engine: remove timestamp function so that we fallback to promql again

[MichaHoffmann]

Prometheus returns the timestamp of the last sample in the lookback period, we currently only return 0 because we never pass a timestamp to the point that is passed to the function. It is not clear how to fix this since we do not save the timestamp of the last sample in the lookback period currently. Notice that prometheus returns 30 even at timestamp 90000

                    Diff:
                    --- Expected
                    +++ Actual
                    @@ -2,46 +2,46 @@
                     0 @[0]
                    -30 @[30000]
                    -30 @[60000]
                    -30 @[90000]
                    -30 @[120000]
                    -30 @[150000]
                    -30 @[180000]
                    -30 @[210000]
                    -30 @[240000]
                    -30 @[270000]
                    -30 @[300000]
                    -30 @[330000]
                    -30 @[360000]
                    -30 @[390000]
                    -30 @[420000]
                    -30 @[450000]
                    -30 @[480000]
                    -30 @[510000]
                    -30 @[540000]
                    -30 @[570000]
                    -30 @[600000]
                    -30 @[630000]
                    +0 @[30000]
                    +0 @[60000]
                    +0 @[90000]
                    +0 @[120000]
                    +0 @[150000]
                    +0 @[180000]
                    +0 @[210000]
                    +0 @[240000]
                    +0 @[270000]
                    +0 @[300000]
                    +0 @[330000]
                    +0 @[360000]
                    +0 @[390000]
                    +0 @[420000]
                    +0 @[450000]
                    +0 @[480000]
                    +0 @[510000]
                    +0 @[540000]
                    +0 @[570000]
                    +0 @[600000]
                    +0 @[630000]
                     {pod="nginx-2"} =>
                     0 @[0]
                    -30 @[30000]
                    -30 @[60000]
                    -30 @[90000]
                    -30 @[120000]
                    -30 @[150000]
                    -30 @[180000]
                    -30 @[210000]
                    -30 @[240000]
                    -30 @[270000]
                    -30 @[300000]
                    -30 @[330000]
                    -30 @[360000]
                    -30 @[390000]
                    -30 @[420000]
                    -30 @[450000]
                    -30 @[480000]
                    -30 @[510000]
                    -30 @[540000]
                    -30 @[570000]
                    -30 @[600000]
                    -30 @[630000])
                    +0 @[30000]
                    +0 @[60000]
                    +0 @[90000]
                    +0 @[120000]
                    +0 @[150000]...(output trimmed)

[fpetkovski]

Thanks, I agree we should roll this back.

[yeya24]

Let's create a separate issue to support timestamp function?

[MichaHoffmann]

Let's create a separate issue to support timestamp function?

That might be non trivial and require changes to StepVector to save timestamps of samples i fear. We could open an issue for it but i wonder if the function warrants the changes.

[yeya24]

Yes, but in the long run we have to support it natively in the engine, right? Then let's have an issue for sure and make required changes if needed.
I guess we don't want to fallback to the old engine forever.

[fpetkovski]

I think we could to push down the timestamp func into the selector and store the timestamp in the value for each series. Having an issue to track this makes sense since it’s a commonly used function.

[MichaHoffmann]

We could push the timestamp function down into the vector selector maybe, thats where we have the information easily available and timestamp only makes sense in timestamp(X) form with X being a vector selector, right? That way we dont need to modify StepVector directly.

[fpetkovski]

I've opened #206 to track this.

Are you okay with merging this @yeya24?

[yeya24]

Thanks. Let's merge it.