__all__ = ['GraphicsItem'] import operator import weakref from collections import OrderedDict from functools import reduce from math import hypot from typing import Optional from xml.etree.ElementTree import Element from .. import functions as fn from ..GraphicsScene import GraphicsScene from ..Point import Point from ..Qt import QtCore, QtWidgets, isQObjectAlive # Recipe from https://docs.python.org/3.8/library/collections.html#collections.OrderedDict # slightly adapted for Python 3.7 compatibility class LRU(OrderedDict): 'Limit size, evicting the least recently looked-up key when full' def __init__(self, maxsize=128, *args, **kwds): self.maxsize = maxsize super().__init__(*args, **kwds) def __getitem__(self, key): value = super().__getitem__(key) self.move_to_end(key) return value def __setitem__(self, key, value): if key in self: self.move_to_end(key) super().__setitem__(key, value) if len(self) > self.maxsize: oldest = next(iter(self)) del self[oldest] class GraphicsItem(object): """ **Bases:** :class:`object` Abstract class providing useful methods to GraphicsObject and GraphicsWidget. (This is required because we cannot have multiple inheritance with QObject subclasses.) A note about Qt's GraphicsView framework: The GraphicsView system places a lot of emphasis on the notion that the graphics within the scene should be device independent--you should be able to take the same graphics and display them on screens of different resolutions, printers, export to SVG, etc. This is nice in principle, but causes me a lot of headache in practice. It means that I have to circumvent all the device-independent expectations any time I want to operate in pixel coordinates rather than arbitrary scene coordinates. A lot of the code in GraphicsItem is devoted to this task--keeping track of view widgets and device transforms, computing the size and shape of a pixel in local item coordinates, etc. Note that in item coordinates, a pixel does not have to be square or even rectangular, so just asking how to increase a bounding rect by 2px can be a rather complex task. """ _pixelVectorGlobalCache = LRU(100) def __init__(self): if not hasattr(self, '_qtBaseClass'): for b in self.__class__.__bases__: if issubclass(b, QtWidgets.QGraphicsItem): self.__class__._qtBaseClass = b break if not hasattr(self, '_qtBaseClass'): raise Exception('Could not determine Qt base class for GraphicsItem: %s' % str(self)) self._pixelVectorCache = [None, None] self._viewWidget = None self._viewBox = None self._connectedView = None self._exportOpts = False ## If False, not currently exporting. Otherwise, contains dict of export options. self._cachedView = None def getViewWidget(self): """ Return the view widget for this item. If the scene has multiple views, only the first view is returned. The return value is cached; clear the cached value with forgetViewWidget(). If the view has been deleted by Qt, return None. """ if self._viewWidget is None: scene = self.scene() if scene is None: return None views = scene.views() if len(views) < 1: return None self._viewWidget = weakref.ref(self.scene().views()[0]) v = self._viewWidget() if v is not None and not isQObjectAlive(v): return None return v def forgetViewWidget(self): self._viewWidget = None def getViewBox(self): """ Return the first ViewBox or GraphicsView which bounds this item's visible space. If this item is not contained within a ViewBox, then the GraphicsView is returned. If the item is contained inside nested ViewBoxes, then the inner-most ViewBox is returned. The result is cached; clear the cache with forgetViewBox() """ if self._viewBox is None: p = self while True: try: p = p.parentItem() except RuntimeError: ## sometimes happens as items are being removed from a scene and collected. return None if p is None: vb = self.getViewWidget() if vb is None: return None else: self._viewBox = weakref.ref(vb) break if hasattr(p, 'implements') and p.implements('ViewBox'): self._viewBox = weakref.ref(p) break return self._viewBox() ## If we made it this far, _viewBox is definitely not None def forgetViewBox(self): self._viewBox = None def deviceTransform(self, viewportTransform=None): """ Return the transform that converts local item coordinates to device coordinates (usually pixels). Extends deviceTransform to automatically determine the viewportTransform. """ if viewportTransform is None: view = self.getViewWidget() if view is None: return None viewportTransform = view.viewportTransform() dt = self._qtBaseClass.deviceTransform(self, viewportTransform) #xmag = abs(dt.m11())+abs(dt.m12()) #ymag = abs(dt.m21())+abs(dt.m22()) #if xmag * ymag == 0: if dt.determinant() == 0: ## occurs when deviceTransform is invalid because widget has not been displayed return None else: return dt def viewTransform(self): """Return the transform that maps from local coordinates