Implement tap selector

[leoparente]

No description provided.

[leoparente]

Changed to draft: still validating manually it as well

[leoparente]

Ready to review

[leoparente]

Tap name concatenated to handler name to ensure uniqueness

{
    "default": {
        "default-default-dns1": {
            "dns": {
                "cardinality": {
                    "qname": 0
                },
                "period": {
                    "length": 1,

And it was added as a label for prometheus return value:

# HELP dns_rates_total Rate of all DNS wire packets (combined ingress and egress) per second
# TYPE dns_rates_total summary
dns_rates_total{module="default-default-dns1",policy="default",quantile="0.5",tap="default"} 0
dns_rates_total{module="default-default-dns1",policy="default",quantile="0.9",tap="default"} 0
dns_rates_total{module="default-default-dns1",policy="default",quantile="0.95",tap="default"} 1
dns_rates_total{module="default-default-dns1",policy="default",quantile="0.99",tap="default"} 1

[leoparente]

Added error 422 when there is no match for tap selector:

[weyrick]

so i think it's almost there - but it's also an extremely hard PR to review since lots of code is moved around and it's tricky logic. i think this means we really need good unit tests because this is a critical section of code. i'd also love to find a way to simplify this policy loading code. ideally that means breaking it up into separate methods performing sections of the transaction which are no more than 20-30 lines long, say. but i don't know how realistic that is.

[leoparente]

so i think it's almost there - but it's also an extremely hard PR to review since lots of code is moved around and it's tricky logic. i think this means we really need good unit tests because this is a critical section of code. i'd also love to find a way to simplify this policy loading code. ideally that means breaking it up into separate methods performing sections of the transaction which are no more than 20-30 lines long, say. but i don't know how realistic that is.

Agreed. That's why I added new unit tests to validate it, did this based on Code Coverage to ensure to test new code.

I think this PR is a improvement on that policy load which add new two methods to break it a bit. I also agree that we can improve more that... But I think we could do that in another PR