Time series functions

Time series functions provide various ways to look up values corresponding to a specific time or offset along a given time axis.

In a certain way this functionality is similar to the window function LAG. The main difference is that LAG is indifferent to the actual values of the dimensions being used, and operates over positional offsets specified in rows, while time series functions use specific values and value offsets in date/time units like days, hours or seconds. This makes them sensitive to missing values in data. As a result of this AGO(SUM([Sales]), [Date], "year") will return NULL if the same-date row for the previous year is missing.

Syntax

Time series functions support extended syntax:

<FUNCTION_NAME>(
    arg1, arg2, ...

    [ BEFORE FILTER BY filtered_field_1, ... ]
    [ IGNORE DIMENSIONS dimension_1, ... ]
)

BEFORE FILTER BY

If any fields are listed in BEFORE FILTER BY, then the function is calculated before data is filtered using these fields. Let's say you are trying to calculate the value of [Sales] a year ago from today's date:

AGO([Sales], [Date], "year", 1)

and also have a filter in the chart limiting the dates to a specific year (2018 < [Year] <= 2019, where Year is calculated as YEAR([Date])). All data pertaining to the year 2018 will be omitted from the result. Because of this the function will return NULL.

If BEFORE FILTER BY is added to the function:

AGO([Sales], [Date], "year", 1 BEFORE FILTER BY [Year])

it will return the value of [Sales].

The date/time dimension specified as the second argument of time series functions is implicitly added to the BEFORE FILTER BY clause. For example, the following to formulas are equivalent:

AGO([Sales], [Date], "week")
AGO([Sales], [Date], "week" BEFORE FILTER BY [Date])

IGNORE DIMENSIONS

The IGNORE DIMENSIONS clause allows the exclusion of dimensions from the search criteria. When looking up a value for a certain date, the search is done by matching dimension values in each row against the values of the same dimensions in the original row. If any of the other dimensions correlate with the date dimension, then the data query may return an empty result (NULL).

For example, if the data request contains [Date], [Month] and AGO([Sales], [Date], "month"), then it will be impossible to find a row with [Date] being a month earlier, but the value of [Month] being the same as in the current row. This will result in AGO always returning NULL.

To get the correct value of [Sales] exclude [Month] using the IGNORE DIMENSIONS clause:

AGO([Sales], [Date], "month" IGNORE DIMENSIONS [Month])

AGO

Syntax:
AGO( measure, date_dimension [ , unit [ , number ] ] )
or
AGO( measure, date_dimension [ , unit [ , number ] ] [ BEFORE FILTER BY ... ] [ IGNORE DIMENSIONS ... ] )

Re-evaluate measure for a date/time with a given offset. The date_dimension argument is the dimension along which the offset is made. The number argument is an integer. It can be negative. The unit argument takes the following values:

  • "year";
  • "month";
  • "day";
  • "hour";
  • "minute";
  • "second".

Can also be used as AGO( measure, date_dimension, number ). In this case, the third argument is interpreted as the number of days.

See also AT_DATE, LAG.

AT_DATE

Syntax:
AT_DATE( measure, date_dimension, date_expr )
or
AT_DATE( measure, date_dimension, date_expr [ BEFORE FILTER BY ... ] [ IGNORE DIMENSIONS ... ] )

Re-evaluate measure for a date/time specified by date_expr. The date_dimension argument is the dimension along which the offset is made.

See also AGO, LAG.

Previous
Next