Excel XY Chart (Scatter plot) Data Label No Overlap

Question

So I've been working on this for the past week. Although it can't do miracles, I can say I've got a pretty good result:
I just wanted to put this code out there for all the poor souls like me that are looking for some kind of vba macro that helps them avoid label overlaps in a scatter plot, because while doing my research on the subject, I wasn't able to find anything helpful.

Solution 1

Const PIXEL_TO_POINT_RATIO As Double = 0.72 '1 Pixel = 72/96*1 Point
Const tStep As Double = 0.1
Const rStep As Double = 0.1
Dim pCount As Integer

Sub ExampleMain()

        RearrangeScatterLabels Sheet5 

        RearrangeScatterLabels Sheet25

End Sub

Sub RearrangeScatterLabels(sht As Worksheet)
    Dim plot As Chart
    Dim sCollection As SeriesCollection
    Dim dLabels() As DataLabel
    Dim dPoints() As Point
    Dim xArr(), yArr(), stDevX, stDevY As Double
    Dim x0, x1, y0, y1 As Double
    Dim temp() As Double
    Dim theta As Double
    Dim r As Double
    Dim isOverlapped As Boolean
    Dim safetyNet, validEntry, currentPoint As Integer

    Set plot = sht.ChartObjects(1).Chart 'XY chart (scatter plot)
    Set sCollection = plot.SeriesCollection 'All points and labels
    safetyNet = 1
    pCount = (sCollection.Count - 1)

    ReDim dLabels(1 To 1)
    ReDim dPoints(1 To 1)
    ReDim xArr(1 To 1)
    ReDim yArr(1 To 1)

    For pt = 1 To sCollection(1).Points.Count
        For i = 1 To pCount
            If sCollection(i).Points.Count <> 0 Then
                'Dynamically expand the arrays
                validEntry = validEntry + 1
                If validEntry <> 1 Then
                    ReDim Preserve dLabels(1 To UBound(dLabels) + 1)
                    ReDim Preserve dPoints(1 To UBound(dPoints) + 1)
                    ReDim Preserve xArr(1 To UBound(xArr) + 1)
                    ReDim Preserve yArr(1 To UBound(yArr) + 1)
                End If

                Set dLabels(i) = sCollection(i).Points(pt).DataLabel 'Store all label objects
                Set dPoints(i) = sCollection(i).Points(pt)           'Store all point objects
                temp = getElementDimensions(, dPoints(i))
                xArr(i) = temp(0) 'Store all points x values
                yArr(i) = temp(2) 'Store all points y values
            End If
        Next
    Next

    If UBound(dLabels) < 2 Then Exit Sub

    pCount = UBound(dLabels)
    stDevX = Application.WorksheetFunction.StDev(xArr) 'Get standard deviation for x
    stDevY = Application.WorksheetFunction.StDev(yArr) 'Get standard deviation for y
    If stDevX = 0 Then stDevX = 1
    If stDevY = 0 Then stDevY = 1
    r = 0

    For currentPoint = 1 To pCount
        theta = Rnd * 2 * Application.WorksheetFunction.Pi()
        x0 = xArr(currentPoint)
        y0 = yArr(currentPoint)
        x1 = xArr(currentPoint)
        y1 = yArr(currentPoint)
        isOverlapped = True

        Do Until Not isOverlapped
            safetyNet = safetyNet + 1

            If safetyNet < 500 Then
                If Not checkForOverlap(dLabels(currentPoint), dLabels, dPoints, plot) Then
                    'No label is within bounds and not overlapping
                    isOverlapped = False
                    r = 0
                    theta = Rnd * 2 * Application.WorksheetFunction.Pi()
                    safetyNet = 1
                Else
                    'Move label so it does not overlap
                    theta = theta + tStep
                    r = r + rStep * tStep / (2 * Application.WorksheetFunction.Pi())
                    x1 = x0 + stDevX * r * Cos(theta)
                    y1 = y0 + stDevY * r * Sin(theta)
                    dLabels(currentPoint).Left = x1
                    dLabels(currentPoint).Top = y1
                End If
            Else
                safetyNet = 1
                Exit Do
            End If
        Loop
    Next
End Sub

