try: import metaarray HAVE_METAARRAY = True except: HAVE_METAARRAY = False from pyqtgraph.Qt import QtCore from GraphicsObject import GraphicsObject from PlotCurveItem import PlotCurveItem from ScatterPlotItem import ScatterPlotItem import numpy as np import scipy import pyqtgraph.functions as fn class PlotDataItem(GraphicsObject): """GraphicsItem for displaying plot curves, scatter plots, or both.""" sigPlotChanged = QtCore.Signal(object) sigClicked = QtCore.Signal(object) def __init__(self, *args, **kargs): """ There are many different ways to create a PlotDataItem: Data initialization: (x,y data only) =================================== ====================================== PlotDataItem(xValues, yValues) x and y values may be any sequence (including ndarray) of real numbers PlotDataItem(yValues) y values only -- x will be automatically set to range(len(y)) PlotDataItem(x=xValues, y=yValues) x and y given by keyword arguments PlotDataItem(ndarray(Nx2)) numpy array with shape (N, 2) where x=data[:,0] and y=data[:,1] =================================== ====================================== Data initialization: (x,y data AND may include spot style) =========================== ========================================= PlotDataItem(recarray) numpy array with dtype=[('x', float), ('y', float), ...] PlotDataItem(list-of-dicts) [{'x': x, 'y': y, ...}, ...] PlotDataItem(dict-of-lists) {'x': [...], 'y': [...], ...} PlotDataItem(MetaArray) 1D array of Y values with X sepecified as axis values OR 2D array with a column 'y' and extra columns as needed. =========================== ========================================= Line style keyword ========== ================================================ pen pen to use for drawing line between points. Default is solid grey, 1px width. Use None to disable line drawing. shadowPen pen for secondary line to draw behind the primary line. disabled by default. fillLevel fill the area between the curve and fillLevel fillBrush fill to use when fillLevel is specified ========== ================================================ Point style keyword arguments: ============ ================================================ symbol symbol to use for drawing points OR list of symbols, one per point. Default is no symbol. options are o, s, t, d, + symbolPen outline pen for drawing points OR list of pens, one per point symbolBrush brush for filling points OR list of brushes, one per point symbolSize diameter of symbols OR list of diameters pxMode (bool) If True, then symbolSize is specified in pixels. If False, then symbolSize is specified in data coordinates. ============ ================================================ Optimization keyword arguments: ========== ================================================ identical spots are all identical. The spot image will be rendered only once and repeated for every point decimate (int) decimate data ========== ================================================ Meta-info keyword arguments: ========== ================================================ name name of dataset. This would appear in a legend ========== ================================================ """ GraphicsObject.__init__(self) self.setFlag(self.ItemHasNoContents) self.xData = None self.yData = None self.xDisp = None self.yDisp = None #self.curves = [] #self.scatters = [] self.curve = PlotCurveItem() self.scatter = ScatterPlotItem() self.curve.setParentItem(self) self.scatter.setParentItem(self) #self.clear() self.opts = { 'fftMode': False, 'logMode': [False, False], 'downsample': False, 'alphaHint': 1.0, 'alphaMode': False, 'pen': (200,200,200), 'shadowPen': None, 'fillLevel': None, 'fillBrush': None, 'symbol': None, 'symbolSize': 10, 'symbolPen': (200,200,200), 'symbolBrush': (50, 50, 150), 'identical': False, } self.setData(*args, **kargs) def implements(self, interface=None): ints = ['plotData'] if interface is None: return ints return interface in ints def boundingRect(self): return QtCore.QRectF() ## let child items handle this def