Announcements
NEW Customer Portal: Initial launch will improve how you submit Support Cases. FIND OUT MORE
Former Employee

A Primer on Set Analysis

Set analysis is one of the more complex things you can use in QlikView or Qlik Sense. Its syntax is often perceived as complicated and there are some misunderstandings around it. So here is my short explanation.

Set analysis is a way to define an aggregation scope different from current selection. Think of it as a way to define a conditional aggregation. The condition, however, is in itself like a selection that is evaluated before the cube (the chart) is expanded. Hence, it is not possible to have a condition that is evaluated row by row.

The set analysis is used inside an aggregation function, e.g. in an expression like

Sum(Sales)

The first step is to add the markers for the set analysis – the curly brackets:

Sum( {…} Sales)

These define the record set over which the aggregation should be made. Inside, you can use different identifiers and operators, e.g. ‘\$’ for records implied by current selection, ‘1’ for all records, ‘1-\$’ for all excluded records, etc.

A set of records that you can define by a simple selection is called natural set. Not all record sets are natural; for instance, {1-\$} cannot always be defined through a selection.

The next step is often to add a set modifier, which is defined by angle brackets. The set modifier adds or changes a selection. It can be used on any natural set and consists of a list of fields, where each field can have a new selection:

Sum( {\$<…>} Sales)

The next step is to define the element set for a field; the set of field values that defines the selection. The element set could be a field reference or a set function, P() or E(). It is more common, though, that it is an explicit list of field values or a search, and then you need the curly brackets to define the element set:

Sum( {\$<Date={…}>} Sales)   or   Sum( {\$<Date=P(…)>} Sales)

A search can be defined through double quotes. This way, field values that match the search string will be selected:

Sum( {\$<Date={"…"}>} Sales)

Do not use single quotes to initiate a search here! Single quotes should denote literals, i.e. explicit field values. (Yes, today single quotes work as a search, but this is a bug that one day will be fixed…)

Often a numeric search is made, defined by a leading relational operator. Then, field values will be selected based on a comparison:

Sum( {\$<Date={"<=…"}>} Sales)

To make it worse, the value to which the field values are compared is often a calculated one. And in order to get a calculated value into the expression, a dollar expansion is needed.:

Sum( {\$<Date={"<=\$(…)"}>} Sales)

Inside the dollar expansion, you need an expression that starts with an equals sign and contains an aggregation function, e.g.:

Sum( {\$<Date={"<=\$(=Max(Date))"}>} Sales)

This aggregation function is evaluated globally, before the cube is expanded.

As you can see, there are many levels of a set expression, and many pairs of brackets and delimiters that need to match. When you write set expressions, you should always write both brackets directly, and then continue with the expression between them. This way you will avoid simple syntax errors.

Good luck with your set analysis expressions!

HIC

Further reading related to this topic:

Why is it called Set Analysis?

Dates in Set Analysis

Excluding values in Set Analysis

Introduction to Set Analysis (video) - Part 1

Partner - Creator III

Hi Henric,

thank you for another brilliant blog post!

I have a question regarding to: "Do not use single quotes here! Single quotes should denote literals, i.e. explicit field values. (Yes, today single quotes work as a search, but this is a bug that one day will be fixed…)"

I am just going through Advanced Topics in Design and Development course book and I found the following there:

String/Numeric values

To select string or numeric values, place the value inside {} and single quotes (double quotes also work, but using them is not best practice).

=Sum({\$<Year={'2011'}>}LineSalesAmount)

=Sum({\$<Country={'USA'}>}LineSalesAmount)

same is then shown for Conditional values

=Sum({\$<OrderDate={'>01/01/2011'}>}LinesSalesAmount)

So where is the true? Should I trust double quotes or single quotes? Can you please clarify?

T.

31,518 Views
Former Employee

In all other places in QlikView/Qlik Sense, single quotes mean literals and double quotes mean something else. The intention for Set Analysis was when it was designed, to follow the same logic. However, the implementation today is that single or double quotes doesn't matter in Set Analysis. But there is a consensus in the Qlik development organization that this is a bug that should be corrected.

For the above examples, this means that

=Sum({\$<Year={'2011'}>} LineSalesAmount)

=Sum({\$<Country={'USA'}>} LineSalesAmount)

are both correct. Both 2011 and USA are literals, so single quotes are correct, if you want a case sensitive comparison.

The third example is correct if you compare how QlikView de facto works, but incorrect given the design intention. Our education department has (correctly) described how it works, but not how it should work.

I'll talk to our education department and make sure the training material is changed.

HIC

31,518 Views
Partner - Creator III

Thank you Henric! It's all clear now.

31,518 Views
Specialist III
31,518 Views
Contributor III

Great! Clear explanation! Thanks.

31,518 Views
Not applicable

DR. Henric Cronström,

A million Thanks for your wonderful insights. This is a very important and complicated concepts and through examples you have made it clear. I highly appreciated it.

I hope to stay in touch for Data Modeling, Incremental load among other challenging questions.

31,518 Views
Contributor III

Hi Henric,

Great post indeed!

I have just a little doubt about how Qlik constructs the global set of data before raise the cube.

Let’s say for instance, that we have a straight table with many expressions. Each one of them have their own set analysis which may be different one from another in some way.

So, When you say: “This aggregation function is evaluated globally, before the cube is expanded”. Do you mean that Qlikview makes the union of all differents sets obtained from each set analysis definition (inside every single aggregation expression from the chart)  in order to obtain a global set of data before expand de cube?

MR

26,337 Views
Creator II

Clear !!

26,337 Views
Former Employee

mrooney‌:

Yes, sort of... In a way, the calculations need to be made on the union of the sets. However, it is slightly more intricate, and not so simple...

If you have two calculations: Sum({BM1} Sales) and Count({BM2} Products), then you have two aggregation functions that operate on two different tables: "Sales" is typically in the fact table, and "Products" is in a dimension table. This means that the two aggregations are independent and not affected by the set in the other aggregation. For example, the count of products does not depend on BM1. Hence, two separate aggregations.

But if both aggregations are calculated from the same table, then it is optimized so that only one pass is made over this table, and both sets are taken into account.

HIC

26,337 Views
Creator II