Source

orange / Orange / OrangeWidgets / Visualize / OWLinProjGraph.py

Full commit
  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
from OWGraph import *
from copy import copy
import time
from operator import add
from math import *
from orngScaleLinProjData import *
import orngVisFuncts
import OWColorPalette
from OWGraphTools import UnconnectedLinesCurve

# indices in curveData
SYMBOL = 0
PENCOLOR = 1
BRUSHCOLOR = 2

LINE_TOOLTIPS = 0
VISIBLE_ATTRIBUTES = 1
ALL_ATTRIBUTES = 2

TOOLTIPS_SHOW_DATA = 0
TOOLTIPS_SHOW_SPRINGS = 1

###########################################################################################
##### CLASS : OWLINPROJGRAPH
###########################################################################################
class OWLinProjGraph(OWGraph, orngScaleLinProjData):
    def __init__(self, widget, parent = None, name = "None"):
        OWGraph.__init__(self, parent, name)
        orngScaleLinProjData.__init__(self)

        self.totalPossibilities = 0 # a variable used in optimization - tells us the total number of different attribute positions
        self.triedPossibilities = 0 # how many possibilities did we already try
        self.p = None

        self.dataMap = {}        # each key is of form: "xVal-yVal", where xVal and yVal are discretized continuous values. Value of each key has form: (x,y, HSVValue, [data vals])
        self.tooltipCurves = []
        self.tooltipMarkers   = []
        self.widget = widget

        # moving anchors manually
        self.shownAttributes = []
        self.selectedAnchorIndex = None

        self.hideRadius = 0
        self.showAnchors = 1
        self.showValueLines = 0
        self.valueLineLength = 5

        self.onlyOnePerSubset = 1
        self.showLegend = 1
        self.useDifferentSymbols = 0
        self.useDifferentColors = 1
        self.tooltipKind = 0        # index in ["Show line tooltips", "Show visible attributes", "Show all attributes"]
        self.tooltipValue = 0       # index in ["Tooltips show data values", "Tooltips show spring values"]
        self.scaleFactor = 1.0
        self.showAttributeNames = 1

        self.showProbabilities = 0
        self.squareGranularity = 3
        self.spaceBetweenCells = 1

        self.showKNN = 0   # widget sets this to 1 or 2 if you want to see correct or wrong classifications
        self.insideColors = None
        self.valueLineCurves = [{}, {}]    # dicts for x and y set of coordinates for unconnected lines

        self.enableXaxis(0)
        self.enableYLaxis(0)
        self.setAxisScale(QwtPlot.xBottom, -1.13, 1.13, 1)
        self.setAxisScale(QwtPlot.yLeft, -1.13, 1.13, 1)

    def setData(self, data, subsetData = None, **args):
        OWGraph.setData(self, data)
        orngScaleLinProjData.setData(self, data, subsetData, **args)
        #self.anchorData = []

        self.setAxisScale(QwtPlot.yLeft, -1.13, 1.13, 1)
        self.setAxisScale(QwtPlot.xBottom, -1.13, 1.13, 1)

        if data and data.domain.classVar and data.domain.classVar.varType == orange.VarTypes.Continuous:
            self.setAxisScale(QwtPlot.xBottom, -1.13, 1.13 + 0.1, 1)   # if we have a continuous class we need a bit more space on the right to show a color legend

    # ####################################################################
    # update shown data. Set labels, coloring by className ....
    def updateData(self, labels = None, setAnchors = 0, **args):
        self.removeDrawingCurves()  # my function, that doesn't delete selection curves
        #self.removeCurves()
        self.removeMarkers()
        self.tooltipMarkers = []

        self.__dict__.update(args)
        if labels == None: labels = [anchor[2] for anchor in self.anchorData]
        self.shownAttributes = labels
        self.dataMap = {}   # dictionary with keys of form "x_i-y_i" with values (x_i, y_i, color, data)
        self.valueLineCurves = [{}, {}]    # dicts for x and y set of coordinates for unconnected lines

        if not self.haveData or len(labels) < 3:
            self.anchorData = []
            self.updateLayout()
            return

        if setAnchors or (args.has_key("XAnchors") and args.has_key("YAnchors")):
            self.potentialsBmp = None
            self.setAnchors(args.get("XAnchors"), args.get("YAnchors"), labels)
            #self.anchorData = self.createAnchors(len(labels), labels)    # used for showing tooltips

        indices = [self.attributeNameIndex[anchor[2]] for anchor in self.anchorData]  # store indices to shown attributes

        # do we want to show anchors and their labels
        if self.showAnchors:
            if self.hideRadius > 0:
                xdata = self.createXAnchors(100)*(self.hideRadius / 10)
                ydata = self.createYAnchors(100)*(self.hideRadius / 10)
                self.addCurve("hidecircle", QColor(200,200,200), QColor(200,200,200), 1, style = QwtPlotCurve.Lines, symbol = QwtSymbol.NoSymbol, xData = xdata.tolist() + [xdata[0]], yData = ydata.tolist() + [ydata[0]])