to the item's ViewBox coordinates If there is no ViewBox, return the scene transform. Returns None if the item does not have a view.""" view = self.getViewBox() if view is None: return None if hasattr(view, 'implements') and view.implements('ViewBox'): return self.itemTransform(view.innerSceneItem())[0] else: return self.sceneTransform() #return self.deviceTransform(view.viewportTransform()) def getBoundingParents(self): """Return a list of parents to this item that have child clipping enabled.""" p = self parents = [] while True: p = p.parentItem() if p is None: break if p.flags() & self.GraphicsItemFlag.ItemClipsChildrenToShape: parents.append(p) return parents def viewRect(self): """Return the visible bounds of this item's ViewBox or GraphicsWidget, in the local coordinate system of the item.""" if self._cachedView is not None: return self._cachedView # Note that in cases of early returns here, the view cache stays empty (None). view = self.getViewBox() if view is None: return None bounds = self.mapRectFromView(view.viewRect()) if bounds is None: return None bounds = bounds.normalized() self._cachedView = bounds ## nah. #for p in self.getBoundingParents(): #bounds &= self.mapRectFromScene(p.sceneBoundingRect()) return bounds def pixelVectors(self, direction=None): """Return vectors in local coordinates representing the width and height of a view pixel. If direction is specified, then return vectors parallel and orthogonal to it. Return (None, None) if pixel size is not yet defined (usually because the item has not yet been displayed) or if pixel size is below floating-point precision limit. """ ## This is an expensive function that gets called very frequently. ## We have two levels of cache to try speeding things up. dt = self.deviceTransform() if dt is None: return None, None ## Ignore translation. If the translation is much larger than the scale ## (such as when looking at unix timestamps), we can get floating-point errors. dt.setMatrix(dt.m11(), dt.m12(), 0, dt.m21(), dt.m22(), 0, 0, 0, 1) if direction is None: direction = QtCore.QPointF(1, 0) elif direction.manhattanLength() == 0: raise Exception("Cannot compute pixel length for 0-length vector.") key = (dt.m11(), dt.m21(), dt.m12(), dt.m22(), direction.x(), direction.y()) ## check local cache if key == self._pixelVectorCache[0]: return tuple(map(Point, self._pixelVectorCache[1])) ## return a *copy* ## check global cache pv = self._pixelVectorGlobalCache.get(key, None) if pv is not None: self._pixelVectorCache = [key, pv] return tuple(map(Point,pv)) ## return a *copy* ## attempt to re-scale direction vector to fit within the precision of the coordinate system ## Here's the problem: we need to map the vector 'direction' from the item to the device, via transform 'dt'. ## In some extreme cases, this mapping can fail unless the length of 'direction' is cleverly chosen. ## Example: ## dt = [ 1, 0, 2 ## 0, 2, 1e20 ## 0, 0, 1 ] ## Then we map the origin (0,0) and direction (0,1) and get: ## o' = 2,1e20 ## d' = 2,1e20 <-- should be 1e20+2, but this can't be represented with a 32-bit float ## ## |o' - d'| == 0 <-- this is the problem. ## Perhaps the easiest solution is to exclude the transformation column from dt. Does this cause any other problems? #if direction.x() == 0: #r = abs(dt.m32())/(abs(dt.m12()) + abs(dt.m22())) ##r = 1.0/(abs(dt.m12()) + abs(dt.m22())) #elif direction.y() == 0: #r = abs(dt.m31())/(abs(dt.m11()) + abs(dt.m21())) ##r = 1.0/(abs(dt.m11()) + abs(dt.m21())) #else: #r = ((abs(dt.m32())/(abs(dt.m12()) + abs(dt.m22()))) * (abs(dt.m31())/(abs(dt.m11()) + abs(dt.m21()))))**0.5 #if r == 0: #r = 1. ## shouldn't need to do this; probably means the math above is wrong? #directionr = direction * r directionr = direction ## map direction vector onto device #viewDir = Point(dt.map(directionr) - dt.map(Point(0,0))) #mdirection = dt.map(directionr) dirLine = QtCore.QLineF(QtCore.QPointF(0,0), directionr) viewDir = dt.map(dirLine) if viewDir.length() == 0: return None, None ## pixel size cannot be represented on this scale ## get unit vector and orthogonal vector (length of pixel) #orthoDir = Point(viewDir[1], -viewDir[0]) ## orthogonal to line in pixel-space try: normView = viewDir.unitVector() #normView = viewDir.norm() ## direction of one pixel orthogonal to line normOrtho = normView.normalVector() #normOrtho = orthoDir.norm() except: raise Exception("Invalid direction %s" %directionr) ## map back to item dti = fn.invertQTransform(dt) #pv = Point(dti.map(normView)-dti.map(Point(0,0))), Point(dti.map(normOrtho)-dti.map(Point(0,0))) pv = Point(dti.map(normView).p2()), Point(dti.map(normOrtho).p2()) self._pixelVectorCache[1] = pv self._pixelVectorCache[0] = key self._pixelVectorGlobalCache[key] = pv return self._pixelVectorCache[1] def pixelLength(self, direction, ortho=False): """Return the