: images/serial-findTransposedMultisets.png :width: 500 >>> TMM = search.serial.TransposedMultisetMatcher(part, includeChords=False) >>> instanceList = TMM.find([[-9, -10, -11]]) >>> for instance in instanceList: ... (instance.activeSegment, instance.startMeasureNumber, instance.matchedSegment) ([0, 1, 2], 1, [-9, -10, -11]) ([2, 4, 3], 3, [-9, -10, -11]) ([3, 4, 2], 5, [-9, -10, -11]) OMIT_FROM_DOCS >>> part2 = stream.Part() >>> n1 = chord.Chord(['c4', 'c5']) >>> n2 = chord.Chord(['c#4', 'c#5']) >>> n3 = chord.Chord(['d4', 'd5']) >>> n4 = chord.Chord(['e4', 'e5']) >>> n5 = chord.Chord(['e-4', 'e-5']) >>> n6 = chord.Chord(['e4', 'e5']) >>> n7 = chord.Chord(['d4', 'd5']) >>> for n in [n1, n2, n3, n4, n5, n6, n7]: ... n.quarterLength = 1 ... part2.append(n) >>> part2 = part2.makeMeasures() >>> TMM = search.serial.TransposedMultisetMatcher(part2, 'ignoreAll') >>> instanceList2 = TMM.find([3, 2, 1]) >>> [seg.segment for seg in instanceList2] [[, , ], [, , ], [, , , ], [, , , , ], [, , , ], [, , ]] >>> [seg.matchedSegment for seg in instanceList2] [[3, 2, 1], [3, 2, 1], [3, 2, 1], [3, 2, 1], [3, 2, 1], [3, 2, 1]] ''' includeMultisetDuplicates = True def checkSearchedAlready(self, multiset): ''' searched already uses counters. ''' searchSegmentCounter = {} for i in range(12): searchSegmentCounter = Counter([(p + i) % 12 for p in multiset]) if searchSegmentCounter in self.searchedAlready: return True self.searchedAlready.append(searchSegmentCounter) return False def equalSubset(self, searchSegment, subsetToCheck): ''' Returns True if there are the same number of each pitchClass in searchSegment as in subsetToCheck ''' subsetCounter = Counter(subsetToCheck) for i in range(12): searchSegmentCounter = Counter([(p + i) % 12 for p in searchSegment]) if bool(searchSegmentCounter == subsetCounter): return True return False class TransposedInvertedMultisetMatcher(TransposedMultisetMatcher): ''' Finds all instances of given multisets of pitch classes, with transpositions and inversions, within a :class:`~music21.stream.Stream`. A multiset is a generalization of a set, as described in :class:`~music21.search.serial.MultisetSegmentMatcher`. The inputStream is :class:`~music21.stream.Stream`; as in :class:`~music21.search.serial.ContiguousSegmentSearcher`, it can contain at most one :class:`~music21.stream.Score`, and its notes must be contained in measures. The multisetList is a list of multisets to be searched for, each multiset being given as a list of pitch classes. Note that the order of pitch classes given in a multiset does not matter. The reps and includeChords settings specify how repeated pitches and chords, respectively, are handled; the possible settings are the same as those in :class:`~music21.search.serial.ContiguousSegmentSearcher` Returns a list of :class:`~music21.search.serial.ContiguousSegmentOfNotes` objects for some transposition or inversion of the :attr:`~music21.search.serial.ContiguousSegmentOfNotes.activeSegment`, interpreted as a multiset, matches at least one of the elements of the searchList, subject to the settings specified in reps and includeChords. >>> s = stream.Stream() >>> n1 = note.Note('c4') >>> n2 = note.Note('e-4') >>> n3 = note.Note('g4') >>> n4 = note.Note('e4') >>> n5 = note.Note('c4') >>> for n in [n1, n2, n3, n4]: ... n.quarterLength = 1 ... s.append(n) >>> n5.quarterLength = 4 >>> s.append(n5) >>> s = s.makeMeasures() >>> #_DOCS_SHOW s.show() .. image:: images/serial-findTransposedAndInvertedMultisets.png :width: 150 >>> transposedMatcher = search.serial.TransposedInvertedMultisetMatcher(s, 'ignoreAll', ... includeChords=False) >>> majOrMinTriads = transposedMatcher.find([0, 3, 7]) >>> majOrMinTriads [, ] Note that if we search for both kinds, we should only find each once. >>> bothTriads = transposedMatcher.find([[0, 4, 7], [0, 3, 7]]) >>> [(maj.segment, maj.startOffset) for