Aggregations in DataLens
- SQL basic concepts
- Data aggregation in DataLens
This section describes how data aggregation and grouping work in DataLens, how to correctly formulate expressions with aggregations, and also gives examples of SQL queries with aggregations for ClickHouse.
The file Selling.csv is used as the source of data on sales in cities.
SQL basic concepts
What is aggregation?
Let's take a look at the
Selling table with data on sales in cities:
|4||San Francisco||Office Supplies||2014-03-11||8||3||0.10|
Aggregation is used to calculate resulting values. Aggregation process essentially means converting a large set of strings into a single value. In SQL, special aggregate functions are used for this. The most common functions include
COUNT. Each of these functions operates with column values in a given table, which results in a single value. For example, for the
SUM function, this is the sum of all the column values, for
AVG — the average value, for
MAX — the highest value.
There are two ways to use aggregate functions:
- Aggregate functions return the resulting value for a single group.
- Aggregate functions return the resulting value for multiple groups.
Aggregation for a single group
In this case, aggregate functions calculate and return a single resulting value for all rows that are combined into a single group.
For example, to get the sum of values in the
Sales column from the table
Selling, you need to run a query:
SELECT sum("Sales") FROM "Selling"
Aggregation for multiple groups
In this case, the entire set of rows returned by the query is split into separate groups. The group is determined by the value of the column for which the grouping is performed. In this way, aggregate functions calculate and return the resulting value separately for each group.
In SQL, the columns for grouping are specified in the
GROUP BY section. Grouping can be performed either by one or several columns.
Calculating the sales amount for each city:
SELECT "City", sum("Sales") FROM "Selling" GROUP BY "City"
Calculating the sales amount for each city and product category:
SELECT "City", "Category", sum("Sales") FROM "Selling" GROUP BY "City","Category"
|San Francisco||Office Supplies||8|
When aggregating with grouping by several groups, keep in mind the following limitations:
The columns used for selection must be specified in the
SELECT "City", sum("Sales") FROM "Selling" GROUP BY "City"
SELECT "City", sum("Sales") FROM "Selling" GROUP BY "Category"
Aggregated and non-aggregated expressions cannot be used at the same query level:CorrectIncorrect
SELECT "City", sum("Sales") as "Detroit Sales" FROM "Selling" WHERE "City" = 'Detroit' GROUP BY "City"
SELECT if("City" = 'Detroit', sum("Sales"), 0) as "Detroit Sales" FROM "Selling" GROUP BY "Category"
You can apply filtering in queries with grouping and aggregate functions. You can filter both the original pre-aggregation set of rows and the resulting values calculated by aggregate functions.
The filtering conditions for the original set of rows are specified in the
SELECT "City", sum("Sales") FROM "Selling" WHERE "Category" = 'Furniture' GROUP BY "City"
To filter aggregated values, specify a condition in the
HAVING section. In this case, the query returns only those rows in which the resulting value of the aggregate function satisfies the specified condition:
SELECT "City", "Category", sum("Sales") FROM "Selling" GROUP BY "City","Category" HAVING sum("Sales") >= 10
Data aggregation in DataLens
Dimensions and measures
In DataLens, aggregation is performed using dimensions and measures.
Measure: A dataset field with a specified aggregation type (for example, sum, average, or quantity). In the dataset and in the wizard, measures are displayed in blue. Usually, a measure is a business metric scrutinized by different slices or groupings, such as revenue, number of customers, or average customer bill.
Dimension: A dataset field without the specified aggregation, such as a region, a product, or category. In the dataset and in the wizard, dimensions are displayed in green. Dimensions are used to group a query in the chart (the GROUP BY section in SQL). To group data in the chart, you need to drag the dimension to the desired section.
Methods to create measures
You can measures both at the dataset and the chart level. We recommend adding dimensions at the dataset level. This allows you to reuse them in different charts and speed up chart rendering.
Creating measures at the dataset level
You can add a measure at the dataset level in the following ways:
In the dataset creation interface, open the Fields tab and select the aggregation type for the field in the Aggregation column.