length of one pixel in the direction indicated (in local coordinates) If ortho=True, then return the length of one pixel orthogonal to the direction indicated. Return None if pixel size is not yet defined (usually because the item has not yet been displayed). """ normV, orthoV = self.pixelVectors(direction) if normV is None or orthoV is None: return None if ortho: return orthoV.length() return normV.length() def pixelSize(self): ## deprecated v = self.pixelVectors() if v == (None, None): return None, None return (hypot(v[0].x(), v[0].y()), hypot(v[1].x(), v[1].y())) # lengths def pixelWidth(self): ## deprecated vt = self.deviceTransform() if vt is None: return 0 vt = fn.invertQTransform(vt) return vt.map(QtCore.QLineF(0, 0, 1, 0)).length() def pixelHeight(self): ## deprecated vt = self.deviceTransform() if vt is None: return 0 vt = fn.invertQTransform(vt) return vt.map(QtCore.QLineF(0, 0, 0, 1)).length() #return Point(vt.map(QtCore.QPointF(0, 1))-vt.map(QtCore.QPointF(0, 0))).length() def mapToDevice(self, obj): """ Return *obj* mapped from local coordinates to device coordinates (pixels). If there is no device mapping available, return None. """ vt = self.deviceTransform() if vt is None: return None return vt.map(obj) def mapFromDevice(self, obj): """ Return *obj* mapped from device coordinates (pixels) to local coordinates. If there is no device mapping available, return None. """ vt = self.deviceTransform() if vt is None: return None if isinstance(obj, QtCore.QPoint): obj = QtCore.QPointF(obj) vt = fn.invertQTransform(vt) return vt.map(obj) def mapRectToDevice(self, rect): """ Return *rect* mapped from local coordinates to device coordinates (pixels). If there is no device mapping available, return None. """ vt = self.deviceTransform() if vt is None: return None return vt.mapRect(rect) def mapRectFromDevice(self, rect): """ Return *rect* mapped from device coordinates (pixels) to local coordinates. If there is no device mapping available, return None. """ vt = self.deviceTransform() if vt is None: return None vt = fn.invertQTransform(vt) return vt.mapRect(rect) def mapToView(self, obj): vt = self.viewTransform() if vt is None: return None return vt.map(obj) def mapRectToView(self, obj): vt = self.viewTransform() if vt is None: return None return vt.mapRect(obj) def mapFromView(self, obj): vt = self.viewTransform() if vt is None: return None vt = fn.invertQTransform(vt) return vt.map(obj) def mapRectFromView(self, obj): vt = self.viewTransform() if vt is None: return None vt = fn.invertQTransform(vt) return vt.mapRect(obj) def pos(self): return Point(self._qtBaseClass.pos(self)) def viewPos(self): return self.mapToView(self.mapFromParent(self.pos())) def parentItem(self): return self._qtBaseClass.parentItem(self) def setParentItem(self, parent): ## Workaround for Qt bug: https://bugreports.qt-project.org/browse/QTBUG-18616 if parent is not None: pscene = parent.scene() if pscene is not None and self.scene() is not pscene: pscene.addItem(self) return self._qtBaseClass.setParentItem(self, parent) def childItems(self): return self._qtBaseClass.childItems(self) def sceneTransform(self): ## Qt bug: do no allow access to sceneTransform() until ## the item has a scene. if self.scene() is None: return self.transform() else: return self._qtBaseClass.sceneTransform(self) def transformAngle(self, relativeItem=None): """Return the rotation produced by this item's transform (this assumes there is no shear in the transform) If relativeItem is given, then the angle is determined relative to that item. """ if relativeItem is None: relativeItem = self.parentItem() tr = self.itemTransform(relativeItem)[0] vec = tr.map(QtCore.QLineF(0,0,1,0)) return vec.angleTo(QtCore.QLineF(vec.p1(), vec.p1()+QtCore.QPointF(1,0))) #def itemChange(self, change, value): #ret = self._qtBaseClass.itemChange(self, change, value) #if change == self.GraphicsItemChange.ItemParentHasChanged or change == self.ItemSceneHasChanged: #print "Item scene changed:", self #self.setChildScene(self) ## This is bizarre. #return ret #def setChildScene(self, ch): #scene = self.scene() #for ch2 in ch.childItems(): #if ch2.scene() is not scene: #print "item", ch2, "has different scene:", ch2.scene(), scene #scene.addItem(ch2) #QtWidgets.QApplication.processEvents() #print " --> ", ch2.scene() #self.setChildScene(ch2) def changeParent(self): """Called when the item's parent has changed. This method handles connecting / disconnecting from ViewBox signals to make sure viewRangeChanged works properly. It should generally be extended, not overridden.""" self._updateView() def parentChanged(self): # deprecated version of changeParent() GraphicsItem.changeParent(self) def _updateView(self): ## called to see whether this item has a new view to connect to ## NOTE: This is called from GraphicsObject.itemChange or