Normalize "IN(?, ?, ...)" substrings in db.statement span attribute to "IN(?)"

[swar8080]

Is your feature request related to a problem? Please describe.

Hello, we're looking to use db.statement as an attribute in span metrics to get RED metrics for each query

Currently, SQL queries using IN with a variable number of values will each generate a different db.statement, varying only by the number of ? in the IN part. That can cause high cardinality span metrics and makes the metrics difficult to analyze

Describe the solution you'd like

Replace all IN(...) statements in SQL db.statements with a constant value like IN( ? )

Describe alternatives you've considered

Using the otel collector to normalize the query string

Additional context

Let me know if this would need to first be approved as a semantic convention

[trask]

this makes sense to me

possibly IN(?, ...) to be more explicit that it could be one or more elements

Let me know if this would need to first be approved as a semantic convention

I don't think so, as there's no specification for how to normalize queries at this point

[laurit]

possibly IN(?, ...) to be more explicit that it could be one or more elements

I suspect that backends might attempt to parse these queries so it might be better to ensure that normalization does not introduce syntax errors

[swar8080]

Would this new behaviour need to be turned on with configuration or a feature flag? Just wondering in case I or someone else ends up implementing this