Aggregations

Aggregations are provided to the `aggregate` operator and specify how the aggregation should happen. All aggregations take two common arguments:

1. window: Window - argument that specifies the length of the duration across which Fennel needs to perform the aggregation. See how duration is specified in Fennel.
2. into_field: str - the name of the field in the output dataset that corresponds to this aggregation. This

Besides these common arguments, here is the rest of the API reference for all the aggregations:

1. Count

Count computes a rolling count for each group key across a window. It returns 0 by default. Its output type is always int. The count aggregate also takes an optional argument unique which is a boolean. If set to true, counts the number of unique values in the given window. The field over which the count is computed is specified by the of parameter of type str. Count also takes approx as an argument that when set to true, makes the count an approximate, but allows Fennel to be more efficient with state storage. Currently, Fennel only supports approximate unique counts, hence if unique is set to true, approx must also be set to true.

2. Sum

Sum aggregate computes a rolling sum across a window for a given field in the dataset. This field is specified by the of parameter of type str. If no data is available, the default value is 0. Its output type can be int or float depending on the input type (and any other input type will fail sync validation)

3. Average

Same as "Sum", but instead maintains a rolling average in the given window. In addition to the of field, also requires a default value to be specified which is returned when average is queried for a window with no data points. The input types can only be int or float and the output type is always float.

4. Min

Same as "Sum", but instead maintains a rolling minimum in the given window. Requires a default value to be specified. Input type can be int or float and the output type is same as the input type.

5. Max

Identical to "min", but instead maintains a rolling maximum in the given window.

6. LastK

Maintains a list of "items" in the given rolling durations. If no events have been logged, returns an empty list. If input field is of type T the output field is of type List[T]

7. Stddev

Like average, but instead maintains the standard deviation for the population in the given window. The population size used is the count in that given window. The input types can only be int or float and the output type is always float. A default value is required for when there is a window with 0 data points.