Function checkForOverlap(ByRef dLabel As DataLabel, ByRef dLabels() As DataLabel, ByRef dPoints() As Point, ByRef dChart As Chart) As Boolean
    checkForOverlap = False 'Return false by default

    'Detect label going over chart area
    If detectOverlap(dLabel, , , dChart) Then
        checkForOverlap = True
        Exit Function
    End If

    'Detect labels overlap
    For i = 1 To pCount
        If Not dLabel.Left = dLabels(i).Left Then
            If detectOverlap(dLabel, dLabels(i)) Then
                checkForOverlap = True
                Exit Function
            End If
        End If
    Next

    'Detect label overlap with point
    For i = 1 To pCount
        If detectOverlap(dLabel, , dPoints(i)) Then
            checkForOverlap = True
            Exit Function
        End If
    Next
End Function

Function getElementDimensions(Optional dLabel As DataLabel, Optional dPoint As Point, Optional dChart As Chart) As Double()
    'Get element dimensions and compensate slack
    Dim eDimensions(3) As Double

    'Working in IV quadrant
    If dPoint Is Nothing And dChart Is Nothing Then
        'Get label dimensions and compensate padding
        eDimensions(0) = dLabel.Left + PIXEL_TO_POINT_RATIO * 3                'Left
        eDimensions(1) = dLabel.Left + dLabel.Width - PIXEL_TO_POINT_RATIO * 3 'Right
        eDimensions(2) = dLabel.Top + PIXEL_TO_POINT_RATIO * 6                 'Top
        eDimensions(3) = dLabel.Top + dLabel.Height - PIXEL_TO_POINT_RATIO * 3 'Bottom
    End If
    If dLabel Is Nothing And dChart Is Nothing Then
        'Get point dimensions
        eDimensions(0) = dPoint.Left - PIXEL_TO_POINT_RATIO * 5 'Left
        eDimensions(1) = dPoint.Left + PIXEL_TO_POINT_RATIO * 5 'Right
        eDimensions(2) = dPoint.Top - PIXEL_TO_POINT_RATIO * 5  'Top
        eDimensions(3) = dPoint.Top + PIXEL_TO_POINT_RATIO * 5  'Bottom
    End If
    If dPoint Is Nothing And dLabel Is Nothing Then
        'Get chart dimensions
        eDimensions(0) = dChart.PlotArea.Left + PIXEL_TO_POINT_RATIO * 22                         'Left
        eDimensions(1) = dChart.PlotArea.Left + dChart.PlotArea.Width - PIXEL_TO_POINT_RATIO * 22 'Right
        eDimensions(2) = dChart.PlotArea.Top - PIXEL_TO_POINT_RATIO * 4                           'Top
        eDimensions(3) = dChart.PlotArea.Top + dChart.PlotArea.Height - PIXEL_TO_POINT_RATIO * 4  'Bottom
    End If

    getElementDimensions = eDimensions 'Return dimensions array in Points
End Function

Function detectOverlap(ByVal dLabel1 As DataLabel, Optional ByVal dLabel2 As DataLabel, Optional ByVal dPoint As Point, Optional ByVal dChart As Chart) As Boolean
    'Left, Right, Top, Bottom
    Dim AxL, AxR, AyT, AyB As Double 'First label coordinates
    Dim BxL, BxR, ByT, ByB As Double 'Second label coordinates
    Dim eDimensions() As Double 'Element dimensions

    eDimensions = getElementDimensions(dLabel1)
    AxL = eDimensions(0)
    AxR = eDimensions(1)
    AyT = eDimensions(2)
    AyB = eDimensions(3)