maj in bothTriads] [([, , ], 0.0), ([, , ], 2.0)] In both these cases, the `[0, 4, 7]` search should have matched: >>> [maj.matchedSegment for maj in bothTriads] [[0, 4, 7], [0, 4, 7]] Note that when we search for both [0, 4, 7] and [0, 3, 7], which are related to each other by the composition of an inversion and a transposition, each found segment is only matched to one of the multisets in the searchList; thus each found segment appears still appears at most once in the returned list of contiguous segments. Accordingly, calling :attr:`~music21.search.serial.ContiguousSegmentOfNotes.matchedSegment` returns only one element of the searchList for each found segment. OMIT_FROM_DOCS >>> transposedMatcher = search.serial.TransposedInvertedMultisetMatcher(s, 'rowsOnly', ... includeChords=True) >>> majAndMinTriads = transposedMatcher.find([0, 4, 7]) >>> [maj.segment for maj in majAndMinTriads] [[, , ], [, , ]] ''' includeMultisetDuplicates = True def checkSearchedAlready(self, multiset): ''' since the parent class adds to the list, we check inversions first and then return the parent class result ''' for i in range(12): searchSegmentCounter = Counter([(-1 * (p + i)) % 12 for p in multiset]) if searchSegmentCounter in self.searchedAlready: return True return super().checkSearchedAlready(multiset) def equalSubset(self, searchSegment, subsetToCheck): ''' Returns True if there are the same number of each pitchClass in searchSegment as in subsetToCheck ''' if super().equalSubset(searchSegment, subsetToCheck): return True subsetCounter = Counter(subsetToCheck) for i in range(12): searchSegmentCounter = Counter([(-1 * (p + i)) % 12 for p in searchSegment]) if bool(searchSegmentCounter == subsetCounter): return True return False def _labelGeneral(segmentsToLabel, inputStream, segmentDict, reps, includeChords, labelTransformations=False): ''' Helper function for all but one of the labelling functions below. Private because this should only be called in conjunction with one of the find(type of set of pitch classes) functions. ''' parts = {} if not inputStream.getElementsByClass(stream.Score): bigContainer = inputStream else: bigContainer = inputStream.getElementsByClass(stream.Score) if not bigContainer.getElementsByClass(stream.Part): hasParts = False else: parts = bigContainer.getElementsByClass(stream.Part) hasParts = True segmentList = [segmentDict[label] for label in segmentDict] labelList = list(segmentDict) # dicts are ordered as officially in 3.7 numSearchSegments = len(segmentList) numSegmentsToLabel = len(segmentsToLabel) reorderedSegmentsToLabel = sorted(segmentsToLabel, key=attrgetter( 'partNumber', 'startMeasureNumber', 'startOffset')) for k in range(numSegmentsToLabel): foundSegment = reorderedSegmentsToLabel[k] lineLabel = spanner.Line(foundSegment.segment[0], foundSegment.segment[-1]) if hasParts: parts[foundSegment.partNumber].insert(0, lineLabel) else: bigContainer.insert(0, lineLabel) rowToMatch = foundSegment.matchedSegment for searchSegmentIndex in range(numSearchSegments): if segmentList[searchSegmentIndex] != rowToMatch: continue label = labelList[searchSegmentIndex] firstNote = foundSegment.segment[0] # for labelTransformedSegments if labelTransformations is False: transformations = [] elif labelTransformations == 'original': transformations = foundSegment.originalCenteredTransformationsFromMatched elif labelTransformations == 'zero': transformations = foundSegment.zeroCenteredTransformationsFromMatched else: transformations = [] for trans in transformations: label = label + ' ,' + str(trans[0]) + str(trans[1]) if label not in [lyr.text for lyr in firstNote.lyrics]: firstNote.addLyric(label) break return inputStream def labelSegments(inputStream, segmentDict, reps='skipConsecutive', includeChords=True): ''' Labels all instances of a given collection of segments of pitch classes