            # draw dots at anchors
            shownAnchorData = filter(lambda p, r=self.hideRadius**2/100: p[0]**2+p[1]**2>r, self.anchorData)

            if not self.normalizeExamples:
                r=self.hideRadius**2/100
                for i,(x,y,a) in enumerate(shownAnchorData):
                    self.addCurve("l%i" % i, QColor(160, 160, 160), QColor(160, 160, 160), 10, style = QwtPlotCurve.Lines, symbol = QwtSymbol.NoSymbol, xData = [0, x], yData = [0, y], showFilledSymbols = 1, lineWidth=2)
                    if self.showAttributeNames:
                        self.addMarker(a, x*1.07, y*1.04, Qt.AlignCenter, bold=1)
            else:
                XAnchors = [a[0] for a in shownAnchorData]
                YAnchors = [a[1] for a in shownAnchorData]
                self.addCurve("dots", QColor(160,160,160), QColor(160,160,160), 10, style = QwtPlotCurve.NoCurve, symbol = QwtSymbol.Ellipse, xData = XAnchors, yData = YAnchors, showFilledSymbols = 1)

                # draw text at anchors
                if self.showAttributeNames:
                    for x, y, a in shownAnchorData:
                        self.addMarker(a, x*1.07, y*1.04, Qt.AlignCenter, bold = 1)

        if self.showAnchors and self.normalizeExamples:
            # draw "circle"
            xdata = self.createXAnchors(100)
            ydata = self.createYAnchors(100)
            self.addCurve("circle", QColor(Qt.black), QColor(Qt.black), 1, style = QwtPlotCurve.Lines, symbol = QwtSymbol.NoSymbol, xData = xdata.tolist() + [xdata[0]], yData = ydata.tolist() + [ydata[0]])

        self.potentialsClassifier = None # remove the classifier so that repaint won't recompute it
        self.updateLayout()

        if self.dataHasDiscreteClass:
            self.discPalette.setNumberOfColors(len(self.dataDomain.classVar.values))

        useDifferentSymbols = self.useDifferentSymbols and self.dataHasDiscreteClass and len(self.dataDomain.classVar.values) <= len(self.curveSymbols)
        dataSize = len(self.rawData)
        validData = self.getValidList(indices)
        # jitter point if Radviz
        transProjData = self.createProjectionAsNumericArray(indices, validData = validData, scaleFactor = self.scaleFactor, normalize = self.normalizeExamples, jitterSize = self.jitter_size if self.normalizeExamples else -1, useAnchorData = 1, removeMissingData = 0)
        if transProjData == None:
            return
        projData = transProjData.T
        x_positions = projData[0]
        y_positions = projData[1]
        xPointsToAdd = {}
        yPointsToAdd = {}


        if self.showProbabilities and self.haveData and self.dataHasClass:
            # construct potentialsClassifier from unscaled positions
            domain = orange.Domain([self.dataDomain[i].name for i in indices]+[self.dataDomain.classVar.name], self.dataDomain)
            offsets = [self.attrValues[self.attributeNames[i]][0] for i in indices]
            normalizers = [self.getMinMaxVal(i) for i in indices]
            selectedData = numpy.take(self.originalData, indices, axis = 0)
            averages = numpy.average(numpy.compress(validData, selectedData, axis=1), 1)
            classData = numpy.compress(validData, self.originalData[self.dataClassIndex])
            if classData.any():
                self.potentialsClassifier = orange.P2NN(domain, numpy.transpose(numpy.array([numpy.compress(validData, self.unscaled_x_positions), numpy.compress(validData, self.unscaled_y_positions), classData])), self.anchorData, offsets, normalizers, averages, self.normalizeExamples, law=1)
            else:
                self.potentialsClassifier = None
            self.potentialsImage = None


        # ##############################################################
        # show model quality
        # ##############################################################
        if self.insideColors != None or self.showKNN and self.haveData:
            # if we want to show knn classifications of the examples then turn the projection into example table and run knn
            if self.insideColors:
                insideData, stringData = self.insideColors
            else:
                shortData = self.createProjectionAsExampleTable([self.attributeNameIndex[attr] for attr in labels], useAnchorData = 1)
                predictions, probabilities = self.widget.vizrank.kNNClassifyData(shortData)
                if self.showKNN == 2: insideData, stringData = [1.0 - val for val in predictions], "Probability of wrong classification = %.2f%%"
                else:                 insideData, stringData = predictions, "Probability of correct classification = %.2f%%"

            if self.dataHasDiscreteClass:        classColors = self.discPalette
            elif self.dataHasContinuousClass:    classColors = self.contPalette

