Add present_over_time

[darshanime]

Signed-off-by: darshanime deathbullet@gmail.com

closes #8255

[darshanime]

ref: dc8f505#r53260620

[darshanime]

already defined here

prometheus/promql/engine.go

Line 50 in f5655c4

env = "query execution"

[roidelapluie]

Thanks for this pull request. It seems that this is a complex approach, I would implement it like other functions:

func funcPresentOverTime(vals []parser.Value, args parser.Expressions, enh *EvalNodeHelper) Vector {
        return aggrOverTime(vals, enh, func(values []Point) float64 {
                return 1
        })
}

[roidelapluie]

Thanks! I will let a few more days for @juliusv to chime in.

[juliusv]

I believe this copied comment only applies to absent_over_time(), for which this engine optimization exists:

prometheus/promql/engine.go

Lines 1366 to 1402 in 7337ecf

	// The absent_over_time function returns 0 or 1 series. So far, the matrix
	// contains multiple series. The following code will create a new series
	// with values of 1 for the timestamps where no series has value.
	if e.Func.Name == "absent_over_time" {
	steps := int(1 + (ev.endTimestamp-ev.startTimestamp)/ev.interval)
	// Iterate once to look for a complete series.
	for _, s := range mat {
	if len(s.Points) == steps {
	return Matrix{}, warnings
	}
	}
	found := map[int64]struct{}{}
	for i, s := range mat {
	for _, p := range s.Points {
	found[p.T] = struct{}{}
	}
	if i > 0 && len(found) == steps {
	return Matrix{}, warnings
	}
	}
	newp := make([]Point, 0, steps-len(found))
	for ts := ev.startTimestamp; ts <= ev.endTimestamp; ts += ev.interval {
	if _, ok := found[ts]; !ok {
	newp = append(newp, Point{T: ts, V: 1})
	}
	}
	return Matrix{
	Series{
	Metric: createLabelsForAbsentFunction(e.Args[0]),
	Points: newp,
	},
	}, warnings
	}

[darshanime]

Hmm, I thought it was appropriate since both the functions are called only when the corresponding data exists.

prometheus/promql/engine.go

Line 1343 in b1ed4a0

outVec := call(inArgs, e.Args, enh)

Removed...

[juliusv]

Ah yeah, you were right, sorry. Still fine to omit that line I think, it's normal for any _over_time() function to only be called when there's data for it :)

For absent_over_time() it probably just made more sense to call it out because the function returns something even (and only if) it was not normally called.

[juliusv]

Thanks! Could you add a documentation entry for the function in the same PR, in https://github.com/prometheus/prometheus/blob/main/docs/querying/functions.md#aggregation_over_time?

[roidelapluie]

The documentation should be a bullet in the over time https://prometheus.io/docs/prometheus/latest/querying/functions/#aggregation_over_time

[roidelapluie]

the value 1 for any series in the specified interval.

I think it would be more clear, WDYT?

[juliusv]

👍 Thank you :)