Aggregate function combinators¶
The name of an aggregate function can have a suffix appended to it. This changes the way the aggregate function works.
The suffix -If can be appended to the name of any aggregate function. In this case, the aggregate function accepts an extra argument – a condition (Uint8 type). The aggregate function processes only the rows that trigger the condition. If the condition was not triggered even once, it returns a default value (usually zeros or empty strings).
quantilesTimingIf(level1, level2)(x, cond),
argMinIf(arg, val, cond) and so on.
With conditional aggregate functions, you can calculate aggregates for several conditions at once, without using subqueries and
JOINs. For example, in Yandex.Metrica, conditional aggregate functions are used to implement the segment comparison functionality.
The -Array suffix can be appended to any aggregate function. In this case, the aggregate function takes arguments of the 'Array(T)' type (arrays) instead of 'T' type arguments. If the aggregate function accepts multiple arguments, this must be arrays of equal lengths. When processing arrays, the aggregate function works like the original aggregate function across all array elements.
sumArray(arr) - Totals all the elements of all 'arr' arrays. In this example, it could have been written more simply:
uniqArray(arr) – Counts the number of unique elements in all 'arr' arrays. This could be done an easier way:
uniq(arrayJoin(arr)), but it's not always possible to add 'arrayJoin' to a query.
-If and -Array can be combined. However, 'Array' must come first, then 'If'. Examples:
quantilesTimingArrayIf(level1, level2)(arr, cond). Due to this order, the 'cond' argument won't be an array.
If you apply this combinator, the aggregate function doesn't return the resulting value (such as the number of unique values for the uniq function), but an intermediate state of the aggregation (for
uniq, this is the hash table for calculating the number of unique values). This is an
AggregateFunction(...) that can be used for further processing or stored in a table to finish aggregating later.
To work with these states, use:
- AggregatingMergeTree table engine.
- finalizeAggregation function.
- runningAccumulate function.
- -Merge combinator.
- -MergeState combinator.
If you apply this combinator, the aggregate function takes the intermediate aggregation state as an argument, combines the states to finish aggregation, and returns the resulting value.
Merges the intermediate aggregation states in the same way as the -Merge combinator. However, it doesn't return the resulting value, but an intermediate aggregation state, similar to the -State combinator.
Converts an aggregate function for tables into an aggregate function for arrays that aggregates the corresponding array items and returns an array of results. For example,
sumForEach for the arrays
[3, 4, 5]and
[6, 7]returns the result
[10, 13, 5] after adding together the corresponding array items.
Fills the default value of the aggregate function's return type if there is nothing to aggregate.
SELECT avg(number), avgOrDefault(number) FROM numbers(0)
┌─avg(number)─┬─avgOrDefault(number)─┐ │ nan │ 0 │ └─────────────┴──────────────────────┘
null if there is nothing to aggregate. The return column will be nullable.
SELECT avg(number), avgOrNull(number) FROM numbers(0)
┌─avg(number)─┬─avgOrNull(number)─┐ │ nan │ ᴺᵁᴸᴸ │ └─────────────┴───────────────────┘
-OrDefault and -OrNull can be combined with other combinators. It is useful when the aggregate function does not accept the empty input.
SELECT avgOrNullIf(x, x > 10) FROM ( SELECT toDecimal32(1.23, 2) AS x )
┌─avgOrNullIf(x, greater(x, 10))─┐ │ ᴺᵁᴸᴸ │ └────────────────────────────────┘
Lets you divide data into groups, and then separately aggregates the data in those groups. Groups are created by splitting the values from one column into intervals.
<aggFunction>Resample(start, end, step)(<aggFunction_params>, resampling_key)
start— Starting value of the whole required interval for
stop— Ending value of the whole required interval for
resampling_keyvalues. The whole interval doesn't include the
step— Step for separating the whole interval into subintervals. The
aggFunctionis executed over each of those subintervals independently.
resampling_key— Column whose values are used for separating data into intervals.
- Array of
aggFunctionresults for each subinterval.
people table with the following data:
┌─name───┬─age─┬─wage─┐ │ John │ 16 │ 10 │ │ Alice │ 30 │ 15 │ │ Mary │ 35 │ 8 │ │ Evelyn │ 48 │ 11.5 │ │ David │ 62 │ 9.9 │ │ Brian │ 60 │ 16 │ └────────┴─────┴──────┘
Let's get the names of the people whose age lies in the intervals of
[60,75). Since we use integer representation for age, we get ages in the
[30, 59] and
To aggregate names in an array, we use the groupArray aggregate function. It takes one argument. In our case, it's the
name column. The
groupArrayResample function should use the
age column to aggregate names by age. To define the required intervals, we pass the
30, 75, 30 arguments into the
SELECT groupArrayResample(30, 75, 30)(name, age) FROM people
┌─groupArrayResample(30, 75, 30)(name, age)─────┐ │ [['Alice','Mary','Evelyn'],['David','Brian']] │ └───────────────────────────────────────────────┘
Consider the results.
Jonh is out of the sample because he's too young. Other people are distributed according to the specified age intervals.
Now let's count the total number of people and their average wage in the specified age intervals.
SELECT countResample(30, 75, 30)(name, age) AS amount, avgResample(30, 75, 30)(wage, age) AS avg_wage FROM people
┌─amount─┬─avg_wage──────────────────┐ │ [3,2] │ [11.5,12.949999809265137] │ └────────┴───────────────────────────┘