            if len(insideData) != len(self.rawData):
                j = 0
                for i in range(len(self.rawData)):
                    if not validData[i]: continue
                    if self.dataHasClass:
                        fillColor = classColors.getRGB(self.originalData[self.dataClassIndex][i], 255*insideData[j])
                        edgeColor = classColors.getRGB(self.originalData[self.dataClassIndex][i])
                    else:
                        fillColor = edgeColor = (0,0,0)
                    self.addCurve(str(i), QColor(*fillColor+ (self.alphaValue,)), QColor(*edgeColor+ (self.alphaValue,)), self.pointWidth, xData = [x_positions[i]], yData = [y_positions[i]])
                    if self.showValueLines:
                        self.addValueLineCurve(x_positions[i], y_positions[i], edgeColor, i, indices)
                    self.addTooltipKey(x_positions[i], y_positions[i], QColor(*edgeColor), i, stringData % (100*insideData[j]))
                    j+= 1
            else:
                for i in range(len(self.rawData)):
                    if not validData[i]: continue
                    if self.dataHasClass:
                        fillColor = classColors.getRGB(self.originalData[self.dataClassIndex][i], 255*insideData[i])
                        edgeColor = classColors.getRGB(self.originalData[self.dataClassIndex][i])
                    else:
                        fillColor = edgeColor = (0,0,0)
                    self.addCurve(str(i), QColor(*fillColor+ (self.alphaValue,)), QColor(*edgeColor+ (self.alphaValue,)), self.pointWidth, xData = [x_positions[i]], yData = [y_positions[i]])
                    if self.showValueLines:
                        self.addValueLineCurve(x_positions[i], y_positions[i], edgeColor, i, indices)
                    self.addTooltipKey(x_positions[i], y_positions[i], QColor(*edgeColor), i, stringData % (100*insideData[i]))

        # ##############################################################
        # do we have a subset data to show?
        # ##############################################################
        elif self.haveSubsetData:
            shownSubsetCount = 0
            subsetIdsToDraw = dict([(example.id,1) for example in self.rawSubsetData])
            subsetIdsAlreadyDrawn = set()
            # draw the rawData data set. examples that exist also in the subset data draw full, other empty
            for i in range(dataSize):
                if not validData[i]:
                    continue
                if subsetIdsToDraw.has_key(self.rawData[i].id):
                    instance_filled = 1
                    subsetIdsAlreadyDrawn.add(self.rawData[i].id)
                else:
                    instance_filled = 0

                if self.dataHasDiscreteClass and self.useDifferentColors:
                    newColor = self.discPalette.getRGB(self.originalData[self.dataClassIndex][i])
                elif self.dataHasContinuousClass and self.useDifferentColors:
                    newColor = self.contPalette.getRGB(self.noJitteringScaledData[self.dataClassIndex][i])
                else:
                    newColor = (0,0,0)

                if useDifferentSymbols:
                    curveSymbol = self.curveSymbols[int(self.originalData[self.dataClassIndex][i])]
                else:
                    curveSymbol = self.curveSymbols[0]

                if not xPointsToAdd.has_key((newColor, curveSymbol, instance_filled)):
                    xPointsToAdd[(newColor, curveSymbol, instance_filled)] = []
                    yPointsToAdd[(newColor, curveSymbol, instance_filled)] = []
                xPointsToAdd[(newColor, curveSymbol, instance_filled)].append(x_positions[i])
                yPointsToAdd[(newColor, curveSymbol, instance_filled)].append(y_positions[i])
                if self.showValueLines:
                    self.addValueLineCurve(x_positions[i], y_positions[i], newColor, i, indices)

                self.addTooltipKey(x_positions[i], y_positions[i], QColor(*newColor), i)

            # if we have a data subset that contains examples that don't exist in the original dataset we show them here
            XAnchors = numpy.array([val[0] for val in self.anchorData])
            YAnchors = numpy.array([val[1] for val in self.anchorData])
            anchorRadius = numpy.sqrt(XAnchors*XAnchors + YAnchors*YAnchors)
            validSubData = self.getValidSubsetList(indices)
            # jitter points if Radviz
            projSubData = self.createProjectionAsNumericArray(indices, validData = validSubData, scaleFactor = self.scaleFactor, normalize = self.normalizeExamples, jitterSize = self.jitter_size if self.normalizeExamples else -1, useAnchorData = 1, removeMissingData = 0, useSubsetData = 1)
            if projSubData is not None:
                projSubData = projSubData.T
                sub_x_positions = projSubData[0]
                sub_y_positions = projSubData[1]

            for i in range(len(self.rawSubsetData)):
                if not validSubData[i]: # check if has missing values
                    continue
                if self.rawSubsetData[i].id in subsetIdsAlreadyDrawn:
                    continue

                if not self.dataHasClass or self.rawSubsetData[i].getclass().isSpecial():
                    newColor = (0,0,0)
                else:
                    if self.dataHasDiscreteClass:
                        newColor = self.discPalette.getRGB(self.originalSubsetData[self.dataClassIndex][i])
                    else:
                        newColor = self.contPalette.getRGB(self.noJitteringScaledSubsetData[self.dataClassIndex][i])

                if useDifferentSymbols and self.validSubsetDataArray[self.dataClassIndex][i]:
                    curveSymbol = self.curveSymbols[int(self.originalSubsetData[self.dataClassIndex][i])]
                else:
                    curveSymbol = self.curveSymbols[0]

