Values passed to lambda_metric of incorrect type do not raise an exception.

[aaron-csetter]

Expected Behavior

When a value passed to lambda_metric() is of an incorrect value or type (not something that can be sent as a metric), a TypeError or ValueError should be raised.

Actual Behavior

In the example below, neither call actually goes through and no exception is raised in the main thread. This makes debugging a nightmare.

Steps to Reproduce the Problem

@datadog_lambda_wrapper
def lambda_handler(lambda_event, _context):
    lambda_metric("should.work", 1)
    lambda_metric("should.raise.exception", "oops")

Specifications

• Datadog Lambda Layer version: v73 and v75
• Python version: 3.10

Stacktrace

I wish there was one.

[astuyve]

We received a similar request here, and the answer will be similar here. We never want to cause a lambda function execution to halt, so raising a TypeError or ValueError will be incorrect here.

We are open to adding validation, ideally we'd do so at the agent level so that it would cover all runtimes. Presumably the correct response here would be to log an error? If we did this at the WARN level, would that seem appropriate?

Thanks!

[nhulston]

Fixed in #553, will be out in the next release (v106)

[nhulston]

Thanks for opening this. It's been fixed in the latest release
https://github.com/DataDog/datadog-lambda-python/releases/tag/v6.106.0