setAlpha(self, alpha, auto): self.opts['alphaHint'] = alpha self.opts['alphaMode'] = auto self.setOpacity(alpha) #self.update() def setFftMode(self, mode): self.opts['fftMode'] = mode self.xDisp = self.yDisp = None self.updateItems() def setLogMode(self, mode): self.opts['logMode'] = mode self.xDisp = self.yDisp = None self.updateItems() def setPointMode(self, mode): self.opts['pointMode'] = mode self.update() def setPen(self, *args, **kargs): """ | Sets the pen used to draw lines between points. | *pen* can be a QPen or any argument accepted by :func:`pyqtgraph.mkPen() ` """ pen = fn.mkPen(*args, **kargs) self.opts['pen'] = pen #self.curve.setPen(pen) #for c in self.curves: #c.setPen(pen) #self.update() self.updateItems() def setShadowPen(self, *args, **kargs): """ | Sets the shadow pen used to draw lines between points (this is for enhancing contrast or emphacizing data). | This line is drawn behind the primary pen (see :func:`setPen() `) and should generally be assigned greater width than the primary pen. | *pen* can be a QPen or any argument accepted by :func:`pyqtgraph.mkPen() ` """ pen = fn.mkPen(*args, **kargs) self.opts['shadowPen'] = pen #for c in self.curves: #c.setPen(pen) #self.update() self.updateItems() def setFillBrush(self, *args, **kargs): brush = fn.mkBrush(*args, **kargs) self.opts['fillBrush'] = brush self.updateItems() def setBrush(self, *args, **kargs): return self.setFillBrush(*args, **kargs) def setFillLevel(self, level): self.opts['fillLevel'] = level self.updateItems() def setSymbol(self, symbol): self.opts['symbol'] = symbol #self.scatter.setSymbol(symbol) self.updateItems() def setSymbolPen(self, *args, **kargs): pen = fn.mkPen(*args, **kargs) self.opts['symbolPen'] = pen #self.scatter.setSymbolPen(pen) self.updateItems() def setSymbolBrush(self, *args, **kargs): brush = fn.mkBrush(*args, **kargs) self.opts['symbolBrush'] = brush #self.scatter.setSymbolBrush(brush) self.updateItems() def setSymbolSize(self, size): self.opts['symbolSize'] = size #self.scatter.setSymbolSize(symbolSize) self.updateItems() def setDownsampling(self, ds): if self.opts['downsample'] != ds: self.opts['downsample'] = ds self.xDisp = self.yDisp = None self.updateItems() def setData(self, *args, **kargs): """ Clear any data displayed by this item and display new data. See :func:`__init__() ` for details; it accepts the same arguments. """ #self.clear() y = None x = None if len(args) == 1: data = args[0] dt = dataType(data) if dt == 'empty': return elif dt == 'listOfValues': y = np.array(data) elif dt == 'Nx2array': x = data[:,0] y = data[:,1] elif dt == 'recarray' or dt == 'dictOfLists': if 'x' in data: x = np.array(data['x']) if 'y' in data: y = np.array(data['y']) elif dt == 'listOfDicts': if 'x' in data[0]: x = np.array([d.get('x',None) for d in data]) if 'y' in data[0]: y = np.array([d.get('y',None) for d in data]) elif dt == 'MetaArray': y = data.view(np.ndarray) x = data.xvals(0).view(np.ndarray) else: raise Exception('Invalid data type %s' % type(data)) elif len(args) == 2: seq = ('listOfValues', 'MetaArray') if dataType(args[0]) not in seq or dataType(args[1]) not in seq: raise Exception('When passing two unnamed arguments, both must be a list or array of values. (got %s, %s)' % (str(type(args[0])), str(type(args[1])))) if not isinstance(args[0], np.ndarray): x = np.array(args[0]) else: x = args[0].view(np.ndarray) if not isinstance(args[1], np.ndarray): y = np.array(args[1]) else: y = args[1].view(np.ndarray) if 'x' in kargs: x = kargs['x'] if 'y' in kargs: y = kargs['y'] ## pull in all style arguments. ## Use self.opts to fill in anything not present in kargs. ## if symbol pen/brush are given with no symbol, then assume symbol is 'o' if 'symbol' not in kargs and ('symbolPen' in kargs or 'symbolBrush' in kargs or 'symbolSize' in kargs): kargs['symbol'] = 'o' if 'brush' in kargs: kargs['fillBrush'] = kargs['brush'] for k in self.opts.keys(): if k in kargs: self.opts[k] = kargs[k] #curveArgs = {} #for k in ['pen', 'shadowPen', 'fillLevel', 'brush']: #if