                if not xPointsToAdd.has_key((newColor, curveSymbol, 1)):
                    xPointsToAdd[(newColor, curveSymbol, 1)] = []
                    yPointsToAdd[(newColor, curveSymbol, 1)] = []
                xPointsToAdd[(newColor, curveSymbol, 1)].append(sub_x_positions[i])
                yPointsToAdd[(newColor, curveSymbol, 1)].append(sub_y_positions[i])

        elif not self.dataHasClass:
            xs = []; ys = []
            for i in range(dataSize):
                if not validData[i]: continue
                xs.append(x_positions[i])
                ys.append(y_positions[i])
                self.addTooltipKey(x_positions[i], y_positions[i], QColor(Qt.black), i)
                if self.showValueLines:
                    self.addValueLineCurve(x_positions[i], y_positions[i], (0,0,0), i, indices)
            self.addCurve(str(1), QColor(0,0,0,self.alphaValue), QColor(0,0,0,self.alphaValue), self.pointWidth, symbol = self.curveSymbols[0], xData = xs, yData = ys, penAlpha = self.alphaValue, brushAlpha = self.alphaValue)

        # ##############################################################
        # CONTINUOUS class
        # ##############################################################
        elif self.dataHasContinuousClass:
            for i in range(dataSize):
                if not validData[i]: continue
                newColor = self.contPalette.getRGB(self.noJitteringScaledData[self.dataClassIndex][i])
                self.addCurve(str(i), QColor(*newColor+ (self.alphaValue,)), QColor(*newColor+ (self.alphaValue,)), self.pointWidth, symbol = QwtSymbol.Ellipse, xData = [x_positions[i]], yData = [y_positions[i]])
                if self.showValueLines:
                    self.addValueLineCurve(x_positions[i], y_positions[i], newColor, i, indices)
                self.addTooltipKey(x_positions[i], y_positions[i], QColor(*newColor), i)

        # ##############################################################
        # DISCRETE class
        # ##############################################################
        elif self.dataHasDiscreteClass:
            for i in range(dataSize):
                if not validData[i]:
                    continue
                if self.useDifferentColors:
                    newColor = self.discPalette.getRGB(self.originalData[self.dataClassIndex][i])
                else:
                    newColor = (0, 0, 0)
                if useDifferentSymbols:
                    curveSymbol = self.curveSymbols[int(self.originalData[self.dataClassIndex][i])]
                else:
                    curveSymbol = self.curveSymbols[0]
                if not xPointsToAdd.has_key((newColor, curveSymbol, self.showFilledSymbols)):
                    xPointsToAdd[(newColor, curveSymbol, self.showFilledSymbols)] = []
                    yPointsToAdd[(newColor, curveSymbol, self.showFilledSymbols)] = []
                xPointsToAdd[(newColor, curveSymbol, self.showFilledSymbols)].append(x_positions[i])
                yPointsToAdd[(newColor, curveSymbol, self.showFilledSymbols)].append(y_positions[i])
                if self.showValueLines:
                    self.addValueLineCurve(x_positions[i], y_positions[i], newColor, i, indices)
                self.addTooltipKey(x_positions[i], y_positions[i], QColor(*newColor), i)

        # first draw value lines
        if self.showValueLines:
            for i, color in enumerate(self.valueLineCurves[0].keys()):
                curve = UnconnectedLinesCurve("", QPen(QColor(*color + (self.alphaValue,))), self.valueLineCurves[0][color], self.valueLineCurves[1][color])
                curve.attach(self)

        # draw all the points with a small number of curves
        for i, (color, symbol, showFilled) in enumerate(xPointsToAdd.keys()):
            xData = xPointsToAdd[(color, symbol, showFilled)]
            yData = yPointsToAdd[(color, symbol, showFilled)]
            self.addCurve(str(i), QColor(*color + (self.alphaValue,)), QColor(*color + (self.alphaValue,)), self.pointWidth, symbol = symbol, xData = xData, yData = yData, showFilledSymbols = showFilled)

        # ##############################################################
        # draw the legend
        # ##############################################################
        if self.showLegend:
            # show legend for discrete class
            if self.dataHasDiscreteClass:
                self.addMarker(self.dataDomain.classVar.name, 0.87, 1.05, Qt.AlignLeft | Qt.AlignVCenter)

                classVariableValues = getVariableValuesSorted(self.dataDomain.classVar)
                for index in range(len(classVariableValues)):
                    if self.useDifferentColors: color = QColor(self.discPalette[index])
                    else:                       color = QColor(Qt.black)
                    y = 1.0 - index * 0.05

                    if not useDifferentSymbols:
                        curveSymbol = self.curveSymbols[0]
                    else:
                        curveSymbol = self.curveSymbols[index]

                    self.addCurve(str(index), color, color, self.pointWidth, symbol = curveSymbol, xData = [0.95], yData = [y], penAlpha = self.alphaValue, brushAlpha = self.alphaValue)
                    self.addMarker(classVariableValues[index], 0.90, y, Qt.AlignLeft | Qt.AlignVCenter)
            # show legend for continuous class
            elif self.dataHasContinuousClass:
                xs = [1.15, 1.20, 1.20, 1.15]
                count = 200
                height = 2 / float(count)
                for i in range(count):
                    y = -1.0 + i*2.0/float(count)
                    col = self.contPalette[i/float(count)]
                    col.setAlpha(self.alphaValue)
                    PolygonCurve(QPen(col), QBrush(col), xData = xs, yData = [y,y, y+height, y+height]).attach(self)

