[sql] Parse db.operation.name and db.collection.name from db.query.text

[alanwest]

Component

OpenTelemetry.Instrumentation.SqlClient

What is the expected behavior?

The attributes db.operation.name and db.collection.name are conditionally required "if readily available".

Alternatively, they MAY be parsed from db.query.text. See relevant footnotes.

[3]: If readily available. The collection name MAY be parsed from the query text, in which case it SHOULD be the first collection name found in the query.
[6]: If readily available. The operation name MAY be parsed from the query text, in which case it SHOULD be the first operation name found in the query.

These attributes are very valuable to end users.

Parsing db.query.text is a costly operation. What this issue does not attempt to address is whether parsing operation/collection from db.query.text should be on by default. For now, it should be implemented as opt-in, though later we should decide whether it should be on by default.

[hannahramadan]

Ruby has an in-progress PR to add db.collection.name. Here is the regex:

# Capture the first word (including letters, digits, underscores, & '.', ) that follows common table commands
TABLE_NAME = /\b(?:(?:FROM|INTO|UPDATE)|(?:(?:CREATE|DROP|ALTER)\s+TABLE(?:\s+IF\s+(?:NOT\s+)?EXISTS)?))\s+["]?([\w.]+)["]?/i

Adding this note as more of a resource/ point of discussion when this ticket gets worked on. Might be a better way to grab this.