    If dPoint Is Nothing And dChart Is Nothing Then
        'Compare with another label
        eDimensions = getElementDimensions(dLabel2)
    End If
    If dLabel2 Is Nothing And dChart Is Nothing Then
        'Compare with a point
        eDimensions = getElementDimensions(, dPoint)
    End If
    If dPoint Is Nothing And dLabel2 Is Nothing Then
        'Compare with chart area
        eDimensions = getElementDimensions(, , dChart)
    End If
    BxL = eDimensions(0)
    BxR = eDimensions(1)
    ByT = eDimensions(2)
    ByB = eDimensions(3)

    If dChart Is Nothing Then
        detectOverlap = (AxL <= BxR And AxR >= BxL And AyT <= ByB And AyB >= ByT) 'Reverse De Morgan's Law
    Else
        detectOverlap = Not (AxL >= BxL And AxR <= BxR And AyT >= ByT And AyB <= ByB) 'Is in chart bounds (working in IV quadrant)
    End If
End Function

I realize the code is kinda rough and not optimized, but I can't spend more time on this project. I've left quite a few notes around to help read it, should anyone choose to continue this project.

Hope this helps.
Best wishes, Schadenfreude.

Solution 2

Building on your function, I made a routine to randomly reposition the labels, assigning a score according to how much overlap it would cause, and thusly optimize. The results aren't great for my own data set, but I think it can be tuned easily for most usages.

There are some issues with the borders and the axis labels which maybe I'll account for later.

Option Explicit

Sub ExampleUsage()

    RearrangeScatterLabels ActiveSheet.ChartObjects(1).Chart, 3

End Sub

Sub RearrangeScatterLabels(plot As Chart, Optional timelimit As Double = 5)

    Dim sCollection As SeriesCollection
    Set sCollection = plot.SeriesCollection

    Dim pCount As Integer
    pCount = sCollection(1).Points.Count
    If pCount < 2 Then Exit Sub

    Dim dPoints() As Point

    Dim xArr() As Double ' Label center position X
    Dim yArr() As Double ' Label center position Y
    Dim wArr() As Double ' Label width
    Dim hArr() As Double ' Label height
    Dim pArr() As Double ' Marker position X
    Dim qArr() As Double ' Marker position Y
    Dim mArr() As Double ' Markersize

    ReDim dPoints(1 To pCount)

    ReDim xArr(1 To pCount)
    ReDim yArr(1 To pCount)
    ReDim wArr(1 To pCount)
    ReDim hArr(1 To pCount)
    ReDim pArr(1 To pCount)
    ReDim qArr(1 To pCount)
    ReDim mArr(1 To pCount)

    Dim theta As Double

    Dim i As Integer
    Dim j As Integer
    Dim dblStart As Double

    ' Loop through all points to get their handles and coordinates
    For i = 1 To pCount

        ' Store all point objects
        Set dPoints(i) = sCollection(1).Points(i)

        ' Extract their coordinates and size
        pArr(i) = dPoints(i).Left
        qArr(i) = dPoints(i).Top
        mArr(i) = dPoints(i).MarkerSize

        ' Store the size of the corresponding labels
        wArr(i) = dPoints(i).DataLabel.Width
        hArr(i) = dPoints(i).DataLabel.Height

        ' Starting position (center of label) is middle below
        xArr(i) = pArr(i)
        yArr(i) = qArr(i) + mArr(i)

    Next

    Dim newX As Double
    Dim newY As Double
    Dim dE As Double

    Dim wgtOverlap As Double
    Dim wgtDistance As Double
    Dim wgtClose As Double

    wgtOverlap = 10000 ' Extra penalty for overlapping
    wgtDistance = 10000 ' Penalty for being nearby other labels
    wgtClose = 10 ' Penalty for being further from marker

    ' Limit the function by time
    dblStart = Timer
    Do Until TimerDiff(dblStart, Timer) > timelimit

        ' Pick a random label to move around
        i = Int(Rnd * pCount + 1)

        ' Pick a new random position by angle
        theta = Rnd * 2 * Application.WorksheetFunction.Pi()

