# -*- coding: utf-8 -*- # ----------------------------------------------------------------------------- # Name: stream/core.py # Purpose: mixin class for the core elements of Streams # # Authors: Michael Scott Asato Cuthbert # Christopher Ariza # # Copyright: Copyright © 2008-2024 Michael Scott Asato Cuthbert # License: BSD, see license.txt # ----------------------------------------------------------------------------- ''' the Stream Core Mixin handles the core attributes of streams that should be thought of almost as private values and not used except by advanced programmers who need the highest speed in programming. Nothing here promises to be stable. The music21 team can make any changes here for efficiency reasons while being considered backwards compatible so long as the public methods that call these methods remain stable. All functions here will eventually begin with `.core`. ''' from __future__ import annotations import copy from fractions import Fraction import typing as t import unittest from music21.base import Music21Object from music21.common.enums import OffsetSpecial from music21.common.numberTools import opFrac from music21.common.types import OffsetQL, OffsetQLSpecial, M21ObjType from music21 import spanner from music21 import tree from music21.exceptions21 import StreamException, ImmutableStreamException from music21.stream.iterator import StreamIterator, RecursiveIterator if t.TYPE_CHECKING: from music21.stream import Stream class StreamCore(Music21Object): ''' Core aspects of a Stream's behavior. Any of these can change at any time. Users are encouraged only to create stream.Stream objects. ''' def __init__(self, **keywords) -> None: super().__init__(**keywords) # hugely important -- keeps track of where the _elements are # the _offsetDict is a dictionary where id(element) is the # index and the value is a tuple of offset and element. # offsets can be floats, Fractions, or a member of the enum OffsetSpecial self._offsetDict: dict[int, tuple[OffsetQLSpecial, Music21Object]] = {} # self._elements stores Music21Object objects. self._elements: list[Music21Object] = [] # self._endElements stores Music21Objects found at # the highestTime of this Stream. self._endElements: list[Music21Object] = [] self.isSorted = True # should isFlat become readonly? self.isFlat = True # does it have no embedded Streams # someday # self._elementTree = tree.trees.ElementTree(source=self) def coreInsert( self, offset: OffsetQL, element: Music21Object, *, ignoreSort=False, setActiveSite=True ) -> bool: ''' N.B. -- a "core" method, not to be used by general users. Run .insert() instead. A faster way of inserting elements that performs no checks, just insertion. Only be used in contexts that we know we have a proper, single Music21Object. Best for usage when taking objects in a known Stream and creating a new Stream When using this method, the caller is responsible for calling Stream.coreElementsChanged after all operations are completed. Do not mix coreInsert with coreAppend operations. Returns boolean if the Stream (assuming it was sorted before) is still guaranteed to be sorted. (False doesn't mean that it's not sorted, just that we can't guarantee it.) If you don't care and plan to sort the stream later, then use `ignoreSort=True`. ''' # environLocal.printDebug(['coreInsert', 'self', self, # 'offset', offset, 'element', element]) # need to compare highest time before inserting the element in # the elements list storeSorted = False if not ignoreSort: # # if sorted and our insertion is > the highest time, then # # are still inserted # if self.isSorted is True and self.highestTime <= offset: # storeSorted = True if self.isSorted is True: ht = self.highestTime # type: ignore if ht < offset: storeSorted = True elif ht == offset: if not self._elements: storeSorted = True else: highestSortTuple = self._elements[-1].sortTuple() thisSortTuple = element.sortTuple().modify(offset=offset) if highestSortTuple < thisSortTuple: storeSorted = True self.coreSetElementOffset( element, float(offset), # why is this not opFrac? addElement=True, setActiveSite=setActiveSite ) element.sites.add(self) # need to explicitly set the activeSite of the element # will be sorted later if