Skip to content

delay

Input Stack:
n: Int
expr: TimeSeriesExpr
 ⇨ Output Stack:
TimeSeriesExpr
 

Delays the values in a time series by shifting them forward by N time steps. This creates a lag effect where each value appears N intervals later than it originally occurred. Unlike :offset, this operates on the computed time series data rather than changing the time window used for data fetching.

Parameters

  • expr: The time series expression to delay
  • n: The number of time steps to delay the values (integer)

Examples

Delaying a time series by 5 time steps:

BeforeAfterCombined
name,requestsPerSecond,:eq,
:sum

name,requestsPerSecond,:eq,
:sum,
5,:delay

name,requestsPerSecond,:eq,
:sum,
:dup,
5,:delay

The "Combined" example shows both the original and delayed series overlaid, demonstrating how the delayed series follows the same pattern but shifted later in time.

Key Differences from :offset

  • :delay: Operates on computed time series values, shifting them within the same time window
  • :offset: Changes the time window used for data fetching, looking at historical data
  • Streaming compatibility: :delay works with streaming queries, :offset does not

Use Cases

  • Change detection: Compare current values with slightly delayed versions to detect trends
  • Alerting: Detect when current metrics deviate from recent patterns
  • Trend analysis: Identify lead/lag relationships between different metrics
  • Smoothing effects: Create visual comparisons between immediate and delayed signals

Since: 1.6

  • :offset - Shift the time window for data fetching (complementary approach)
  • :dup - Duplicate series for comparison (commonly used with delay)
  • :sub - Calculate differences between original and delayed series
  • :time - Generate time-based values for reference