k in kargs: #self.opts[k] = kargs[k] #curveArgs[k] = self.opts[k] #scatterArgs = {} #for k,v in [('symbolPen','pen'), ('symbolBrush','brush'), ('symbol','symbol')]: #if k in kargs: #self.opts[k] = kargs[k] #scatterArgs[v] = self.opts[k] if y is None: return if y is not None and x is None: x = np.arange(len(y)) if isinstance(x, list): x = np.array(x) if isinstance(y, list): y = np.array(y) self.xData = x.view(np.ndarray) ## one last check to make sure there are no MetaArrays getting by self.yData = y.view(np.ndarray) self.xDisp = None self.yDisp = None self.updateItems() view = self.getViewBox() if view is not None: view.itemBoundsChanged(self) ## inform view so it can update its range if it wants self.sigPlotChanged.emit(self) def updateItems(self): #for c in self.curves+self.scatters: #if c.scene() is not None: #c.scene().removeItem(c) curveArgs = {} for k,v in [('pen','pen'), ('shadowPen','shadowPen'), ('fillLevel','fillLevel'), ('fillBrush', 'brush')]: curveArgs[v] = self.opts[k] scatterArgs = {} for k,v in [('symbolPen','pen'), ('symbolBrush','brush'), ('symbol','symbol'), ('symbolSize', 'size')]: scatterArgs[v] = self.opts[k] x,y = self.getData() if curveArgs['pen'] is not None or (curveArgs['brush'] is not None and curveArgs['fillLevel'] is not None): self.curve.setData(x=x, y=y, **curveArgs) self.curve.show() else: self.curve.hide() #curve = PlotCurveItem(x=x, y=y, **curveArgs) #curve.setParentItem(self) #self.curves.append(curve) if scatterArgs['symbol'] is not None: self.scatter.setData(x=x, y=y, **scatterArgs) self.scatter.show() else: self.scatter.hide() #sp = ScatterPlotItem(x=x, y=y, **scatterArgs) #sp.setParentItem(self) #self.scatters.append(sp) def getData(self): if self.xData is None: return (None, None) if self.xDisp is None: nanMask = np.isnan(self.xData) | np.isnan(self.yData) if any(nanMask): x = self.xData[~nanMask] y = self.yData[~nanMask] else: x = self.xData y = self.yData ds = self.opts['downsample'] if ds > 1: x = x[::ds] #y = resample(y[:len(x)*ds], len(x)) ## scipy.signal.resample causes nasty ringing y = y[::ds] if self.opts['fftMode']: f = np.fft.fft(y) / len(y) y = abs(f[1:len(f)/2]) dt = x[-1] - x[0] x = np.linspace(0, 0.5*len(x)/dt, len(y)) if self.opts['logMode'][0]: x = np.log10(x) if self.opts['logMode'][1]: y = np.log10(y) self.xDisp = x self.yDisp = y #print self.yDisp.shape, self.yDisp.min(), self.yDisp.max() #print self.xDisp.shape, self.xDisp.min(), self.xDisp.max() return self.xDisp, self.yDisp def dataBounds(self, ax, frac=1.0): (x, y) = self.getData() if x is None or len(x) == 0: return (0, 0) if ax == 0: d = x elif ax == 1: d = y if frac >= 1.0: return (np.min(d), np.max(d)) elif frac <= 0.0: raise Exception("Value for parameter 'frac' must be > 0. (got %s)" % str(frac)) else: return (scipy.stats.scoreatpercentile(d, 50 - (frac * 50)), scipy.stats.scoreatpercentile(d, 50 + (frac * 50))) def clear(self): #for i in self.curves+self.scatters: #if i.scene() is not None: #i.scene().removeItem(i) #self.curves = [] #self.scatters = [] self.xData = None self.yData = None self.xDisp = None self.yDisp = None self.curve.setData([]) self.scatter.setData([]) def appendData(self, *args, **kargs): pass def dataType(obj): if hasattr(obj, '__len__') and len(obj) == 0: return 'empty' if isSequence(obj): first = obj[0] if isinstance(obj, np.ndarray): if HAVE_METAARRAY and isinstance(obj, metaarray.MetaArray): return 'MetaArray' if obj.ndim == 1: if obj.dtype.names is None: return 'listOfValues' else: return 'recarray' elif obj.ndim == 2 and obj.dtype.names is None and obj.shape[1] == 2: return 'Nx2array' else: raise Exception('array shape must be (N,) or (N,2); got %s instead' % str(obj.shape)) elif isinstance(first, dict): return 'listOfDict' else: return 'listOfValues' elif isinstance(obj, dict): return 'dict' def isSequence(obj): return isinstance(obj, list) or isinstance(obj, np.ndarray) #class TableData: #""" #Class for presenting multiple forms of tabular data through a consistent interface. #May contain: #- numpy record array #- list-of-dicts (all dicts are _not_ required to have the same keys) #- dict-of-lists #- dict (single record) #Note: if all the values in this record are lists, it will be interpreted as multiple records #Data can be accessed and modified by column, by row, or by value #data[columnName] #data[rowId] #data[columnName, rowId] = value #data[columnName] = [value, value, ...] #data[rowId] = {columnName: value, ...