necessary self._elements.append(element) # self._elementTree.insert(float(offset), element) return storeSorted def coreAppend( self, element: Music21Object, *, setActiveSite=True ) -> None: ''' N.B. -- a "core" method, not to be used by general users. Run .append() instead. Low level appending; like `coreInsert` does not error check, determine elements changed, or similar operations. When using this method, the caller is responsible for calling Stream.coreElementsChanged after all operations are completed. ''' # NOTE: this is not called by append, as that is optimized # for looping multiple elements ht = self.highestTime # type: ignore self.coreSetElementOffset(element, ht, addElement=True) element.sites.add(self) # need to explicitly set the activeSite of the element if setActiveSite: self.coreSelfActiveSite(element) self._elements.append(element) # Make this faster # self._elementTree.insert(self.highestTime, element) # does not change sorted state self._setHighestTime(ht + element.duration.quarterLength) # type: ignore # -------------------------------------------------------------------------- # adding and editing Elements and Streams -- all need to call coreElementsChanged # most will set isSorted to False def coreSetElementOffset( self, element: Music21Object, offset: int|float|Fraction|OffsetSpecial, *, addElement=False, setActiveSite=True ) -> None: ''' Sets the Offset for an element, very quickly. Caller is responsible for calling :meth:`~music21.stream.core.coreElementsChanged` afterward. >>> s = stream.Stream() >>> s.id = 'Stream1' >>> n = note.Note('B-4') >>> s.insert(10, n) >>> n.offset 10.0 >>> s.coreSetElementOffset(n, 20.0) >>> n.offset 20.0 >>> n.getOffsetBySite(s) 20.0 ''' # Note: not documenting 'highestTime' is on purpose, since can only be done for # elements already stored at end. Infinite loop. try: # try first, for the general case of not OffsetSpecial. offset = opFrac(offset) # type: ignore except TypeError: if offset not in OffsetSpecial: # pragma: no cover raise StreamException(f'Cannot set offset to {offset!r} for {element}') idEl = id(element) if not addElement and idEl not in self._offsetDict: raise StreamException( f'Cannot set the offset for element {element}, not in Stream {self}.') self._offsetDict[idEl] = (offset, element) # fast if setActiveSite: self.coreSelfActiveSite(element) def coreElementsChanged( self, *, updateIsFlat: bool = True, clearIsSorted: bool = True, memo: list[int]|None = None, keepIndex: bool = False, ) -> None: ''' NB -- a "core" stream method that is not necessary for most users. This method is called automatically any time the elements in the Stream are changed. However, it may be called manually in case sites or other advanced features of an element have been modified. It was previously a private method and for most users should still be treated as such. The various arguments permit optimizing the clearing of cached data in situations when completely dropping all cached data is excessive. >>> a = stream.Stream() >>> a.isFlat True Here we manipulate the private `._elements` storage (which generally shouldn't be done) using coreAppend and thus need to call `.coreElementsChanged` directly. >>> a.coreAppend(stream.Stream()) >>> a.isFlat # this is wrong. True >>> a.coreElementsChanged() >>> a.isFlat False ''' # experimental if not getattr(self, '_mutable', True): raise ImmutableStreamException( 'coreElementsChanged should not be triggered on an immutable stream' ) if memo is None: memo = [] if id(self) in memo: return memo.append(id(self)) # WHY??? THIS SEEMS OVERKILL, esp. since the first call to .sort() in .flatten() will # invalidate it! TODO: Investigate if this is necessary and then remove if not necessary # should not need to do this # if this Stream is a flat representation of something, and its # elements have changed, than we must clear the cache of that # ancestor so that subsequent calls get a new representation of this derivation; # we can do that by calling coreElementsChanged on # the derivation.origin if self._derivation is not None: sdm = self._derivation.method if sdm in ('flat', 'semiflat'): origin: 'music21.stream.Stream' = t.cast('music21.stream.Stream', self._derivation.origin) origin.clearCache() # may