pathpol: Do not use policy.Policy directly

[lukedirtwalker]

Instead use interfaces. This way tests no longer need to instantiate actual policies.
Also we can much easier refactor the Policy object without breaking anything else.

Furthermore remove the Act method instead offer Filter and FilterOpt

Fixes #3087

---

This change is 

[scrye]

Reviewed 35 of 35 files at r1.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @lukedirtwalker)

---

go/lib/pathdb/helpers.go, line 33 at r1 (raw file):

// NoPolicy should be used instead of nil when inserting policy hashes into
// fields which can't be null in the DB.
var NoPolicy = PolicyHash{0x44, 0x13, 0x6f, 0xa3, 0x55, 0xb3, 0x67, 0x8a, 0x11, 0x46, 0xad, 0x16,

How to generate this value should be added to its docstring.

---

go/lib/pathdb/helpers.go, line 39 at r1 (raw file):

// HashPolicy is a default implementation of a policy hash function. It creates
// a sha256 hash of the json serialized policy.
func HashPolicy(policy *pathpol.Policy) (PolicyHash, error) {

I realized something when looking at Dominik's JSON PRs for the TRC.

The same policy can lead to different hashes, because JSON can represent the same data in multiple ways. Don't know if this would cause problems though.

[lukedirtwalker]

Reviewable status: 34 of 35 files reviewed, 2 unresolved discussions (waiting on @scrye)

---

go/lib/pathdb/helpers.go, line 33 at r1 (raw file):

Previously, scrye (Sergiu Costea) wrote…

How to generate this value should be added to its docstring.

Done.

---

go/lib/pathdb/helpers.go, line 39 at r1 (raw file):

Previously, scrye (Sergiu Costea) wrote…

I realized something when looking at Dominik's JSON PRs for the TRC.

The same policy can lead to different hashes, because JSON can represent the same data in multiple ways. Don't know if this would cause problems though.

Clarified in the comment.

[scrye]

Reviewed 1 of 1 files at r2.
Reviewable status: complete! all files reviewed, all discussions resolved