pyqtgraph/pyqtgraph/widgets/ScatterPlotWidget.py

from ..Qt import QtGui, QtCore
from .PlotWidget import PlotWidget
from .DataFilterWidget import DataFilterParameter
from .ColorMapWidget import ColorMapParameter
from .. import parametertree as ptree
from .. import functions as fn
from .. import getConfigOption
from ..graphicsItems.TextItem import TextItem
import numpy as np
from collections import OrderedDict

__all__ = ['ScatterPlotWidget']

class ScatterPlotWidget(QtGui.QSplitter):
    """
    This is a high-level widget for exploring relationships in tabular data.
        
    Given a multi-column record array, the widget displays a scatter plot of a
    specific subset of the data. Includes controls for selecting the columns to
    plot, filtering data, and determining symbol color and shape.
    
    The widget consists of four components:
    
    1) A list of column names from which the user may select 1 or 2 columns
       to plot. If one column is selected, the data for that column will be
       plotted in a histogram-like manner by using :func:`pseudoScatter()
       <pyqtgraph.pseudoScatter>`. If two columns are selected, then the
       scatter plot will be generated with x determined by the first column
       that was selected and y by the second.
    2) A DataFilter that allows the user to select a subset of the data by 
       specifying multiple selection criteria.
    3) A ColorMap that allows the user to determine how points are colored by
       specifying multiple criteria.
    4) A PlotWidget for displaying the data.
    """
    sigScatterPlotClicked = QtCore.Signal(object, object, object)
    sigScatterPlotHovered = QtCore.Signal(object, object, object)

    def __init__(self, parent=None):
        QtGui.QSplitter.__init__(self, QtCore.Qt.Orientation.Horizontal)
        self.ctrlPanel = QtGui.QSplitter(QtCore.Qt.Orientation.Vertical)
        self.addWidget(self.ctrlPanel)
        self.fieldList = QtGui.QListWidget()
        self.fieldList.setSelectionMode(self.fieldList.SelectionMode.ExtendedSelection)
        self.ptree = ptree.ParameterTree(showHeader=False)
        self.filter = DataFilterParameter()
        self.colorMap = ColorMapParameter()
        self.params = ptree.Parameter.create(name='params', type='group', children=[self.filter, self.colorMap])
        self.ptree.setParameters(self.params, showTop=False)
        
        self.plot = PlotWidget()
        self.ctrlPanel.addWidget(self.fieldList)
        self.ctrlPanel.addWidget(self.ptree)
        self.addWidget(self.plot)
        
        fg = fn.mkColor(getConfigOption('foreground'))
        fg.setAlpha(150)
        self.filterText = TextItem(border=getConfigOption('foreground'), color=fg)
        self.filterText.setPos(60,20)
        self.filterText.setParentItem(self.plot.plotItem)
        
        self.data = None
        self.indices = None
        self.mouseOverField = None
        self.scatterPlot = None
        self.selectionScatter = None
        self.selectedIndices = []
        self.style = dict(pen=None, symbol='o')
        self._visibleXY = None  # currently plotted points
        self._visibleData = None  # currently plotted records
        self._visibleIndices = None
        self._indexMap = None
        
        self.fieldList.itemSelectionChanged.connect(self.fieldSelectionChanged)
        self.filter.sigFilterChanged.connect(self.filterChanged)
        self.colorMap.sigColorMapChanged.connect(self.updatePlot)
    
    def setFields(self, fields, mouseOverField=None):
        """
        Set the list of field names/units to be processed.
        
        The format of *fields* is the same as used by 
        :func:`ColorMapWidget.setFields <pyqtgraph.widgets.ColorMapWidget.ColorMapParameter.setFields>`
        """
        self.fields = OrderedDict(fields)
        self.mouseOverField = mouseOverField
        self.fieldList.clear()
        for f,opts in fields:
            item = QtGui.QListWidgetItem(f)
            item.opts = opts
            item = self.fieldList.addItem(item)
        self.filter.setFields(fields)
        self.colorMap.setFields(fields)

    def setSelectedFields(self, *fields):
        self.fieldList.itemSelectionChanged.disconnect(self.fieldSelectionChanged)
        try:
            self.fieldList.clearSelection()
            for f in fields:
                i = list(self.fields.keys()).index(f)
                item = self.fieldList.item(i)
                item.setSelected(True)
        finally:
            self.fieldList.itemSelectionChanged.connect(self.fieldSelectionChanged)
        self.fieldSelectionChanged()

    def setData(self, data):
        """
        Set the data to be processed and displayed. 
        Argument must be a numpy record array.
        """
        self.data = data
        self.indices = np.arange(len(data))
        self.filtered = None
        self.filteredIndices = None
        self.updatePlot()
        
    def setSelectedIndices(self, inds):
        """Mark the specified indices as selected.

        Must be a sequence of integers that index into the array given in setData().
        """
        self.selectedIndices = inds
        self.updateSelected()

    def setSelectedPoints(self, points):
        """Mark the specified points as selected.

