contrib/redis/rueidis, contrib/valkey: improve pipeline span resource names

[snutiise]

Package Name

contrib/redis/rueidis, contrib/valkey-io/valkey-go

Package Version(s)

latest (v2)

Describe the feature you'd like

Deduplicate and sort command names alphabetically for pipeline span resource names, and add structured tags for detailed pipeline information

• Resource: "DEL GET SET" (unique command types only, sorted)
• redis.pipeline.length: Total number of commands (e.g. 150)
• redis.pipeline.command_counts: Breakdown per command type (e.g. "DEL:5 GET:100 SET:45")

This produces low-cardinality resource names that group well in Datadog APM, while preserving full pipeline details via tags.

Is your feature request related to a problem?

Currently, DoMulti / DoMultiCache / DoMultiStream pipeline spans produce resource names by concatenating up to 5 commands with their full arguments. For example

GET key1 SET key2 value2 GET key3 SET key4 value4 GET key5

This causes two issues

1. High cardinality: The same pipeline pattern produces different resource names depending on key arguments, making it hard to aggregate in APM.
2. Arbitrary truncation: Only the first 5 commands are shown, which may not represent the full pipeline.

Describe alternatives you've considered

• Join all command verbs without deduplication: Simple but produces long resource names for large pipelines and still has higher cardinality than needed.
• Trace each command as individual spans: Provides per-command visibility but loses the pipeline-level view and increases span volume significantly.

Additional context

I apologize for opening a PR before this discussion. I've already submitted an implementation at #4511 — happy to adjust based on feedback here.