1
votes

Normally, PivotTables are used to present data in a certain order. For this specific "issue" I need to present numbers and a target. This conflicts in Pivots as the percentage of column takes into account the target as well.

No worries, with PowerPivot it should be possible to address this issue. So far I have been able to create a table with the following layout:

         1     2    3
cat A    5    10    7
cat B    10    8    9
cat C    0     2    1

Where 1, 2 and 3 are the first three days of a month (for sake of simplicity rest is left out).

I can get totals of the columns as follows:

= Table.FromRows({List.Transform({"1","2","3"}, each List.Sum(Table.Column(prevStep, _)))}, {"1","2","3"})

Furthermore I am able to divide each value of a column by a number:

= Table.TransformColumns(prevStep, List.Transform({"1","2","3"}, each {_, (this) =>  this / 42, type number}))

Now I would like to replace 42 with the totals as calculated previously for the columns columns. Note that "{"1","2","3"}" will be calculated automatically in another step.

Can someone elaborate how to achieve this? Expected result:

         1     2    3
cat A    0.33  0.5    0.41
cat B    0.67  0.4    0.53
cat C    0     0.1    0.06
2

2 Answers

0
votes

I couldn't come up with a way to simply replace 42 in your code, but since I saw this was matrix math, I based this below solution to your problem on Gil Raviv's blog on matrix multiplication.

I started with this as Table1:

enter image description here

Then, to get to this...

enter image description here

...I used Table1 as the source for a new query, and this code:

let
    Source = Table1,
    #"Removed Columns" = Table.RemoveColumns(Source,{""}),
    MatrixA = Table.TransformColumnTypes(#"Removed Columns",{{"1", type number}, {"2", type number}, {"3", type number}}),
    Summed = Table.FromRows({List.Transform({"1","2","3"}, each List.Sum(Table.Column(MatrixA, _)))}, {"1","2","3"}),
    MatrixB = Table.TransformColumnTypes(Summed,{{"1", type number}, {"2", type number}, {"3", type number}}),
    IndexedMatrixA = Table.AddIndexColumn(MatrixA, "Index", 1, 1),
    #"Unpivoted Other Columns" = Table.UnpivotOtherColumns(IndexedMatrixA, {"Index"}, "Attribute", "Value"),
    #"Changed Type" = Table.TransformColumnTypes(#"Unpivoted Other Columns",{{"Attribute", Int64.Type}}),
    RenamedColumnsMatrixA = Table.RenameColumns(#"Changed Type",{{"Index", "Row"}, {"Attribute", "Column"}}),
    IndexedMatrixB = Table.AddIndexColumn(MatrixB, "Index", 1, 1),
    #"Unpivoted Other Columns1" = Table.UnpivotOtherColumns(IndexedMatrixB, {"Index"}, "Attribute", "Value"),
    #"Changed Type1" = Table.TransformColumnTypes(#"Unpivoted Other Columns1",{{"Attribute", Int64.Type}}),
    RenamedColumnsMatrixB = Table.RenameColumns(#"Changed Type1",{{"Index", "Row"}, {"Attribute", "Column"}}),
    #"Merged Queries" = Table.NestedJoin(RenamedColumnsMatrixA,{"Column"},RenamedColumnsMatrixB,{"Column"},"RenamedColumnsMatrixB",JoinKind.LeftOuter),
    #"Expanded RenamedColumnsMatrixB" = Table.ExpandTableColumn(#"Merged Queries", "RenamedColumnsMatrixB", {"Row", "Column", "Value"}, {"B.Row", "B.Column", "B.Value"}),
    #"Added Custom" = Table.AddColumn(#"Expanded RenamedColumnsMatrixB", "AB", each [Value]/[B.Value]),
    #"Grouped Rows" = Table.Group(#"Added Custom", {"Row", "B.Column"}, {{"AB", each List.Sum([AB]), type number}}),
    #"Pivoted Column" = Table.Pivot(Table.TransformColumnTypes(#"Grouped Rows", {{"B.Column", type text}}, "en-US"), List.Distinct(Table.TransformColumnTypes(#"Grouped Rows", {{"B.Column", type text}}, "en-US")[B.Column]), "B.Column", "AB"),
    #"Added Index" = Table.AddIndexColumn(#"Pivoted Column", "Index", 1, 1),
    IndexedSource = Table.AddIndexColumn(Source, "Index", 1, 1),
    #"Merged Queries1" = Table.NestedJoin(#"Added Index",{"Index"},IndexedSource,{"Index"},"IndexedSource",JoinKind.LeftOuter),
    #"Expanded IndexedSource" = Table.ExpandTableColumn(#"Merged Queries1", "IndexedSource", {""}, {"Column1"}),
    #"Removed Columns1" = Table.RemoveColumns(#"Expanded IndexedSource",{"Row", "Index"}),
    #"Reordered Columns" = Table.ReorderColumns(#"Removed Columns1",{"Column1", "1", "2", "3"})
in
    #"Reordered Columns"

You'll notice I kinda "jumped around" a bit, using previous Applied Steps (which are basically query table states) as "tables" in the code...like in these Applied Steps:

  • Index MatrixB (IndexedMatrixB = Table.AddIndexColumn(MatrixB, "Index", 1, 1), where MatrixB is a previous Applied Step); and

  • Merged Queries (#"Merged Queries" = Table.NestedJoin(RenamedColumnsMatrixA,{"Column"},RenamedColumnsMatrixB,"Column"},"RenamedColumnsMatrixB",JoinKind.LeftOuter), where RenamedColumnsMatrixA and RenamedColumnsMatrixB are previous Applied Steps).

0
votes

@Marc Pince; thanks for your suggestions. After some struggeling I was capable to cook the following:

First, I swapped the total sum calculation line as follows:

totMinDay =  Table.ToRows    (Table.FromRows({List.Transform(daysOfMonth , each List.Sum(Table.Column(prevStep, _)))}, daysOfMonth )){0},

Where daysOfMonth are the days of that actual month, obtained elsewhere.

Next I calculate the percentage as follows:

perc= Table.TransformColumns(prevStep, List.Transform(daysOfMonth , each {_, (this) =>  Number.Round(this/ ( totMinDay {Number.FromText( _ ) - 1} ), 3), type number})),

Here I use the fact that the index of the sum is based on the day of month minus 1 (starting from zero) providing solid results.

Thanks for all inputs.