Combining X Variety of Rows in Energy Question for Energy BI, Excel and Energy Question On-line

Some time again, I used to be engaged on a venture involving getting information from Excel recordsdata. The Excel recordsdata comprise the information in sheets and tables. Getting the information from the tables is simple. Nonetheless, the information within the sheets have some commentaries on prime of the sheet, then the column names after which the information itself. One thing like beneath:

This strategy is fairly constant throughout many Excel recordsdata. The client desires to have the commentary within the column names when the information is imported into Energy BI. So the ultimate outcome should seem like this:

The enterprise requirement although is to mix the primary 3 rows of information and advertise because the column title.

The Problem

Let’s join the Excel file and take a look at the information in Energy BI Desktop.

As you’ll be able to see within the previous picture, Energy BI, or extra exactly, Energy Question, sees the information in Desk format. After we click on the Remodel Information button, that is what we get in Energy Question Editor:

Everyone knows that tables encompass Columns and Rows. The conjunction of a column and a row is a Cell. What we require to do is to concatenate the values of cells from the primary three rows. We even have to make use of a House character to separate the values of every cell from the others.

In Energy Question, we are able to get every row of information in as a Document with the next syntax:

Desk{RecordIndex}

Within the above syntax, the Desk will be the outcomes of the earlier transformation step, and the RecordIndex begins from 0. So to get the primary row of the desk within the previous picture, we use the next syntax:

#"Modified Kind"{0}

The place the #"Modified Kind" is the earlier step. Listed below are the outcomes of working the previous expression:

So we are able to get the second and third rows with comparable expressions. The next picture exhibits all the codes within the Superior Editor:

However how can we concatenate the values of the rows?

My strategy is:

• Making a desk from the primary three rows
• Transposing the resutls to get the rows as columns
• Merging the columns right into a single column
• Transposing the brand new column
• Appending the transposed values to the unique desk
• Selling the primary row as column names

Right here is how we achieve this:

1. We create a Desk from Data utilizing the Desk.FromRecords() perform:

2. Transposing the resutls to get the rows as columns

3. Merging all columns and eradicating the columns. To take action we choose all columns, right-click a specific column and click on Merge Columns. We use House because the separator

4. Transposing the brand new merged column

5. Appending the transposed outcomes to the unique desk. The resutls of the third transformation step provides us the orignial desk. We use the Desk.Mix() perform

Desk.Mix({#"Transposed Table1", #"Modified Kind"})

As you’ll be able to within the previous picture, the outcomes should not what we would like as we nonetheless have the primary 3 rows showing within the values. So earlier than we append the transposed outcomes to the unique desk, we have now to take away the primary 3 rows from the unique desk, after which append the outcomes. To take away the primary N rows we use Desk.Skip() perform. So we have now to alter the earlier step as follows:

Desk.Mix({#"Transposed Table1", Desk.Skip(#"Modified Kind", 3)})

6. Selling the primary row as column names

Simple!

Hmm… Probably not. Particularly when we have now a number of different Excel recordsdata; some have two, some have three rows of feedback, and a few have much more. So not solely is the duty time-consuming, however it is also fairly boring and susceptible to human errors. What if there’s a customized perform that I can invoke to get the job accomplished?

The Resolution

To beat the problem, I wrote a fairly dynamic customized perform. It accepts a desk and an iterator. The iterator is the variety of rows we need to concatenate and use because the column names. After invoking the perform, it iterated by the primary X variety of rows following the state of affairs I defined within the earlier part. It then appends the transposed outcomes to the unique desk whereas eradicating the primary X variety of rows.

Right here you go…

Once I examined my authentic perform with totally different information sorts, I famous that I needed to convert the remark values to textual content information sort; in any other case, dynamically, the merging columns step breaks because it can not concatenate textual content and quantity values.

A giant shout-out to Imke Feldman for writing such useful blogposts. I used her approach in a single block of the next code to dynamically convert the columns to textual content.

// fn_CombineTopXRows
(InputTable as desk, Iterator as quantity) as desk =>
let
  TopXRows = Record.Generate(
        () => 0
        , every _ < Iterator
        , every _ + 1
        , every InputTable{_}
        ),
    TableFromRecords = Desk.FromRecords(
        Record.Generate(
            () => 0
            , every _ < Iterator
            , every _ + 1
            , every InputTable{_}
            )
        ),
    TransposeTable = Desk.Transpose(
        TableFromRecords
        , Record.Generate(
            () => 0
            , every _ < Iterator
            , every _ + 1
            , every "Col" & Textual content.From(_)
            )
        ),
    
    // Particular due to Imke Feldman for this line of code?
    // https://www.thebiccountant.com/2017/01/09/dynamic-bulk-type-transformation-in-power-query-power-bi-and-m/
    DataTypeToText = Desk.TransformColumnTypes(
        TransposeTable
        , Record.Remodel(
            Desk.ColumnNames(TransposeTable)
            , every {_, sort textual content}
            )
        ),
    
    CombineColumns = Desk.CombineColumns(
        DataTypeToText
        , Record.Generate(
            () => 0
            , every _ < Iterator
            , every _ + 1
            , every "Col" & Textual content.From(_)
            )
        , Combiner.CombineTextByDelimiter(
            " "
            , QuoteStyle.None
            )
        ,"Merged"
        ),
    TransposeCombined = Desk.Transpose(CombineColumns),
    AppendToOriginalTable = Desk.Mix({TransposeCombined, Desk.Skip(InputTable, Iterator)}),
    PromotedHeaders = Desk.PromoteHeaders(AppendToOriginalTable, [PromoteAllScalars=true])
in
    PromotedHeaders

You possibly can obtain the code from right here.