                # add markers for min and max value of color attribute
                [minVal, maxVal] = self.attrValues[self.dataDomain.classVar.name]
                self.addMarker("%s = %%.%df" % (self.dataDomain.classVar.name, self.dataDomain.classVar.numberOfDecimals) % (minVal), xs[0] - 0.02, -1.0 + 0.04, Qt.AlignLeft)
                self.addMarker("%s = %%.%df" % (self.dataDomain.classVar.name, self.dataDomain.classVar.numberOfDecimals) % (maxVal), xs[0] - 0.02, +1.0 - 0.04, Qt.AlignLeft)

        self.replot()


    # ##############################################################
    # create a dictionary value for the data point
    # this will enable to show tooltips faster and to make selection of examples available
    def addTooltipKey(self, x, y, color, index, extraString = None):
        dictValue = "%.1f-%.1f"%(x, y)
        if not self.dataMap.has_key(dictValue): self.dataMap[dictValue] = []
        self.dataMap[dictValue].append((x, y, color, index, extraString))


    def addValueLineCurve(self, x, y, color, exampleIndex, attrIndices):
        XAnchors = numpy.array([val[0] for val in self.anchorData])
        YAnchors = numpy.array([val[1] for val in self.anchorData])
        xs = numpy.array([x] * len(self.anchorData))
        ys = numpy.array([y] * len(self.anchorData))
        dists = numpy.sqrt((XAnchors-xs)**2 + (YAnchors-ys)**2)
        xVect = 0.01 * self.valueLineLength * (XAnchors - xs) / dists
        yVect = 0.01 * self.valueLineLength * (YAnchors - ys) / dists
        exVals = [self.noJitteringScaledData[attrInd, exampleIndex] for attrInd in attrIndices]

        xs = []; ys = []
        for i in range(len(exVals)):
            xs += [x, x + xVect[i]*exVals[i]]
            ys += [y, y + yVect[i]*exVals[i]]
        self.valueLineCurves[0][color] = self.valueLineCurves[0].get(color, []) + xs
        self.valueLineCurves[1][color] = self.valueLineCurves[1].get(color, []) + ys


    def mousePressEvent(self, e):
        if self.manualPositioning:
            self.mouseCurrentlyPressed = 1
            self.selectedAnchorIndex = None
            if not self.normalizeExamples:
                marker, dist = self.closestMarker(e.x(), e.y())
                if dist < 15:
                    self.selectedAnchorIndex = self.shownAttributes.index(str(marker.label().text()))
            else:
                (curve, dist, x, y, index) = self.closestCurve(e.x(), e.y())
                if dist < 5 and str(curve.title().text()) == "dots":
                    self.selectedAnchorIndex = index
        else:
            OWGraph.mousePressEvent(self, e)


    def mouseReleaseEvent(self, e):
        if self.manualPositioning:
            self.mouseCurrentlyPressed = 0
            self.selectedAnchorIndex = None
        else:
            OWGraph.mouseReleaseEvent(self, e)

    # ##############################################################
    # draw tooltips
    def mouseMoveEvent(self, e):
        redraw = (self.tooltipCurves != [] or self.tooltipMarkers != [])

        for curve in self.tooltipCurves:  curve.detach()
        for marker in self.tooltipMarkers: marker.detach()
        self.tooltipCurves = []
        self.tooltipMarkers = []

        canvasPos = self.canvas().mapFrom(self, e.pos())
        xFloat = self.invTransform(QwtPlot.xBottom, canvasPos.x())
        yFloat = self.invTransform(QwtPlot.yLeft, canvasPos.y())

        # in case we are drawing a rectangle, we don't draw enhanced tooltips
        # because it would then fail to draw the rectangle
        if self.mouseCurrentlyPressed:
            if not self.manualPositioning:
                OWGraph.mouseMoveEvent(self, e)
                if redraw: self.replot()
            else:
                if self.selectedAnchorIndex != None:
                    if self.widget.freeVizDlg.restrain == 1:
                        rad = sqrt(xFloat**2 + yFloat**2)
                        xFloat /= rad
                        yFloat /= rad
                    elif self.widget.freeVizDlg.restrain == 2:
                        rad = sqrt(xFloat**2 + yFloat**2)
                        phi = 2 * self.selectedAnchorIndex * math.pi / len(self.anchorData)
                        xFloat = rad * cos(phi)
                        yFloat = rad * sin(phi)
                    self.anchorData[self.selectedAnchorIndex] = (xFloat, yFloat, self.anchorData[self.selectedAnchorIndex][2])
                    self.updateData(self.shownAttributes)
                    self.replot()
                    #self.widget.recomputeEnergy()
            return

        dictValue = "%.1f-%.1f"%(xFloat, yFloat)
        if self.dataMap.has_key(dictValue):
            points = self.dataMap[dictValue]
            bestDist = 100.0
            for (x_i, y_i, color, index, extraString) in points:
                currDist = sqrt((xFloat-x_i)*(xFloat-x_i) + (yFloat-y_i)*(yFloat-y_i))
                if currDist < bestDist:
                    bestDist = currDist
                    nearestPoint = (x_i, y_i, color, index, extraString)