        ' Determine the position it would shift to
        If Abs(Sin(theta) * wArr(i)) > Abs(hArr(i) * Cos(theta)) Then
            ' above or below
            If Sin(theta) > 0 Then
                ' above
                newX = pArr(i) + wArr(i) * Cos(theta) / 2
                newY = qArr(i) - hArr(i) / 2 - mArr(i) / 2
            Else
                ' below
                newX = pArr(i) + wArr(i) * Cos(theta) / 2
                newY = qArr(i) + hArr(i) / 2 + mArr(i) / 2
            End If
        Else
            ' left or right side
            If Cos(theta) < 0 Then
                ' left
                newX = pArr(i) - wArr(i) / 2 - mArr(i) / 2
                newY = qArr(i) - hArr(i) * Sin(theta) / 2
            Else
                ' right
                newX = pArr(i) + wArr(i) / 2 + mArr(i) / 2
                newY = qArr(i) - hArr(i) * Sin(theta) / 2
            End If
        End If

        ' Determine increase in energy caused by this shift
        dE = 0
        For j = 1 To pCount
            If i <> j Then
                ' Current overlap with labels
                If 2 * Abs(xArr(i) - xArr(j)) < wArr(i) + wArr(j) _
                    And 2 * Abs(yArr(i) - yArr(j)) < hArr(i) + hArr(j) Then
                    dE = dE - Abs(xArr(i) - xArr(j) + (wArr(i) + wArr(j)) / 2) _
                        * Abs(yArr(i) - yArr(j) + (hArr(i) + hArr(j)) / 2)
                    dE = dE - wgtOverlap
                End If
                ' New overlap with labels
                If 2 * Abs(newX - xArr(j)) < wArr(i) + wArr(j) _
                    And 2 * Abs(newY - yArr(j)) < hArr(i) + hArr(j) Then
                    dE = dE + Abs(newX - xArr(j) + (wArr(i) + wArr(j)) / 2) _
                        * Abs(newY - yArr(j) + (hArr(i) + hArr(j)) / 2)
                    dE = dE + wgtOverlap
                End If
                ' Current overlap with labels
                If Abs(xArr(i) - pArr(j)) < wArr(i) / 2 + mArr(j) _
                    And Abs(yArr(i) - qArr(j)) < hArr(i) / 2 + mArr(j) Then
                    dE = dE - wgtOverlap
                End If
                ' New overlap with points
                If Abs(newX - pArr(j)) < wArr(i) / 2 + mArr(j) _
                    And Abs(newY - qArr(j)) < hArr(i) / 2 + mArr(j) Then
                    dE = dE + wgtOverlap
                End If
                ' We like the neighbours to be far away
                dE = dE - wgtDistance / ((xArr(i) - xArr(j)) ^ 2 + (yArr(i) - yArr(j)) ^ 2)
                dE = dE + wgtDistance / ((newX - xArr(j)) ^ 2 + (newY - yArr(j)) ^ 2)
            End If
            ' We like the offsets to be low
            dE = dE - wgtClose * (Abs(xArr(i) - pArr(i)) + Abs(yArr(i) - qArr(i)))
            dE = dE + wgtClose * (Abs(newX - pArr(i)) + Abs(newY - qArr(i)))
        Next

        ' If it didn't get worse, adjust to new position
        If dE <= 0 Then
            xArr(i) = newX
            yArr(i) = newY
        End If

    Loop

    ' Actually adjust the labels
    For i = 1 To pCount
        dPoints(i).DataLabel.Left = xArr(i) - wArr(i) / 2
        dPoints(i).DataLabel.Top = yArr(i) - hArr(i) / 2
    Next

End Sub

' Timer function from Peter Albert
' http://stackoverflow.com/questions/15634623
Function TimerDiff(dblTimerStart As Double, dblTimerEnd As Double)
    Dim dblTemp As Double
    dblTemp = dblTimerEnd - dblTimerStart
    If dblTemp < -43200 Then
        dblTemp = dblTemp + 86400
    End If
    TimerDiff = dblTemp
End Function