---
title: `count_window`
url: https://nexyron.tech/docs/cypher/functions/temporal/count-window
description: Returns event count windows using the same list window primitive. It is documented separately because time construction, sequence ordering, window parameters, and as of behavior ma
---

# `count_window`

## Generic Description

Returns event-count windows using the same list-window primitive. It is documented separately because time construction, sequence ordering, window parameters, and as-of behavior materially change analytical meaning.

Simple example:

```cypher
WITH [10, 13, 9, 16] AS readings
RETURN count_window(readings, 3, 1) AS windows
```

## Consumer-Level Explanation

Use it when the business wording is count-window oriented but downstream logic needs windows. Prefer this function when the temporal assumption should be visible to the planner, especially for event streams, freshness checks, and sequence features.

## More Detailed Explanation

`count_window` works on ordered temporal or sequence-style values inside Cypher. The practical contract is to assemble the sequence deliberately, choose the time or window parameter explicitly, and return named fields so downstream planner tooling does not invent unsupported window syntax.

## Advanced Example

This example names the event stream and operational context before applying `count_window`, which helps planner tooling preserve the intended sequence semantics.

```cypher
WITH 'sensor-1' AS subject_id,
     [{ts: 1735689600000, value: 10.0, state: 'ok'}, {ts: 1735689660000, value: 13.0, state: 'ok'}, {ts: 1735689840000, value: 9.0, state: 'warn'}, {ts: 1735690200000, value: 16.0, state: 'ok'}] AS readings
RETURN subject_id,
       count_window(readings, 3, 1) AS windows
```

## Real Use Cases

-   sessionizing user or device events before feature extraction
-   smoothing, rate calculation, or anomaly screening on irregular telemetry
-   joining observations to the correct historical context without future leakage

## Real Limitations And Tradeoffs

-   sequence functions assume the input list is already the intended ordered series
-   window size, inactivity threshold, and tolerance values are modeling choices that change results materially
-   these helpers are not a replacement for registering temporal properties when the workload needs cataloged timeseries history