[timtofan]

What issue is this change attempting to solve?

Why

groupByNode and groupByNodes panic when a callback returns a non-nil slice containing zero series. The existing nil check lets that result reach r[0], turning expression evaluation into an index-out-of-range error. The expression evaluator can return this result shape when its context is cancelled.

How does this change solve the problem? Why is this the best approach?

Architecture

This is a local change to expression grouping. Group construction and callback evaluation keep their current boundaries and error-handling behavior. Check the result length before renaming its first series and appending the callback output.

Before / after

Callback result	Before	After
Nil slice	Skip the group	Skip the group
Non-nil empty slice	Panic at r[0]	Skip the group and continue
One or more series	Rename the first series to the group key and append the output	Same behavior

A non-empty series whose points are all null remains present. This guard operates on the number of returned series, rather than their point values.

How can we be sure this works as expected?

• The new empty-result regression reproduces the index-out-of-range panic on unchanged master.
• Regression cases cover both function names, nil and empty callback results, cancellation-shaped results, and a valid group following an empty one. Additional cases cover empty input and preservation of an all-null aggregation.
• go test ./pkg/expr/functions/groupByNode -count=1 passed.
• make test passed: 50 packages tested with -tags cairo,nativegzip -race.
• golangci-lint v2.14.0 passed for the full repository with zero issues.