            (x_i, y_i, color, index, extraString) = nearestPoint
            intX = self.transform(QwtPlot.xBottom, x_i)
            intY = self.transform(QwtPlot.yLeft, y_i)

            if self.tooltipKind == LINE_TOOLTIPS and bestDist < 0.05:
                shownAnchorData = filter(lambda p, r=self.hideRadius**2/100: p[0]**2+p[1]**2>r, self.anchorData)
                if not self.normalizeExamples:
                    for (xAnchor,yAnchor,label) in shownAnchorData:
                        attrVal = self.scaledData[self.attributeNameIndex[label]][index]
                        markerX, markerY = xAnchor*(attrVal+0.03), yAnchor*(attrVal+0.03)
                        curve = self.addCurve("", color, color, 1, style = QwtPlotCurve.Lines, symbol = QwtSymbol.NoSymbol, xData = [0, xAnchor*attrVal], yData = [0, yAnchor*attrVal], lineWidth=3)

                        marker = None
                        fontsize = 9
                        markerAlign = Qt.AlignCenter
                        self.tooltipCurves.append(curve)
                        labelIndex = self.attributeNameIndex[label]
                        if self.tooltipValue == TOOLTIPS_SHOW_DATA:
                            if self.dataDomain[labelIndex].varType == orange.VarTypes.Continuous:
                                text = "%%.%df" % (self.dataDomain[labelIndex].numberOfDecimals) % (self.rawData[index][labelIndex])
                            else:
                                text = str(self.rawData[index][labelIndex].value)
                            marker = self.addMarker(text, markerX, markerY, markerAlign, size = fontsize)
                        elif self.tooltipValue == TOOLTIPS_SHOW_SPRINGS:
                            marker = self.addMarker("%.3f" % (self.scaledData[labelIndex][index]), markerX, markerY, markerAlign, size = fontsize)
                        self.tooltipMarkers.append(marker)

            elif self.tooltipKind == VISIBLE_ATTRIBUTES or self.tooltipKind == ALL_ATTRIBUTES:
                if self.tooltipKind == VISIBLE_ATTRIBUTES:
                    shownAnchorData = filter(lambda p, r=self.hideRadius**2/100: p[0]**2+p[1]**2>r, self.anchorData)
                    labels = [s for (xA, yA, s) in shownAnchorData]
                else:
                    labels = []

                text = self.getExampleTooltipText(self.rawData[index], labels)
                text += "<hr>Example index = %d" % (index)
                if extraString:
                    text += "<hr>" + extraString
                self.showTip(intX, intY, text)

        OWGraph.mouseMoveEvent(self, e)
        self.replot()


    # send 2 example tables. in first is the data that is inside selected rects (polygons), in the second is unselected data
    def getSelectionsAsExampleTables(self, attrList, useAnchorData = 1, addProjectedPositions = 0):
        if not self.haveData: return (None, None)
        if addProjectedPositions == 0 and not self.selectionCurveList: return (None, self.rawData)       # if no selections exist
        if (useAnchorData and len(self.anchorData) < 3) or len(attrList) < 3: return (None, None)

        xAttr=orange.FloatVariable("X Positions")
        yAttr=orange.FloatVariable("Y Positions")
        if addProjectedPositions == 1:
            domain=orange.Domain([xAttr,yAttr] + [v for v in self.dataDomain.variables])
        elif addProjectedPositions == 2:
            domain=orange.Domain(self.dataDomain)
            domain.addmeta(orange.newmetaid(), xAttr)
            domain.addmeta(orange.newmetaid(), yAttr)
        else:
            domain = orange.Domain(self.dataDomain)

        domain.addmetas(self.dataDomain.getmetas())

        if useAnchorData: indices = [self.attributeNameIndex[val[2]] for val in self.anchorData]
        else:             indices = [self.attributeNameIndex[label] for label in attrList]
        validData = self.getValidList(indices)
        if len(validData) == 0: return (None, None)

        array = self.createProjectionAsNumericArray(attrList, scaleFactor = self.scaleFactor, useAnchorData = useAnchorData, removeMissingData = 0)
        if array == None:       # if all examples have missing values
            return (None, None)