} #""" #def __init__(self, data): #self.data = data #if isinstance(data, np.ndarray): #self.mode = 'array' #elif isinstance(data, list): #self.mode = 'list' #elif isinstance(data, dict): #types = set(map(type, data.values())) ### dict may be a dict-of-lists or a single record #types -= set([list, np.ndarray]) ## if dict contains any non-sequence values, it is probably a single record. #if len(types) != 0: #self.data = [self.data] #self.mode = 'list' #else: #self.mode = 'dict' #elif isinstance(data, TableData): #self.data = data.data #self.mode = data.mode #else: #raise TypeError(type(data)) #for fn in ['__getitem__', '__setitem__']: #setattr(self, fn, getattr(self, '_TableData'+fn+self.mode)) #def originalData(self): #return self.data #def toArray(self): #if self.mode == 'array': #return self.data #if len(self) < 1: ##return np.array([]) ## need to return empty array *with correct columns*, but this is very difficult, so just return None #return None #rec1 = self[0] #dtype = functions.suggestRecordDType(rec1) ##print rec1, dtype #arr = np.empty(len(self), dtype=dtype) #arr[0] = tuple(rec1.values()) #for i in xrange(1, len(self)): #arr[i] = tuple(self[i].values()) #return arr #def __getitem__array(self, arg): #if isinstance(arg, tuple): #return self.data[arg[0]][arg[1]] #else: #return self.data[arg] #def __getitem__list(self, arg): #if isinstance(arg, basestring): #return [d.get(arg, None) for d in self.data] #elif isinstance(arg, int): #return self.data[arg] #elif isinstance(arg, tuple): #arg = self._orderArgs(arg) #return self.data[arg[0]][arg[1]] #else: #raise TypeError(type(arg)) #def __getitem__dict(self, arg): #if isinstance(arg, basestring): #return self.data[arg] #elif isinstance(arg, int): #return dict([(k, v[arg]) for k, v in self.data.iteritems()]) #elif isinstance(arg, tuple): #arg = self._orderArgs(arg) #return self.data[arg[1]][arg[0]] #else: #raise TypeError(type(arg)) #def __setitem__array(self, arg, val): #if isinstance(arg, tuple): #self.data[arg[0]][arg[1]] = val #else: #self.data[arg] = val #def __setitem__list(self, arg, val): #if isinstance(arg, basestring): #if len(val) != len(self.data): #raise Exception("Values (%d) and data set (%d) are not the same length." % (len(val), len(self.data))) #for i, rec in enumerate(self.data): #rec[arg] = val[i] #elif isinstance(arg, int): #self.data[arg] = val #elif isinstance(arg, tuple): #arg = self._orderArgs(arg) #self.data[arg[0]][arg[1]] = val #else: #raise TypeError(type(arg)) #def __setitem__dict(self, arg, val): #if isinstance(arg, basestring): #if len(val) != len(self.data[arg]): #raise Exception("Values (%d) and data set (%d) are not the same length." % (len(val), len(self.data[arg]))) #self.data[arg] = val #elif isinstance(arg, int): #for k in self.data: #self.data[k][arg] = val[k] #elif isinstance(arg, tuple): #arg = self._orderArgs(arg) #self.data[arg[1]][arg[0]] = val #else: #raise TypeError(type(arg)) #def _orderArgs(self, args): ### return args in (int, str) order #if isinstance(args[0], basestring): #return (args[1], args[0]) #else: #return args #def __iter__(self): #for i in xrange(len(self)): #yield self[i] #def __len__(self): #if self.mode == 'array' or self.mode == 'list': #return len(self.data) #else: #return max(map(len, self.data.values())) #def columnNames(self): #"""returns column names in no particular order""" #if self.mode == 'array': #return self.data.dtype.names #elif self.mode == 'list': #names = set() #for row in self.data: #names.update(row.keys()) #return list(names) #elif self.mode == 'dict': #return self.data.keys() #def keys(self): #return self.columnNames()