11-28-2018 10:27 AM
At a recent user group meeting in Columbus Ohio, one of my members presented me with a problem they were having. The member was working with a SharePoint list and attempting to unpivot a number of columns on thousands of list items. For some reason, the unpivoting, specifically, was causing massive performance issues. Therefore, I suggested unpivoting the data in DAX instead and created this example for the member to demonstrate the technique. Fortunately, this worked very well and was very performant even at scale. So, if you are having trouble with unpivoting data in Power Query, this technique may work for you.
The original example data looks like this:
|Meeting 1||Monday, January 1, 2018||Greg||Jason|
|Meeting 2||Friday, January 5, 2018||Jason||Frank||George||Battina|
|Meeting 3||Wednesday, January 10, 2018||John||Greg||Jason||Battina||George||Frank|
The DAX Unpivot formula is:
Table = FILTER( UNION( SELECTCOLUMNS('Table1',"Name",[Name],"Date",[Date],"Column1",[Column1]), SELECTCOLUMNS('Table1',"Name",[Name],"Date",[Date],"Column1",[Column2]), SELECTCOLUMNS('Table1',"Name",[Name],"Date",[Date],"Column1",[Column3]), SELECTCOLUMNS('Table1',"Name",[Name],"Date",[Date],"Column1",[Column4]), SELECTCOLUMNS('Table1',"Name",[Name],"Date",[Date],"Column1",[Column5]), SELECTCOLUMNS('Table1',"Name",[Name],"Date",[Date],"Column1",[Column6]) ), [Column1]<>"")
This will make the above data look like this:
|Meeting 1||1/1/2018 0:00||Greg|
|Meeting 2||1/5/2018 0:00||Jason|
|Meeting 3||1/10/2018 0:00||John|
|Meeting 1||1/1/2018 0:00||Jason|
|Meeting 2||1/5/2018 0:00||Frank|
|Meeting 3||1/10/2018 0:00||Greg|
|Meeting 2||1/5/2018 0:00||George|
|Meeting 3||1/10/2018 0:00||Jason|
|Meeting 2||1/5/2018 0:00||Battina|
|Meeting 3||1/10/2018 0:00||Battina|
|Meeting 3||1/10/2018 0:00||George|
|Meeting 3||1/10/2018 0:00||Frank|
I think you need to add an 'addcolumn' in there to bring in the Column Heading information. The way you have it now, assumes that the value data is all the same thing. ie people. But usually the different columns contain different 'types' of information like, color, size, cost, etc. So that info needs to be included in the unpivot...
This is great, but one thing is missing for me.
In the table on the right, the column headres are stored in the "Purchase type" column (what would be the Attribute column in Power Query), while this DAX code doesn't create that attribute column. Any ideas on how to achieve that?
I am not sure what I amissing but I am with Nart, the column headers NEED to be dynamic. The Union trick is helpful but you are hardcoding "Column1", etc. How do we generate the "Purchase Type" column from large number of columns like attribute column in Power Query, except of course with distinct values...?
Hi @Greg_Deckler ,
This is awsome technique, I am trying to implement it but getting following error, could you please guide me on this. All columns which I am trying to unpivot are of date type you can see it on the right side in fields.
Thanks in advance.
@PoojaDarbhe - I can't see all of the formula so not sure exactly where you are getting the error in your DAX. If we are talking about the last line in your screen shot, you could try to replace "" with BLANK().
I have same error message as@PoojaDarbhe , i tred to change "" to (), then i got another error as below, please help, thanks.
changed "" to () as suggested but got another error as below
Hi @Anonymous ,
Instead of [Output All] write [Outputfg], it should be column name not the value.
Please try this and let me know if that works.
Hello @Greg_Deckler ,
Just stumbled upon your tip. I already knew it, but I was wondering, as the first section of selectcolumns is the same across all, can you bring in a string value created by DAX and apply it to the last bit? A little bit like using Indirect in Excel.