in a :class:`~music21.stream.Stream`. The segmentDict is a dictionary whose keys are names of the segments to be searched for, and whose values are the segments of pitch classes. The values will be turned in to a segmentList, as in :func:`~music21.search.serial.findSegments`. All other settings are as in :func:`~music21.search.serial.findSegments` as well. Returns a deepcopy of the inputStream with a :class:`~music21.spanner.Line` connecting the first and last notes of each found segment, and the first note of each found segment labeled with a :class:`~music21.note.Lyric`, the label being the key corresponding to the segment of pitch classes. One should make sure not to call this function with too large of a segmentDict, as a note being contained in too many segments will result in some spanners not showing. >>> part = stream.Part() >>> sig = meter.TimeSignature('2/4') >>> part.append(sig) >>> n1 = note.Note('e4') >>> n1.quarterLength = 6 >>> part.append(n1) >>> n2 = note.Note('f4') >>> n2.quarterLength = 1 >>> part.append(n2) >>> n3 = chord.Chord(['g4', 'b4']) >>> n3.quarterLength = 1 >>> part.append(n3) >>> n4 = note.Note('g4') >>> n4.quarterLength = 1 >>> part.repeatAppend(n4, 2) >>> n5 = note.Note('a4') >>> n5.quarterLength = 3 >>> part.repeatAppend(n5, 2) >>> n6 = note.Note('b4') >>> n6.quarterLength = 1 >>> part.append(n6) >>> n7 = note.Note('c5') >>> n7.quarterLength = 1 >>> part.append(n7) >>> newPart = part.makeMeasures() >>> newPart.makeTies(inPlace=True) We can then label the segment of pitch classes [7, 9, 11], which corresponds to a G, followed by an A, followed by a B. Let us call this segment "GAB". >>> labelGAB = search.serial.labelSegments(newPart, {'GAB':[7, 9, 11]}, ... includeChords=False) >>> #_DOCS_SHOW labelGAB.show() .. image:: images/serial-labelSegments.png :width: 500 >>> len(labelGAB.getElementsByClass(spanner.Line)) 1 ''' streamCopy = copy.deepcopy(inputStream) segmentList = [segmentDict[label] for label in segmentDict] segmentsToLabel = SegmentMatcher(streamCopy, reps, includeChords).find(segmentList) return _labelGeneral(segmentsToLabel, streamCopy, segmentDict, reps, includeChords) def labelTransposedSegments(inputStream, segmentDict, reps='skipConsecutive', includeChords=True): ''' Labels all instances of a given collection of segments of pitch classes, with transpositions, in a :class:`~music21.stream.Stream`. The segmentDict is a dictionary whose keys are names of the segments to be searched for, and whose values are the segments of pitch classes. The values will be turned in to a segmentList, as in :func:`~music21.search.serial.findTransposedSegments`. All other settings are as in :func:`~music21.search.serial.findTransposedSegments` as well. Returns a deep copy of the inputStream with a :class:`~music21.spanner.Line` connecting the first and last notes of each found segment, and the first note of each found segment labeled with a :class:`~music21.note.Lyric`, the label being the key corresponding to the segment of pitch classes. One should make sure not to call this function with too large of a segmentDict, as a note being contained in too many segments will result in some spanners not showing. >>> part = stream.Part() >>> n1 = note.Note('e4') >>> n1.quarterLength = 6 >>> part.append(n1) >>> n2 = note.Note('f4') >>> n2.quarterLength = 1 >>> part.append(n2) >>> n3 = chord.Chord(['g4', 'b4']) >>> n3.quarterLength = 1 >>> part.append(n3) >>> n4 = note.Note('g4') >>> n4.quarterLength = 1 >>> part.repeatAppend(n4, 2) >>> n5 = note.Note('a4') >>> n5.quarterLength = 3 >>> part.repeatAppend(n5, 2) >>> n6 = note.Note('b4') >>> n6.quarterLength = 1 >>> part.append(n6) >>> n7 = note.Note('c5') >>> n7.quarterLength = 1 >>> part.append(n7) >>> newPart = part.makeMeasures() >>> newPart.makeTies(inPlace=True) We have a soprano line; let us now form a bass line. >>> bass = stream.Part() >>> n8 = note.Note('c3') >>> n8.quarterLength = 4 >>> bass.append(n8) >>> r1 = note.Rest() >>> r1.quarterLength = 4 >>> bass.append(r1) >>> n9 = note.Note('b2') >>> n9.quarterLength = 4 >>> bass.append(n9) >>> r2 = note.Rest() >>> r2.quarterLength = 4 >>> bass.append(r2) >>> n10 = note.Note('c3') >>> n10.quarterLength = 4 >>> bass.append(n10) >>> newBass = bass.makeMeasures() >>> sc = stream.Score() >>> import copy >>> sc.insert(0, copy.deepcopy(newPart)) >>> sc.insert(0, copy.deepcopy(newBass)) >>> labeledSC = search.serial.labelTransposedSegments(sc, {'half':[0, 1]}, 'rowsOnly') >>> #_DOCS_SHOW labeledSC.show() .. image:: images/serial-labelTransposedSegments.png :width: 500 OMIT_FROM_DOCS >>> len(labeledSC.parts[0].getElementsByClass(spanner.Line)) 2 ''' streamCopy = inputStream.coreCopyAsDerivation('labelTransposedSegments') segmentList = [segmentDict[label] for label in segmentDict] segmentsToLabel = TransposedSegmentMatcher(streamCopy, reps, includeChords).find(segmentList) return _labelGeneral(segmentsToLabel, streamCopy, segmentDict, reps, includeChords) def labelTransformedSegments(inputStream, segmentDict, reps='skipConsecutive', includeChords=True, convention='original'): ''' Labels all instances of a given collection of segments of pitch classes, with transformations, in a :class:`~music21.stream.Stream`. The segmentDict is a dictionary whose keys are names of the segments to be searched for, and whose values are the segments of pitch classes. The values will be turned in to a segmentList, as in :func:`~music21.search.serial.findTransposedSegments`. The last argument specifies the convention ('zero' or 'original') used for naming serial transformations, as explained in :meth:`~music21.search.serial.ToneRow.zeroCenteredTransformation` and :meth:`~music21.search.serial.ToneRow.originalCenteredTransformation`. All other settings are as in :func:`~music21.search.serial.findTransposedSegments` as well. Returns a deep copy of the inputStream with a :class:`~music21.spanner.Line` connecting the first and last notes of each found segment, and the first note of each found segment labeled with a :class:`~music21.note.Lyric`, the label being the key corresponding to the segment of pitch classes. One should make sure not to call this function with too large of a segmentDict, as a note being contained in too many segments will result in some spanners not showing. >>> c1 = chord.Chord(['c#4', 'e4']) >>> c2 = chord.Chord(['d#4', 'f4']) >>> c3 = chord.Chord(['e4', 'g4']) >>> chordList = [c1, c2, c3] >>> part = stream.Part() >>> for c in chordList: ... c.quarterLength = 4 ... part.append(c) >>> part = part.makeMeasures() >>> labeledPart = search.serial.labelTransformedSegments(part, {'row':[2, 5, 4]}) >>> #_DOCS_SHOW labeledPart.show() .. image:: images/serial-labelTransformedSegments.png :width: 500 Note: the spanners above were moved manually so that they can be more easily distinguished from one another. OMIT_FROM_DOCS >>> [len(n.lyrics) for n in labeledPart.flatten().notes] [1, 1, 0] ''' streamCopy = copy.deepcopy(inputStream) segmentList = [segmentDict[label] for label in segmentDict] segmentsToLabel = TransformedSegmentMatcher(streamCopy, reps, includeChords).find(segmentList) return _labelGeneral( segmentsToLabel, streamCopy, segmentDict, reps, includeChords, labelTransformations=convention ) def labelMultisets(inputStream, multisetDict, reps='skipConsecutive', includeChords=True): ''' Labels all instances of a given collection of multisets of pitch classes in a :class:`~music21.stream.Stream`. A multiset is a generalization of a set, as described in :class:`~music21.search.serial.MultisetSegmentMatcher`. The multisetDict is a dictionary whose keys are names of the multisets to be searched for, and whose values are the segments of pitch classes. The values will be turned in to