EXPAND ALL
- Home
- About Pixie
- Installing Pixie
- Using Pixie
- Tutorials
- Reference
Summaries describe distributions by recording quantile values (aka percentile values) such as p50, p90 and p99. User specify as many of these point values as they like, labeling them with the float equivalent position in the distribution. Ie p50 would be labeled 0.5
, p99 as 0.99
. The source DataFrame must have a time_
column of type TIME64NS
or the compiler will throw an error. The unit of this metric will be inferred from the SemanticType of the column value
.
Name | Type | Description |
---|---|---|
name | string | The name of the metric. |
count | Column | The column of the count of elements inside each summary. Must be INT64 |
sum | Column | The column of the sum of elements in inside each summary to use for the Summary. Must be FLOAT64 |
quantile_values | Dict[float, Column] | The mapping of the quantile value to the DataFrame column containing the quantile value information. Must be FLOAT64. |
description | string, optional | A description of what the metric tracks. |
attributes | Dict[string, Column|string], optional | A mapping of attribute name to a string or the column that stores data about the attribute. |
unit | string, optional | The unit string to use for the metric. If not specified, will attempt to use the Semantic Type of the quantile values to infer the unit string. |