        #selIndices, unselIndices = self.getSelectionsAsIndices(attrList, useAnchorData, validData)
        selIndices, unselIndices = self.getSelectedPoints(array.T[0], array.T[1], validData)

        if addProjectedPositions:
            selected = orange.ExampleTable(domain, self.rawData.selectref(selIndices))
            unselected = orange.ExampleTable(domain, self.rawData.selectref(unselIndices))
            selIndex = 0; unselIndex = 0
            for i in range(len(selIndices)):
                if selIndices[i]:
                    selected[selIndex][xAttr] = array[i][0]
                    selected[selIndex][yAttr] = array[i][1]
                    selIndex += 1
                else:
                    unselected[unselIndex][xAttr] = array[i][0]
                    unselected[unselIndex][yAttr] = array[i][1]
                    unselIndex += 1
        else:
            selected = self.rawData.selectref(selIndices)
            unselected = self.rawData.selectref(unselIndices)

        if len(selected) == 0: selected = None
        if len(unselected) == 0: unselected = None
        return (selected, unselected)


    def getSelectionsAsIndices(self, attrList, useAnchorData = 1, validData = None):
        if not self.haveData: return [], []

        attrIndices = [self.attributeNameIndex[attr] for attr in attrList]
        if validData == None:
            validData = self.getValidList(attrIndices)

        array = self.createProjectionAsNumericArray(attrList, scaleFactor = self.scaleFactor, useAnchorData = useAnchorData, removeMissingData = 0)
        if array == None:
            return [], []
        array = numpy.transpose(array)
        return self.getSelectedPoints(array[0], array[1], validData)


    # update shown data. Set labels, coloring by className ....
    def savePicTeX(self):
        lastSave = getattr(self, "lastPicTeXSave", "C:\\")
        qfileName = QFileDialog.getSaveFileName(None, "Save to..", lastSave + "graph.pictex","PicTeX (*.pictex);;All files (*.*)")
        fileName = unicode(qfileName)
        if fileName == "":
            return

        if not os.path.splitext(fileName)[1][1:]:
            fileName = fileName + ".pictex"

        self.lastSave = os.path.split(fileName)[0]+"/"
        file = open(fileName, "wt")

        file.write("\\mbox{\n")
        file.write("  \\beginpicture\n")
        file.write("  \\setcoordinatesystem units <0.4\columnwidth, 0.4\columnwidth>\n")
        file.write("  \\setplotarea x from -1.1 to 1.1, y from -1 to 1.1\n")

        if not self.normalizeExamples:
            file.write("\\circulararc 360 degrees from 1 0 center at 0 0\n")

        if self.showAnchors:
            if self.hideRadius > 0:
                file.write("\\setdashes\n")
                file.write("\\circulararc 360 degrees from %5.3f 0 center at 0 0\n" % (self.hideRadius/10.))
                file.write("\\setsolid\n")

            if self.showAttributeNames:
                shownAnchorData = filter(lambda p, r=self.hideRadius**2/100: p[0]**2+p[1]**2>r, self.anchorData)
                if not self.normalizeExamples:
                    for x,y,l in shownAnchorData:
                        file.write("\\plot 0 0 %5.3f %5.3f /\n" % (x, y))
                        file.write("\\put {{\\footnotesize %s}} [b] at %5.3f %5.3f\n" % (l.replace("_", "-"), x*1.07, y*1.04))
                else:
                    file.write("\\multiput {\\small $\\odot$} at %s /\n" % (" ".join(["%5.3f %5.3f" % tuple(i[:2]) for i in shownAnchorData])))
                    for x,y,l in shownAnchorData:
                        file.write("\\put {{\\footnotesize %s}} [b] at %5.3f %5.3f\n" % (l.replace("_", "-"), x*1.07, y*1.04))

        symbols = ("{\\small $\\circ$}", "{\\tiny $\\times$}", "{\\tiny $+$}", "{\\small $\\star$}",
                   "{\\small $\\ast$}", "{\\tiny $\\div$}", "{\\small $\\bullet$}", ) + tuple([chr(x) for x in range(97, 123)])
        dataSize = len(self.rawData)
        labels = self.widget.getShownAttributeList()
        indices = [self.attributeNameIndex[label] for label in labels]
        selectedData = numpy.take(self.scaledData, indices, axis = 0)
        XAnchors = numpy.array([a[0] for a in self.anchorData])
        YAnchors = numpy.array([a[1] for a in self.anchorData])

        r = numpy.sqrt(XAnchors*XAnchors + YAnchors*YAnchors)     # compute the distance of each anchor from the center of the circle
        XAnchors *= r                                               # we need to normalize the anchors by r, otherwise the anchors won't attract points less if they are placed at the center of the circle
        YAnchors *= r

        x_positions = numpy.dot(XAnchors, selectedData)
        y_positions = numpy.dot(YAnchors, selectedData)

        if self.normalizeExamples:
            sum_i = self._getSum_i(selectedData, useAnchorData = 1, anchorRadius = r)
            x_positions /= sum_i
            y_positions /= sum_i

        if self.scaleFactor:
            self.trueScaleFactor = self.scaleFactor
        else:
            abss = x_positions*x_positions + y_positions*y_positions
            self.trueScaleFactor =  1 / sqrt(abss[numpy.argmax(abss)])

        x_positions *= self.trueScaleFactor
        y_positions *= self.trueScaleFactor

        validData = self.getValidList(indices)

        pos = [[] for i in range(len(self.dataDomain.classVar.values))]
        for i in range(dataSize):
            if validData[i]:
                pos[int(self.originalData[self.dataClassIndex][i])].append((x_positions[i], y_positions[i]))