In the data creation interface, add a calculated field using aggregate functions. For more information, see How to create a calculated field. In the formula of the calculated field, you can use substitution of other measures.
When you create a calculated field using an aggregate function, it is assigned the Auto aggregation type, which cannot be changed.
Creating measures at the chart level
You can add a measure at the chart level in the following ways:
In the wizard, drag the dimension to the section for measures and select the aggregation type. The field color changes from green to blue.
Measures can consist of more than one aggregate function and have more complex expressions. For example, in this chart, to calculate the average sales amount for the day, the
Sales per day measure calculated with the formula
SUM([Sales])/COUNTD([Date]) is used.
Using dimensions and measures in charts
When building any chart in DataLens, data is grouped and aggregated.
Let's look at table Selling, for we need to calculate the sales amount (
Sales) for all dates (
Date) separately for each city (
City). To do this, you need to group the data by the
City field. When grouping, rows are combined in such a way that each
City value occupies one row. All source rows where the
City values match and are equal, form a group of rows. As a result, there are three groups for which the
Sales value will be summed up:
- Rows 1 and 5 will be added to the
- Rows 2, 3, and 6 — to
- Rows 4, 7, and 8 —
For example, in the Column chart, the result looks like this:
You can group by several fields rather than one. In this case, each row is defined by a set of values of all fields by which grouping is performed. There will be as many rows in the final result as there are unique sets of such values.
For example, when adding the
Category field to the Colors section, it will affect grouping. The chart looks like this:
Measures in the Colors section also affect data grouping.
In some chart sections, you can drag only a dimension or only a measure. This depends on the chart type. For example, in the Y section of the column chart, you can only drag a measure. If you drag a dimension to this section, it'll be automatically converted to a measure as a result of the Number of unique aggregation.
Like in SQL, in DataLens, you can't use aggregated and non-aggregated values in the same expression.
For example, in the chart with groupings by the
Category dimensions, you can't add the
SUM([Sales]) * (1 - [Day's discount]) measure to calculate the sales amount taking with discounts. In this case, the
Category dimensions determine group splitting, and therefore have fixed values in each group. For each group, you can calculate the
SUM([Sales]) value. However, the
Day's discount field is neither aggregation nor a measure within the group. It doesn't have a fixed value — it can vary from row to row in the group. Therefore, it's impossible to determine what specific value of the
Day's discount field needs to be selected to calculate the
SUM([Sales]) * (1 - [Day's discount]) measure for each group. Thus, the expression
SUM([Sales]) * (1 - [Day's discount]) can't be calculated. In DataLens, such cases result in the
Inconsistent aggregation among operands error.
You can prevent this error in different ways:
Day's discountfield to the dimension section. In this case, data is grouped by the
Day's discountdimensions, so the fixed value of the
Day's discountfield is used for each group to calculate the value of the
SUM([Sales]) * (1 - [Day's discount])measure.
Specify the aggregation type for the
Day's discountfield. In this case, this field will become a measure and the original formula will be correct.
Filtering dimensions and measures
In charts, you can filter the values of dimensions and measures. To do this, drag a dimension or a measure to Filters and set filtering conditions:
Sales by city in the
Sales by city and category, where the
SUM([Sales]) measure is greater than or equal to 10:
When creating calculated fields in a formula, you can use pre-existing measures. These measures can be set either using a formula or the dataset creation interface. The created calculated field is assigned the Auto aggregation type.
[TotalSales] field is set using the
SUM([Sales]) aggregate function. Then the
[TotalSales]/10 calculated field is assigned the Auto aggregation type.
If the measure set using the dataset creation interface is substituted in the calculated field, you can redefine the aggregation type. To do this, use a function with a different aggregation type in the formula.
[Sales] field, the Amount aggregation type is set in the dataset creation interface. Then the
AVG([Sales]) calculated field is assigned the Auto aggregation type and calculated as an average. The Amount aggregation will be ignored.