Let’s see the fn_CombineTopXRows customized perform in motion. Listed below are the outcomes after invoking the perform with our earlier pattern information:

Here’s a extra excessive pattern information containing particular characters, nulls and Unicode characters:

And listed below are the outcomes of invoking the fn_CombineTopXRows customized perform:

As all the time, when you have any feedback or you realize a greater approach than I defined right here, please let me know by way of the feedback part beneath.

Replace

After sharing this publish on LinkedIn, Muneer Marzouq, commented that he would write the customized perform otherwise. I appreciated his strategy, so I share his code right here.

// fn_CombineTopNRows
// Supply: https://www.linkedin.com/feed/replace/urn:li:exercise:6866137870068002816?commentUrn=urnpercent3Alipercent3Acommentpercent3Apercent28activitypercent3A6866137870068002816percent2C6866640507234254848percent29&replyUrn=urnpercent3Alipercent3Acommentpercent3Apercent28activitypercent3A6866137870068002816percent2C6867174468557529088percent29

(InputTable as desk, TopNRows as quantity, non-obligatory Delimiter as nullable textual content) as desk =>
let
  Delimiter = if Delimiter <> null then Delimiter else "#(lf)",
  ToTables = Desk.Cut up(InputTable, TopNRows),
  ColsNames = Desk.ColumnNames(ToTables{0}),
  ToText = 
    Desk.TransformColumnTypes(
      ToTables{0}
      , Record.Remodel(
          ColsNames
          , every {_, sort textual content}
        )
    ),
  ToColumns = Desk.ToColumns(ToText),
  CombineText = 
    Record.Remodel(
      ToColumns
      , every Textual content.Mix(_
        , Delimiter
      )
    ),
  ToTable = Desk.FromRows({CombineText}, ColsNames),
  Mix = Desk.Mix({ToTable, ToTables{1}})
in
  Mix

Replace 2022 Jan

Right here is one other resolution that Mahmoud Baniasadi put within the feedback part. What I like about Mahmoud’s code are:

• Utilizing the Desk.SplitAt() perform which returns an inventory of two tables as an alternative of Desk.Cut up() perform which returns an inventory of tables
• Implementing the perform documentation

let func=
(InputTable as desk, TopNRows as quantity, non-obligatory Delimiter as nullable textual content) as desk =>
let

Delimiter = if Delimiter <> null then Delimiter else "#(lf)",
// Correction: Change Desk.Cut up to Desk.SplitAt
ToTables = Desk.SplitAt(InputTable, TopNRows),
ColsNames=Desk.ColumnNames(ToTables{0}),
ToText= Desk.TransformColumnTypes(
ToTables{0}
, Record.Remodel(
ColsNames
, every {_, sort textual content}
)
),

/* My various methodology for this half
CombineText=
Record.Remodel(
Desk.ToList(
Desk.Transpose(ToText),
Combiner.CombineTextByDelimiter(Delimiter,0)),
every Textual content.Trim(_)),
*/

// Muneer Marzouq Resolution
ToColumns=Desk.ToColumns(ToText),
CombineText =
Record.Remodel(
ToColumns
, every Textual content.Mix(_
, Delimiter
)
),
//Finish of Muneer code which is totally different from mine

ToTable = Desk.FromRows({CombineText}, ColsNames),
Mix = Desk.Mix({ToTable, ToTables{1}}),
// Correction: add one step to advertise mixed rows as header
PromotedHeaders = Desk.PromoteHeaders(Mix, [PromoteAllScalars=true])
in
PromotedHeaders,
// Add documentation
documentation = [
Documentation.Name = " fn_CombineTopNRows ",
Documentation.Description = " Returns a Table with Combining X Number of Rows. ",
Documentation.LongDescription = " Returns a Table with Combining X Number of Rows and Promote them as header. ",
Documentation.Category = " Table ",
Documentation.Source = " https://www.biinsight.com/combining-x-number-of-rows-in-power-query-for-power-bi-excel-and-power-query-online/ ",
Documentation.Version = " 1.0 ",
Documentation.Author = " Muneer Marzouq, Mahmoud Bani Asadi, Soheil Bakhshi ",
Documentation.Examples = {[Description = " see this blogpost: https://www.biinsight.com/combining-x-number-of-rows-in-power-query-for-power-bi-excel-and-power-query-online/ ",
Code = " fn_CombineTopNRows(#table({""Column1"", ""Column2"", ""Column3""}, {{""TitleA"", ""TitleC"", ""Title1""}, {""TitleB"", ""TitleD"", ""Title2""}, {7, 8, 9}}), 2, "" "")",
Result = " #table({""TitleA TitleB"", ""TitleC TitleD"", ""Title1 Title2""}, {{7, 8, 9}}) "]}]
in
Worth.ReplaceType(func, Worth.ReplaceMetadata(Worth.Kind(func), documentation))

For this reason I like information sharing. You be taught one thing new from others every single day.