        for i in range(len(self.dataDomain.classVar.values)):
            file.write("\\multiput {%s} at %s /\n" % (symbols[i], " ".join(["%5.3f %5.3f" % p for p in pos[i]])))

        if self.showLegend:
            classVariableValues = getVariableValuesSorted(self.dataDomain.classVar)
            file.write("\\put {%s} [lB] at 0.87 1.06\n" % self.dataDomain.classVar.name)
            for index in range(len(classVariableValues)):
                file.write("\\put {%s} at 1.0 %5.3f\n" % (symbols[index], 0.93 - 0.115*index))
                file.write("\\put {%s} [lB] at 1.05 %5.3f\n" % (classVariableValues[index], 0.9 - 0.115*index))

        file.write("\\endpicture\n}\n")
        file.close()

    def computePotentials(self):
        import orangeom
        #rx = self.transform(QwtPlot.xBottom, 1) - self.transform(QwtPlot.xBottom, 0)
        #ry = self.transform(QwtPlot.yLeft, 0) - self.transform(QwtPlot.yLeft, 1)

        rx = self.transform(QwtPlot.xBottom, 1) - self.transform(QwtPlot.xBottom, -1)
        ry = self.transform(QwtPlot.yLeft, -1) - self.transform(QwtPlot.yLeft, 1)
        ox = self.transform(QwtPlot.xBottom, 0) - self.transform(QwtPlot.xBottom, -1)
        oy = self.transform(QwtPlot.yLeft, -1) - self.transform(QwtPlot.yLeft, 0)

        rx -= rx % self.squareGranularity
        ry -= ry % self.squareGranularity

        if not getattr(self, "potentialsImage", None) \
           or getattr(self, "potentialContext", None) != (rx, ry, self.shownAttributes, self.trueScaleFactor, self.squareGranularity, self.jitterSize, self.jitterContinuous, self.spaceBetweenCells):
            if self.potentialsClassifier.classVar.varType == orange.VarTypes.Continuous:
                imagebmp = orangeom.potentialsBitmap(self.potentialsClassifier, rx, ry, ox, oy, self.squareGranularity, self.trueScaleFactor/2, self.spaceBetweenCells)
                palette = [qRgb(255.*i/255., 255.*i/255., 255-(255.*i/255.)) for i in range(255)] + [qRgb(255, 255, 255)]
            else:
                imagebmp, nShades = orangeom.potentialsBitmap(self.potentialsClassifier, rx, ry, ox, oy, self.squareGranularity, self.trueScaleFactor/2, self.spaceBetweenCells) # the last argument is self.trueScaleFactor (in LinProjGraph...)
                palette = []
                sortedClasses = getVariableValuesSorted(self.potentialsClassifier.domain.classVar)
                for cls in self.potentialsClassifier.classVar.values:
                    color = self.discPalette.getRGB(sortedClasses.index(cls))
                    towhite = [255-c for c in color]
                    for s in range(nShades):
                        si = 1-float(s)/nShades
                        palette.append(qRgb(*tuple([color[i]+towhite[i]*si for i in (0, 1, 2)])))
                palette.extend([qRgb(255, 255, 255) for i in range(256-len(palette))])

            self.potentialsImage = QImage(imagebmp, rx, ry, QImage.Format_Indexed8)
            self.potentialsImage.setColorTable(OWColorPalette.signedPalette(palette) if qVersion() < "4.5" else palette)
            self.potentialsImage.setNumColors(256)
            self.potentialContext = (rx, ry, self.shownAttributes, self.trueScaleFactor, self.squareGranularity, self.jitterSize, self.jitterContinuous, self.spaceBetweenCells)
            self.potentialsImageFromClassifier = self.potentialsClassifier



    def drawCanvas(self, painter):
        if self.showProbabilities and getattr(self, "potentialsClassifier", None):
            if not (self.potentialsClassifier is getattr(self, "potentialsImageFromClassifier", None)):
                self.computePotentials()
            target = QRectF(self.transform(QwtPlot.xBottom, -1), self.transform(QwtPlot.yLeft, 1),
                            self.transform(QwtPlot.xBottom, 1) - self.transform(QwtPlot.xBottom, -1),
                            self.transform(QwtPlot.yLeft, -1) - self.transform(QwtPlot.yLeft, 1))
            source = QRectF(0, 0, self.potentialsImage.size().width(), self.potentialsImage.size().height())
            painter.drawImage(target, self.potentialsImage, source)
#            painter.drawImage(self.transform(QwtPlot.xBottom, -1), self.transform(QwtPlot.yLeft, 1), self.potentialsImage)
        OWGraph.drawCanvas(self, painter)

OWLinProjGraph = graph_deprecator(OWLinProjGraph)

if __name__== "__main__":
    #Draw a simple graph
    import os
    a = QApplication(sys.argv)
    graph = OWLinProjGraph(None)
    data = orange.ExampleTable(r"E:\Development\Orange Datasets\UCI\wine.tab")
    graph.setData(data)
    graph.updateData([attr.name for attr in data.domain.attributes])
    graph.show()
    a.exec_()