RollingMax / Computation Layer

Calculate the rolling (moving) maximum over a sliding window of values. Similar to pandas.rolling().max() or R's rollapply(FUN=max).

Mathematical form: $y_{i} = j = i - w + 1 max i (x_{j} \cdot w_{j})$ where $w$ is the window size and $w_{j}$ are optional weights.

Common applications:

Financial analysis (peak prices over time)
Environmental monitoring (maximum pollution levels)
Performance tracking (best results in period)
Quality control (peak measurements)
Network monitoring (peak usage)
Sensor data analysis (maximum readings)
Risk assessment (worst-case scenarios)

Note: Results include null values for windows with insufficient observations based on MinPeriods setting.

f32, f64

Table

f32, f64

Select

column

The numeric column to compute rolling maximums for. Common input types:

Price series (stock prices, commodities)
Measurement data (temperatures, pressures)
Performance metrics (response times, scores)
Resource usage (CPU, memory, bandwidth)
Sensor readings (IoT data streams)

WindowSize

u32

Number of observations in each sliding window. Common values:

7: Weekly maximum from daily data
30: Monthly maximum from daily data
252: Annual maximum for trading days
60: Hourly maximum from minute data

Must be ≥ 1. Larger windows provide more smoothing.

Weights

[f64, ...]

Optional weights for window values. Length must match WindowSize. Uses:

Exponential weighting (recent data more important)
Custom smoothing patterns
Seasonal adjustments
Trend emphasis

If not provided, all values weighted equally.

MinPeriods

oneof

AsWindow

Set minimum required observations equal to window size. Ensures complete windows only. Use when you need strict window completeness, such as:

Regulatory compliance calculations
Complete data requirements
Strict quality control measures
Full-window-only analysis

Period

Value

u32

Minimum number of non-null observations required in window (Value) to compute result. Must be ≥ 1 and ≤ window size. Examples:

1: Calculate with any non-null value
window_size/2: Require half window
window_size: Require complete window

Controls trade-off between data availability and calculation reliability.

Center

bool

false

When true, aligns the result label to the center of the window instead of the end. Effects:

true: Better for identifying peak timing
false: Better for real-time processing

Important for temporal alignment in analysis.

Name for the new column. If not provided, the system generates a unique name. If AsColumn matches an existing column, the existing column is replaced. The name should follow valid column naming conventions.