not always need to clear cache of all living sites, but may # always be a good idea since .flatten() has changed etc. # should not need to do derivation.origin sites. for livingSite in self.sites: livingSite.coreElementsChanged(memo=memo) # clear these attributes for setting later if clearIsSorted: self.isSorted = False if updateIsFlat: self.isFlat = True # do not need to look in _endElements for e in self._elements: # Only need to find one case, and if so, no longer flat. # The fastest method here is isinstance() # if isinstance(e, Stream): if e.isStream: self.isFlat = False break # resetting the cache removes lowest and highest time storage # a slight performance optimization: not creating unless needed if self._cache: indexCache = None if keepIndex and 'index' in self._cache: indexCache = self._cache['index'] # always clear cache when elements have changed # for instance, Duration will change. self.clearCache() if keepIndex and indexCache is not None: self._cache['index'] = indexCache # core method that has to live in Stream itself for typing purposes. def coreCopyAsDerivation(self: M21ObjType, methodName: str, *, recurse=True, deep=True) -> M21ObjType: ''' *This is a Core method that most users will not need to use.* Make a copy of this stream with the proper derivation set. >>> s = stream.Stream() >>> n = note.Note() >>> s.append(n) >>> s2 = s.coreCopyAsDerivation('exampleCopy') >>> s2.derivation.method 'exampleCopy' >>> s2.derivation.origin is s True >>> s2[0].derivation.method 'exampleCopy' ''' from music21 import stream if deep: post = copy.deepcopy(self) else: # pragma: no cover post = copy.copy(self) post.derivation.method = methodName if recurse and deep and isinstance(post, stream.Stream): post.setDerivationMethod(methodName, recurse=True) return post def coreGetElementByMemoryLocation(self, objId): ''' NB -- a "core" stream method that is not necessary for most users. Low-level tool to get an element based only on the object id. This is not the same as getElementById, which refers to the id attribute which may be manually set and not unique. However, some implementations of python will reuse object locations, sometimes quickly, so don't keep these around. Used by spanner and variant. >>> s = stream.Stream() >>> n1 = note.Note('g') >>> n2 = note.Note('g#') >>> s.append(n1) >>> s.coreGetElementByMemoryLocation(id(n1)) is n1 True >>> s.coreGetElementByMemoryLocation(id(n2)) is None True >>> b = bar.Barline() >>> s.storeAtEnd(b) >>> s.coreGetElementByMemoryLocation(id(b)) is b True ''' # NOTE: this may be slightly faster than other approaches # as it does not sort. for e in self._elements: if id(e) == objId: return e for e in self._endElements: if id(e) == objId: return e return None # -------------------------------------------------------------------------- def coreGuardBeforeAddElement(self, element, *, checkRedundancy=True): ''' Before adding an element, this method performs important checks on that element. Used by: - :meth:`~music21.stream.Stream.insert` - :meth:`~music21.stream.Stream.append` - :meth:`~music21.stream.Stream.storeAtEnd` - `Stream.__init__()` Returns None or raises a StreamException >>> s = stream.Stream() >>> s.coreGuardBeforeAddElement(s) Traceback (most recent call last): music21.exceptions21.StreamException: this Stream cannot be contained within itself >>> s.append(s.iter()) Traceback (most recent call last): music21.exceptions21.StreamException: cannot insert StreamIterator into a Stream Iterate over it instead (User's Guide chs. 6 and 26) >>> s.insert(4, 3.14159) Traceback (most recent call last): music21.exceptions21.StreamException: The object you tried to add to the Stream, 3.14159, is not a Music21Object. Use an ElementWrapper object if this is what you intend. ''' if element is self: # cannot add this Stream into itself raise StreamException('this Stream cannot be contained within itself') if not isinstance(element, Music21Object): if isinstance(element, StreamIterator): raise StreamException( 'cannot insert StreamIterator into a Stream\n' "Iterate over it instead (User's Guide chs. 6 and 26)") raise StreamException( f'The object you tried to add to the Stream, {element!r}, ' + 'is not a