SSAS DAX formula for same period last year not same month last year

Question

The DAX function sameperiodlastyear, only works if the period we are interested in, is the same as a month.
However our Period 3 for 2014 was 05-march to 04-April and Period 3 in 2013 was 01-March to 31-March
I have a dim_Date table with a separate Year Number and Period Number so you be able to match on Period Number and Year - 1
The sort of formula I am after is this, so that the P3 total for 2013 matches the value reported as 'PY' against P3 2014
so for example
PY:=calculate(sum(amount), dim_date(year) = dim_date(year-1) && dim_date(period) = dim_date(period)
I cannot find an exact example of how to construct the DAX formula to achieve this.
I have tried DAXPatterns.com, but that does not have an exact example.
Thanks in advance
Sotn

Answer 1

All built-in time intelligence functions work on the standard Gregorian calendar. The only fiscal calendars that can be handled by the built-in functions are those that end on another day than December 31, but otherwise respect the standard months/quarters.
The problem you are encountering in your [PY] is that CALCULATE() only allows simple boolean predicates (no calculations allowed. FILTER() is the function that allows you to execute arbitrary logic in your CALCULATE() filter arguments.
Try this:
PY:=
CALCULATE(
SUM( <fact table>[amount] )
, FILTER(
ALL( dim_date )
, dim_date[year] = MAX( dim_date[year] ) - 1
&& dim_date[period] = MAX( dim_date[period] )
Here's what's going on in that FILTER(). FILTER() is an iterator - it steps row-by-row through the table that you pass as the first argument, and returns only those rows that meet the criterion you set in the second argument.
We pass ALL( dim_date ). This strips the current filter context from the pivot table and returns every row of dim_date.
In the second argument to FILTER() we define the logic that will determine which rows to return. This comes in two parts, connected with the logical and operator, &&. I will only explain one in detail as we're using the same pattern in each part.
dim_date[year] is referring to the value of the column [year] based on where you are in the table (remember, FILTER() is stepping row-by-row through), so will return one value from the current row based on FILTER()'s iteration.
It then compares that value to what is returned by MAX( dim_date[year] ) - 1. This expression is not evaluated in the current row context of FILTER(), rather in the filter context of the pivot table (in a pivot table your row and column labels are referred
to as rowfilters and columnfilters - this nomenclature can get confusing). Thus, it returns whatever the current year of the pivot table is and subtracts one. When the value of [year] in dim_date matches the value returned by this expression, that row is part
of the set returned by FILTER().
Once FILTER() has completed its iteration, it returns the filtered table, and that table is used to set the filter context that SUM( <fact table>[amount] ) is evaluated in.