a segmentList, as in :class:`~music21.search.serial.MultisetSegmentMatcher`. All other settings are as in :class:`~music21.search.serial.MultisetSegmentMatcher` as well. Returns a deep copy of the inputStream with a :class:`~music21.spanner.Line` connecting the first and last notes of each found multiset, and the first note of each found multiset labeled with a :class:`~music21.note.Lyric`, the label being the key corresponding to the segment of pitch classes. One should make sure not to call this function with too large of a segmentDict, as a note being contained in too many segments will result in some spanners not showing. At the present time a relatively large number of multisets are found using the 'ignoreAll' setting, particularly when there are many repetitions of pitch classes (immediate or otherwise). As a result, it is possible that at points in the stream there will be more than six spanners active simultaneously, which may result in some spanners not showing correctly in XML format, or not at all. >>> part = stream.Part() >>> n1 = note.Note('e4') >>> n1.quarterLength = 4 >>> n2 = note.Note('e4') >>> n2.quarterLength = 4 >>> n3 = note.Note('f4') >>> n3.quarterLength = 4 >>> n4 = note.Note('e4') >>> n4.quarterLength = 4 >>> part.append(n1) >>> part.append(n2) >>> part.append(n3) >>> part.append(n4) >>> part = part.makeMeasures() >>> labeledPart = search.serial.labelMultisets(part, {'EEF':[4, 5, 4]}, ... reps='includeAll', includeChords=False) >>> #_DOCS_SHOW labeledPart.show() .. image:: images/serial-labelMultisets.png :width: 500 Note: the spanners above were moved manually so that they can be more easily distinguished from one another. OMIT_FROM_DOCS >>> [len(n.lyrics) for n in labeledPart.flatten().notes] [1, 1, 0, 0] ''' streamCopy = copy.deepcopy(inputStream) segmentList = [multisetDict[label] for label in multisetDict] segmentsToLabel = MultisetSegmentMatcher(streamCopy, reps, includeChords).find(segmentList) return _labelGeneral(segmentsToLabel, streamCopy, multisetDict, reps, includeChords) def labelTransposedMultisets(inputStream, multisetDict, reps='skipConsecutive', includeChords=True): ''' Labels all instances of a given collection of multisets, with transpositions, of pitch classes in a :class:`~music21.stream.Stream`. A multiset is a generalization of a set, as described in :class:`~music21.search.serial.MultisetSegmentMatcher`. The multisetDict is a dictionary whose keys are names of the multisets to be searched for, and whose values are the segments of pitch classes. The values will be turned in to a segmentList, as in :class:`~music21.search.serial.MultisetSegmentMatcher`. All other settings are as in :func:`~music21.search.serial.findTransposedMultisets` as well. Returns a deep copy of the inputStream with a :class:`~music21.spanner.Line` connecting the first and last notes of each found multiset, and the first note of each found multiset labeled with a :class:`~music21.note.Lyric`, the label being the key corresponding to the segment of pitch classes. One should make sure not to call this function with too large of a segmentDict, as a note being contained in too many segments will result in some spanners not showing. At the present time a relatively large number of multisets are found using the 'ignoreAll' setting, particularly when there are many repetitions of pitch classes (immediate or otherwise). As a result, it is possible that at points in the stream there will be more than six spanners active simultaneously, which may result in some spanners not showing correctly in XML format, or not at all. As a diversion, instead of using this tool on atonal music, let us do so on Bach. We can label all instances of three of the same pitch classes occurring in a row in one of the chorales. We learn the obvious - it appears that the alto section would be the most bored