cancel
Showing results for
Did you mean:
Anonymous
Not applicable

Please explain cumulative sum work principle

So, I have the followint formula, that calculates cumulative sum for a measure

```CumulativeAmount =
CALCULATE (
SUM ( sheet1[amount] ),
FILTER ( ALL ( calendar[Date] ), calendar[Date] <= MAX ( calendar[Date] ) )
)```

This works perfect for me, but I can't understand fully how it works.

Let's say I have transactions for year 2017. The formula calculates sum of amount, but filters only the cases where Date was below the max selected date, shifting the context, so that it will include ALL the dates, not only the ones included in current context. But why then MAX(Calendar[Date]) is not 2017-12-31 ? If you override current date context, and calculate MAX, it should be the highest value of a column, which is the last day of the year (if the date table was automatically generated), no ?

Why for the left part of condition (calendar[Date]) in filter expression shifts current context (and returns all dates from the very beginning), but the right part (Max(calendar[Date])) does not, and returns the maximum date inside current context ?

I have seen this method of cumulative sum calculation in many sources, but none of them explains this particular part.

Thanks !

1 ACCEPTED SOLUTION
Microsoft

Hi @Anonymous,

The use of ALL(Date) in order to ignore the current context. In fact, FILTER iterates over the entire table, analyzing dates that are outside of the current filter context. In this way, it will return date that are lower than or equal to the current filter.

The comparisons of Date[DateKey] against MAX(Date[Datekey]). When you are not familiar with DAX, these expressions look strange. However, if you recall the exact meaning of MAX , you see that it means "the maximum value of DateKey in the current context." Because the expression is part of CALCULATE filters, it still works in the original filter context. On the other hand, the expression Date[DateKey] is a column name, meaning "The value of DateKey in the current row context which is created by the FILTER during its iteration."

Regards

4 REPLIES 4
Anonymous
Not applicable

Thanks everyone, looks like I figured it out... I think...

Microsoft

Hi @Anonymous,

The use of ALL(Date) in order to ignore the current context. In fact, FILTER iterates over the entire table, analyzing dates that are outside of the current filter context. In this way, it will return date that are lower than or equal to the current filter.

The comparisons of Date[DateKey] against MAX(Date[Datekey]). When you are not familiar with DAX, these expressions look strange. However, if you recall the exact meaning of MAX , you see that it means "the maximum value of DateKey in the current context." Because the expression is part of CALCULATE filters, it still works in the original filter context. On the other hand, the expression Date[DateKey] is a column name, meaning "The value of DateKey in the current row context which is created by the FILTER during its iteration."

Regards

Regular Visitor

Hello, thanks for your explanation. But I still don't understant why the Date[Datekey] in Max() is under the current context. Why does the first parameter of Filter() not work on it?

Helper II

I am going through this same issue on understanding, and the root of the confusion is the fact that semantically, Date[Datekey] means different things in this equation based on where it is in the equation.  By itself, it means the row context, but within a filter that is within a calculate it means the filter context of the calculate.  This isn't the only language where a variable means something different based on its location.  Javascript's scope chain is also lexically scoped, but the rule there is really easy, you go up the scope chain, and if you can't figure out what the scope chain is, look in the debugger.  Here, if it's in a "max" function it's in the filter context of the calculate, but what happens when you nest calculate functions.  Further, what other functions have their own filter contexts?  Sumx, sum?

I have to admit, among obiee, cognos, crystal and tableau, Dax is the hardest most inconsistent language I have ever tried to pick up.  The others can be picked up in a few days, but not dax.  I don't know who the target user is of the dax language, but it's not normal people.  You can't just be a person of average programming ability and figure it out.  Reducing 3sat to vertex cover, I figured out in 40 minutes.  This calculate function is way harder for me.  Congrats to all of you who have conquered it.

Announcements

CollabDays Lisbon - 26 November 2022

Sessions include practical, hands-on experience that will help you take the next step in your career and know-how.

Difinity Conference 2022

Difinity is the largest Microsoft Data, AI, Power BI, Power Platform and Business Applications Conference in New Zealand.

European Share Point Conference

This conference returns live and in-person in Copenhagen this 28 November-01 December with 4 Microsoft Keynotes, 9 Tutorials, and 120 Sessions.

Microsoft 365 Conference - 06-08 December

Join us in Las Vegas to experience community, incredible learning opportunities, and connections that will help grow skills, know-how, and more.

Top Solution Authors
Top Kudoed Authors