GraphicsWidget.itemChange. if not hasattr(self, '_connectedView'): # Happens when Python is shutting down. return ## It is possible this item has moved to a different ViewBox or widget; ## clear out previously determined references to these. self.forgetViewBox() self.forgetViewWidget() ## check for this item's current viewbox or view widget view = self.getViewBox() #if view is None: ##print " no view" #return oldView = None if self._connectedView is not None: oldView = self._connectedView() if view is oldView: #print " already have view", view return ## disconnect from previous view if oldView is not None: Device = 'Device' if hasattr(oldView, 'sigDeviceRangeChanged') else '' for signal, slot in [(f'sig{Device}RangeChanged', self.viewRangeChanged), (f'sig{Device}TransformChanged', self.viewTransformChanged)]: try: getattr(oldView, signal).disconnect(slot) except (TypeError, AttributeError, RuntimeError): # TypeError and RuntimeError are from pyqt and pyside, respectively pass self._connectedView = None ## connect to new view if view is not None: #print "connect:", self, view if hasattr(view, 'sigDeviceRangeChanged'): # connect signals from GraphicsView view.sigDeviceRangeChanged.connect(self.viewRangeChanged) view.sigDeviceTransformChanged.connect(self.viewTransformChanged) else: # connect signals from ViewBox view.sigRangeChanged.connect(self.viewRangeChanged) view.sigTransformChanged.connect(self.viewTransformChanged) self._connectedView = weakref.ref(view) self.viewRangeChanged() self.viewTransformChanged() ## inform children that their view might have changed self._replaceView(oldView) self.viewChanged(view, oldView) def viewChanged(self, view, oldView): """Called when this item's view has changed (ie, the item has been added to or removed from a ViewBox)""" def _replaceView(self, oldView, item=None): if item is None: item = self for child in item.childItems(): if isinstance(child, GraphicsItem): if child.getViewBox() is oldView: child._updateView() #self._replaceView(oldView, child) else: self._replaceView(oldView, child) def viewRangeChanged(self): """ Called whenever the view coordinates of the ViewBox containing this item have changed. """ # when this is called, _cachedView is not invalidated. # this means that for functions overriding viewRangeChanged, viewRect() may be stale. def viewTransformChanged(self): """ Called whenever the transformation matrix of the view has changed. (eg, the view range has changed or the view was resized) Invalidates the viewRect cache. """ self._cachedView = None #def prepareGeometryChange(self): #self._qtBaseClass.prepareGeometryChange(self) #self.informViewBoundsChanged() def informViewBoundsChanged(self): """ Inform this item's container ViewBox that the bounds of this item have changed. This is used by ViewBox to react if auto-range is enabled. """ view = self.getViewBox() if view is not None and hasattr(view, 'implements') and view.implements('ViewBox'): view.itemBoundsChanged(self) ## inform view so it can update its range if it wants def childrenShape(self): """Return the union of the shapes of all descendants of this item in local coordinates.""" shapes = [self.mapFromItem(c, c.shape()) for c in self.allChildItems()] return reduce(operator.add, shapes) def allChildItems(self, root=None): """Return list of the entire item tree descending from this item.""" if root is None: root = self tree = [] for ch in root.childItems(): tree.append(ch) tree.extend(self.allChildItems(ch)) return tree def setExportMode(self, export, opts=None): """ This method is called by exporters to inform items that they are being drawn for export with a specific set of options. Items access these via self._exportOptions. When exporting is complete, _exportOptions is set to False. """ if opts is None: opts = {} if export: self._exportOpts = opts #if 'antialias' not in opts: #self._exportOpts['antialias'] = True else: self._exportOpts = False #def update(self): #self._qtBaseClass.update(self) #print "Update:", self def getContextMenus(self, event): return [self.getMenu()] if hasattr(self, "getMenu") else [] def generateSvg( self, nodes: dict[str, Element] ) -> Optional[tuple[Element, list[Element]]]: """Method to override to manually specify the SVG writer mechanism. Parameters ---------- nodes Dictionary keyed by the name of graphics items and the XML representation of the the item that can be written as valid SVG. Returns ------- tuple First element is the top level group for this item. The second element is a list of xml Elements corresponding to the child nodes of the item. None Return None if no XML is needed for rendering Raises ------ NotImplementedError override method to implement in subclasses of GraphicsItem See Also -------- pyqtgraph.exporters.SVGExporter._generateItemSvg The generic and default implementation """ raise NotImplementedError