while performing this chorale. >>> bach = corpus.parse('bach/bwv57.8') >>> bachLabeled = search.serial.labelTransposedMultisets(bach, ... {'x3': [0, 0, 0]}, ... reps='includeAll', ... includeChords=False) >>> #_DOCS_SHOW bachLabeled.show() .. image:: images/serial-labelTransposedMultisets.png :width: 500 Note: the spanners above were moved manually so that they can be more easily distinguished from one another. ''' streamCopy = copy.deepcopy(inputStream) segmentList = [multisetDict[label] for label in multisetDict] segmentsToLabel = TransposedMultisetMatcher(streamCopy, reps, includeChords).find(segmentList) return _labelGeneral(segmentsToLabel, streamCopy, multisetDict, reps, includeChords) def labelTransposedAndInvertedMultisets(inputStream, multisetDict, reps='skipConsecutive', includeChords=True): ''' Labels all instances of a given collection of multisets, with transpositions and inversions, of pitch classes in a :class:`~music21.stream.Stream`. A multiset is a generalization of a set, as described in :class:`~music21.search.serial.MultisetSegmentMatcher`. The multisetDict is a dictionary whose keys are names of the multisets to be searched for, and whose values are the segments of pitch classes. The values will be turned in to a segmentList, as in :class:`~music21.search.serial.MultisetSegmentMatcher`. All other settings are as in :func:`~music21.search.serial.findTransposedMultisets` as well. Returns a deep copy of the inputStream with a :class:`~music21.spanner.Line` connecting the first and last notes of each found multiset, and the first note of each found multiset labeled with a :class:`~music21.note.Lyric`, the label being the key corresponding to the segment of pitch classes. One should make sure not to call this function with too large of a segmentDict, as a note being contained in too many segments will result in some spanners not showing. At the present time a relatively large number of multisets are found using the 'ignoreAll' setting, particularly when there are many repetitions of pitch classes (immediate or otherwise). As a result, it is possible that at points in the stream there will be more than six spanners active simultaneously, which may result in some spanners not showing correctly in XML format, or not at all. >>> s = stream.Stream() >>> n1 = note.Note('c4') >>> n2 = note.Note('e-4') >>> n3 = note.Note('g4') >>> n4 = note.Note('e4') >>> n5 = note.Note('c4') >>> for n in [n1, n2, n3, n4]: ... n.quarterLength = 1 ... s.append(n) >>> n5.quarterLength = 4 >>> s.append(n5) >>> s = s.makeMeasures() >>> l = search.serial.labelTransposedAndInvertedMultisets >>> #_DOCS_SHOW l(s, {'triad':[0, 4, 7]}, includeChords=False).show() .. image:: images/serial-labelTransposedAndInvertedMultisets.png :width: 500 Note: the spanners above were moved manually so that they can be more easily distinguished from one another. ''' streamCopy = copy.deepcopy(inputStream) segmentList = [multisetDict[label] for label in multisetDict] segmentsToLabel = TransposedInvertedMultisetMatcher(streamCopy, reps, includeChords).find(segmentList) return _labelGeneral(segmentsToLabel, streamCopy, multisetDict, reps, includeChords) # -------------------------------------------------------------------- class Test(unittest.TestCase): pass # ------------------------------------------------------------------------------ # define presented order in documentation _DOC_ORDER = [ 'ContiguousSegmentSearcher', 'ContiguousSegmentOfNotes', 'SegmentMatcher', 'TransposedSegmentMatcher', 'TransformedSegmentMatcher', 'MultisetSegmentMatcher', 'TransposedMultisetMatcher', 'TransposedInvertedMultisetMatcher', 'labelSegments', 'labelTransposedSegments', 'labelTransformedSegments', 'labelMultisets', 'labelTransposedMultisets', 'labelTransposedAndInvertedMultisets' ] if __name__ == '__main__': import music21 music21.mainTest(Test)