        Must be a list of points as generated by the sigScatterPlotClicked signal.
        """
        self.setSelectedIndices([pt.originalIndex for pt in points])

    def fieldSelectionChanged(self):
        sel = self.fieldList.selectedItems()
        if len(sel) > 2:
            self.fieldList.blockSignals(True)
            try:
                for item in sel[1:-1]:
                    item.setSelected(False)
            finally:
                self.fieldList.blockSignals(False)
                
        self.updatePlot()
        
    def filterChanged(self, f):
        self.filtered = None
        self.updatePlot()
        desc = self.filter.describe()
        if len(desc) == 0:
            self.filterText.setVisible(False)
        else:
            self.filterText.setText('\n'.join(desc))
            self.filterText.setVisible(True)
        
    def updatePlot(self):
        self.plot.clear()
        if self.data is None or len(self.data) == 0:
            return
        
        if self.filtered is None:
            mask = self.filter.generateMask(self.data)
            self.filtered = self.data[mask]
            self.filteredIndices = self.indices[mask]
        data = self.filtered
        if len(data) == 0:
            return
        
        colors = np.array([fn.mkBrush(*x) for x in self.colorMap.map(data)])
        
        style = self.style.copy()
        
        ## Look up selected columns and units
        sel = list([str(item.text()) for item in self.fieldList.selectedItems()])
        units = list([item.opts.get('units', '') for item in self.fieldList.selectedItems()])
        if len(sel) == 0:
            self.plot.setTitle('')
            return
        

        if len(sel) == 1:
            self.plot.setLabels(left=('N', ''), bottom=(sel[0], units[0]), title='')
            if len(data) == 0:
                return
            #x = data[sel[0]]
            #y = None
            xy = [data[sel[0]], None]
        elif len(sel) == 2:
            self.plot.setLabels(left=(sel[1],units[1]), bottom=(sel[0],units[0]))
            if len(data) == 0:
                return
            
            xy = [data[sel[0]], data[sel[1]]]
            #xydata = []
            #for ax in [0,1]:
                #d = data[sel[ax]]
                ### scatter catecorical values just a bit so they show up better in the scatter plot.
                ##if sel[ax] in ['MorphologyBSMean', 'MorphologyTDMean', 'FIType']:
                    ##d += np.random.normal(size=len(cells), scale=0.1)
                    
                #xydata.append(d)
            #x,y = xydata

        ## convert enum-type fields to float, set axis labels
        enum = [False, False]
        for i in [0,1]:
            axis = self.plot.getAxis(['bottom', 'left'][i])
            if xy[i] is not None and (self.fields[sel[i]].get('mode', None) == 'enum' or xy[i].dtype.kind in ('S', 'O')):
                vals = self.fields[sel[i]].get('values', list(set(xy[i])))
                xy[i] = np.array([vals.index(x) if x in vals else len(vals) for x in xy[i]], dtype=float)
                axis.setTicks([list(enumerate(vals))])
                enum[i] = True
            else:
                axis.setTicks(None)  # reset to automatic ticking
        
        ## mask out any nan values
        mask = np.ones(len(xy[0]), dtype=bool)
        if xy[0].dtype.kind == 'f':
            mask &= np.isfinite(xy[0])
        if xy[1] is not None and xy[1].dtype.kind == 'f':
            mask &= np.isfinite(xy[1])
        
        xy[0] = xy[0][mask]
        style['symbolBrush'] = colors[mask]
        data = data[mask]
        indices = self.filteredIndices[mask]

        ## Scatter y-values for a histogram-like appearance
        if xy[1] is None:
            ## column scatter plot
            xy[1] = fn.pseudoScatter(xy[0])
        else:
            ## beeswarm plots
            xy[1] = xy[1][mask]
            for ax in [0,1]:
                if not enum[ax]:
                    continue
                imax = int(xy[ax].max()) if len(xy[ax]) > 0 else 0
                for i in range(imax+1):
                    keymask = xy[ax] == i
                    scatter = fn.pseudoScatter(xy[1-ax][keymask], bidir=True)
                    if len(scatter) == 0:
                        continue
                    smax = np.abs(scatter).max()
                    if smax != 0:
                        scatter *= 0.2 / smax
                    xy[ax][keymask] += scatter


        if self.scatterPlot is not None:
            try:
                self.scatterPlot.sigPointsClicked.disconnect(self.plotClicked)
            except:
                pass
        
        self._visibleXY = xy
        self._visibleData = data
        self._visibleIndices = indices
        self._indexMap = None
        self.scatterPlot = self.plot.plot(xy[0], xy[1], data=data, **style)
        self.scatterPlot.sigPointsClicked.connect(self.plotClicked)
        self.scatterPlot.sigPointsHovered.connect(self.plotHovered)
        self.updateSelected()

    def updateSelected(self):
        if self._visibleXY is None:
            return
        # map from global index to visible index
        indMap = self._getIndexMap()
        inds = [indMap[i] for i in self.selectedIndices if i in indMap]
        x,y = self._visibleXY[0][inds], self._visibleXY[1][inds]

        if self.selectionScatter is not None:
            self.plot.plotItem.removeItem(self.selectionScatter)
        if len(x) == 0:
            return
        self.selectionScatter = self.plot.plot(x, y, pen=None, symbol='s', symbolSize=12, symbolBrush=None, symbolPen='y')

    def _getIndexMap(self):
        # mapping from original data index to visible point index
        if self._indexMap is None:
            self._indexMap = {j:i for i,j in enumerate(self._visibleIndices)}
        return self._indexMap

    def plotClicked(self, plot, points, ev):
        # Tag each point with its index into the original dataset
        for pt in points:
            pt.originalIndex = self._visibleIndices[pt.index()]
        self.sigScatterPlotClicked.emit(self, points, ev)

    def plotHovered(self, plot, points, ev):
        self.sigScatterPlotHovered.emit(self, points, ev)