[Question] incomplete method list of SyncedEnforcer

[JalinWang]

There are some functions lacking in SyncedEnforcer, i.e., SetAdapter(). I don't think we can always keep them consistent, so can we add a lock-wrapped interface for compatibility? A proof-of-concept:

func (e *SyncedEnforcer) RunSynced(f func()) {
	e.m.Lock()
	defer e.m.Unlock()
	f()
}

//usage

var res [][]string
e.RunSynced(func() {
	e.Enforcer.SetAdapter(a)
	res = e.Enforcer.GetPolicy()
})

More comparision on supported methods:

[casbin-bot]

@tangyang9464 @JalinWang

[hsluoyz]

@JalinWang your method looks interesting. It can be added. But we cannot remove the old way for back compatibility either.

[tangyang9464]

@JalinWang I don't see any difference from the previous approach, doesn't this approach also need to deal with each method?

[JalinWang]

@tangyang9464 e.m is a private member which can't be accessed outside. However, we don't always keep SyncedEnforcer updated with Enforcer, e.g., SetAdapter() is missed in the synced one. I want to provide such a function (or just exposing the internal lock is okay) for users to use the missed method and leave the existing supported methods unchanged.

[tangyang9464]

@JalinWang I think we should always maintain the consistency of the two SyncedEnforcer and Enforcer APIs, instead of providing RunSync, it is not easy to use, and users are prone to misuse

[hsluoyz]

I agree with @tangyang9464 , so closed here for now, we can do offline discussions further if needed