Optional Metrics

In InfraSonar, you may encounter metrics with a question mark (?) appended to their name (e.g., item.some_metric?). This visual cue indicates that the metric is optional and its value can potentially be null. Unlike regular metrics which are guaranteed to always have a value, optional metrics might not always be present or valid in the incoming data.

When working with such metrics in your expressions, InfraSonar requires you to explicitly mark them with a question mark (e.g., item.payload?). While the question mark itself doesn't change how InfraSonar evaluates the expression at runtime, its presence generates awareness. It visually reminds you that the value might be null, prompting you to handle this possibility in your logic.

Consider the following expression, which might be used with data from the http or httpsvc collector:

// Test if the payload contains the word "ERROR"
case "ERROR" in item.payload.upper():
    return ERROR, "Payload contains an error"

If the payload metric is optional and you omit the question mark, InfraSonar will prevent you from saving the expression, displaying a clear message like this:

metric `payload` is nullable and must be marked with a question mark like `item.payload?

To correctly handle the optional payload and allow the expression to be saved, you must include the question mark:

case "ERROR" in item.payload?.upper():
    return ERROR, "Payload contains an error"

It's crucial to understand how InfraSonar expressions behave when an optional metric (marked with ?) is actually null.

When an optional metric like item.payload? is null, any attempt to access its properties or call its methods (like item.payload?.upper()) will automatically cause that specific part of the expression to evaluate as false. The operation on the null value is simply skipped, and the overall condition will not be met.

In essence, if a nullable metric is null, any operation on it (unless it's an explicit null check) will result in false.

If you want to explicitly detect and act upon a missing or null metric, you must test for null directly:

case item.payload? == null:
    return ERROR, "Payload is missing"

This ensures you have full control over how your expressions respond to the absence of data, allowing you to differentiate between a metric being present but not matching a condition, and a metric being entirely null.