Music21Object. Use an ElementWrapper object ' + 'if this is what you intend.') if checkRedundancy: # using id() here b/c we do not want to get __eq__ comparisons idElement = id(element) if idElement in self._offsetDict: # now go slow for safety -- maybe something is amiss in the index. # this should not happen, but we have slipped many times in not clearing out # old _offsetDict entries. for search_place in (self._elements, self._endElements): for eInStream in search_place: if eInStream is element: raise StreamException( f'the object ({element!r}, id()={id(element)} ' + f'is already found in this Stream ({self!r}, id()={id(self)})' ) # something was old. delete from _offsetDict # environLocal.warn('stale object') del self._offsetDict[idElement] # pragma: no cover # if we do not purge locations here, we may have ids() for # Streams that no longer exist stored in the "locations" entry for element. # Note that dead locations are also purged from .sites during # all get() calls. element.purgeLocations() def coreStoreAtEnd(self, element, setActiveSite=True): ''' NB -- this is a "core" method. General users should use .storeAtEnd() instead. Core method for adding end elements. To be called by other methods. ''' # noinspection PyTypeChecker self.coreSetElementOffset(element, OffsetSpecial.AT_END, addElement=True) element.sites.add(self) # need to explicitly set the activeSite of the element if setActiveSite: self.coreSelfActiveSite(element) # self._elements.append(element) self._endElements.append(element) @property def spannerBundle(self): ''' A low-level object for Spanner management. This is a read-only property. ''' if 'spannerBundle' not in self._cache or self._cache['spannerBundle'] is None: spanners = self.recurse(classFilter=(spanner.Spanner,), restoreActiveSites=False) self._cache['spannerBundle'] = spanner.SpannerBundle(list(spanners)) return self._cache['spannerBundle'] def asTimespans(self, *, flatten=True, classList=None): r''' Convert stream to a :class:`~music21.tree.trees.TimespanTree` instance, a highly optimized data structure for searching through elements and offsets. >>> score = tree.makeExampleScore() >>> scoreTree = score.asTimespans() >>> print(scoreTree) > > > > > > > > > > > > > > > > > > > > > ''' hashedAttributes = hash((tuple(classList or ()), flatten)) cacheKey = 'timespanTree' + str(hashedAttributes) if cacheKey not in self._cache or self._cache[cacheKey] is None: hashedTimespanTree = tree.fromStream.asTimespans(self, flatten=flatten, classList=classList) self._cache[cacheKey] = hashedTimespanTree return self._cache[cacheKey] def coreSelfActiveSite(self, el): ''' Set the activeSite of el to be self. Override for SpannerStorage, VariantStorage, which should never become the activeSite ''' el.activeSite = self def asTree(self, *, flatten=False, classList=None, useTimespans=False, groupOffsets=False): ''' Returns an elementTree of the score, using exact positioning. See tree.fromStream.asTree() for more details. >>> score = tree.makeExampleScore() >>> scoreTree = score.asTree(flatten=True) >>> scoreTree to 8.0) > ''' if t.TYPE_CHECKING: assert isinstance(self, Stream) hashedAttributes = hash((tuple(classList or ()), flatten, useTimespans, groupOffsets)) cacheKey = 'elementTree' + str(hashedAttributes) if cacheKey not in self._cache or self._cache[cacheKey] is None: hashedElementTree = tree.fromStream.asTree(self, flatten=flatten, classList=classList, useTimespans=useTimespans, groupOffsets=groupOffsets) self._cache[cacheKey] = hashedElementTree return self._cache[cacheKey] def coreGatherMissingSpanners( self, *, recurse: bool = True, requireAllPresent: bool = True, insert: bool = True, constrainingSpannerBundle: spanner.SpannerBundle|None = None ) -> list[spanner.Spanner]|None: ''' find all spanners that are referenced by elements in the (recursed if recurse=True) stream and either inserts them in the Stream (if insert is True) or returns them if insert is False. If requireAllPresent is True (default) then only those spanners whose complete spanned elements are in the Stream are returned. Because spanners are stored weakly in .sites this is only guaranteed to find the spanners in cases where the spanner is in another stream that is still