Fix incorrect model schema parsing for enum type columns

[alexanderConstantinescu]

Enum type columns, such as: LoadBalancer.Protocol was currently parsed
to:

// LoadBalancer defines an object in Load_Balancer table
type LoadBalancer struct {
	Protocol        []LoadBalancerProtocol        `ovsdb:"protocol"`
}

Which is incorrect since the protocol is an enumerated field, not a set
of enumerated fields. This happened because the table schema for
load_balancer specified a max / min field in OVN. The table schema
definition in OVN is a bit dubious since it defines:

protocol":{"type":{"key":{"enum":["set",["sctp","tcp","udp"]],"type":"string"},"min":0}}

but we can work-around such particularities by switching the order at
which we parse the fields.

/assign @dave-tucker

[amorenoz]

The RFC states:

  If "min" is not 1 or "max" is not 1, or both, and "value" is not
  specified, the type is a set of scalar type "key".

So I think the encoding is correct. It's not the only case btw, enums are very commonly typed as sets so that an empty value is allowed.

[alexanderConstantinescu]

The RFC states:

  If "min" is not 1 or "max" is not 1, or both, and "value" is not
  specified, the type is a set of scalar type "key".

So I think the encoding is correct. It's not the only case btw, enums are very commonly typed as sets so that an empty value is allowed.

Thinking a little bit more about this I might have jumped the gun on this PR. I am almost leaning towards saying that the OVN NB DB schema seems wrong: https://github.com/ovn-org/ovn/blob/master/ovn-nb.ovsschema#L172-L175. Thanks for the RFE definition: should probably be:

               "protocol": {
                    "type": {"key": {"type": "string",
                             "enum": ["set", ["tcp", "udp", "sctp"]]},
                             "min": 1, "max": 1}},

based on that.

[dave-tucker]

Closing as the encoding is correct, but there is no client-side validation (tracked in #160) and the Go type generated for and Enum Set doesn't reflect min/max constraints tracked in #154
@alexanderConstantinescu I'll work on a PR to get the generator to give you a more type that reflects the min = 0, max = 1 constraint.