active. Here's a little helper function since we'll make the same Stream several times, with two slurred notes, but without the slur itself. Python's garbage collection will get rid of the slur if we do not prevent it >>> preventGarbageCollection = [] >>> def getStream(): ... s = stream.Stream() ... n = note.Note('C') ... m = note.Note('D') ... sl = spanner.Slur(n, m) ... preventGarbageCollection.append(sl) ... s.append([n, m]) ... return s Okay now we have a Stream with two slurred notes, but without the slur. `coreGatherMissingSpanners()` will put it in at the beginning. >>> s = getStream() >>> s.show('text') {0.0} {1.0} >>> s.coreGatherMissingSpanners() >>> s.show('text') {0.0} {0.0} > {1.0} Now, the same Stream, but insert is False, so it will return a list of Spanners that should be inserted, rather than inserting them. >>> s = getStream() >>> spList = s.coreGatherMissingSpanners(insert=False) >>> spList [>] >>> s.show('text') {0.0} {1.0} Now we'll remove the second note so not all elements of the Slur are present. This, by default, will not insert the Slur: >>> s = getStream() >>> s.remove(s[-1]) >>> s.show('text') {0.0} >>> s.coreGatherMissingSpanners() >>> s.show('text') {0.0} But with `requireAllPresent=False`, the spanner appears! >>> s.coreGatherMissingSpanners(requireAllPresent=False) >>> s.show('text') {0.0} {0.0} > With `recurse=False`, then spanners are not gathered inside the inner stream: >>> part = stream.Part() >>> s = getStream() >>> part.insert(0, s) >>> part.coreGatherMissingSpanners(recurse=False) >>> part.show('text') {0.0} {0.0} {1.0} But the default acts with recursion: >>> part.coreGatherMissingSpanners() >>> part.show('text') {0.0} {0.0} {1.0} {0.0} > Spanners already in the stream are not put there again: >>> s = getStream() >>> sl = s.notes.first().getSpannerSites()[0] >>> sl > >>> s.insert(0, sl) >>> s.coreGatherMissingSpanners() >>> s.show('text') {0.0} {0.0} > {1.0} Also does not happen with recursion. >>> part = stream.Part() >>> s = getStream() >>> sl = s.notes.first().getSpannerSites()[0] >>> s.insert(0, sl) >>> part.insert(0, s) >>> part.coreGatherMissingSpanners() >>> part.show('text') {0.0} {0.0} {0.0} > {1.0} If `constrainingSpannerBundle` is set then only spanners also present in that spannerBundle are added. This can be useful, for instance, in restoring spanners from an excerpt that might already have spanners removed. In Jacob Tyler Walls's brilliant phrasing, it prevents regrowing zombie spanners that you thought you had killed. Here we will constrain only to spanners also present in another Stream: >>> s = getStream() >>> s2 = stream.Stream() >>> s.coreGatherMissingSpanners(constrainingSpannerBundle=s2.spannerBundle) >>> s.show('text') {0.0} {1.0} Now with the same constraint, but we will put the Slur into the other stream. >>> sl = s.notes.first().getSpannerSites()[0] >>> s2.insert(0, sl) >>> s.coreGatherMissingSpanners(constrainingSpannerBundle=s2.spannerBundle) >>> s.show('text') {0.0} {0.0} > {1.0} ''' sb = self.spannerBundle sIter: StreamIterator|RecursiveIterator if recurse is True: sIter = self.recurse() # type: ignore else: sIter = self.iter() # type: ignore collectList = [] for el in list(sIter): for sp in el.getSpannerSites(): if sp in sb: continue if sp in collectList: continue if constrainingSpannerBundle is not None and sp not in constrainingSpannerBundle: continue if requireAllPresent: allFound = True for spannedElement in sp.getSpannedElements(): if spannedElement not in sIter: allFound = False break if allFound is False: continue collectList.append(sp) if insert is False: return collectList if collectList: # do not run elementsChanged if nothing here. for sp in collectList: self.coreInsert(0, sp) self.coreElementsChanged(updateIsFlat=False) return None # timing before: Macbook Air 2012, i7 # In [3]: timeit('s = stream.Stream()', setup='from music21 import stream', number=100000) # Out[3]: 1.6291376419831067 # after adding subclass -- actually faster, showing the rounding error: # In [2]: timeit('s = stream.Stream()', setup='from music21 import stream', number=100000) # Out[2]: 1.5247003990225494 class Test(unittest.TestCase): pass if __name__ == '__main__': import music21 music21.mainTest(Test)