README.md

SUSPENDED!

This library has been replaced by LTView for more simplicity and won't be maintained anymore until I need for it.

OT4

A persistent datastructure based idempotent outline-view. Eonil, May 2019.

This is one of trials to make AppKit less painful. Using Apple AppKit is painful even for Apple fanatics, and one of the most painful point is NSOutlineView. As like other classes in AppKit, this class is designed as a half-finished assembly, and you need to implement the rest to use it.

There're many opinions about the rest part. Apple's original answer for convenience was Cocoa Binding with NSTreeController, but IMO, Cocoa Binding is deprecated, and not the solution.

Painless Tree Interaction

OT4View is persistent datastructure based tree rendering view with fully idempotent rendering behavior.

typealias Source = OT4Source<Int,String>
typealias View = OT4View<Source,OT4ItemView<String>> 
let v = View()
var s = Source() 
s.timeline.insert("This", for: 1, at: [])
s.timeline.insert("is", for: 2, at: [0])
s.timeline.insert("a", for: 3, at: [0,0])
s.timeline.insert("demo!", for: 4, at: [1])
v.control(.render(s))

// Done!

You'll get this.

Let's consider hash-table look-up and tree navigation as O(1).

As Source is persistent, it keeps all versions of snapshots and changed key set. With that informations, view can recover diff information quickly in O(d) time where d is number of changed nodes. Tree mutators like inserting/updating/removing node takes O(1) time.

Regardless of how many operations you performed, only last 4 operations will be tracked at maximum. Tracking informations for any older operation will be lost. (only tracking informations, final snapshot will be kept) If diff information is lost and the view cannot continuously track the state, rendering takes O(n) time where n is total number of nodes in the tree.

Also rendering is fully idempotent. Passing same value to the view always provides same final rendering.

v.control(.render(s))
v.control(.render(s))
v.control(.render(s))

// Same result.

You also can track selections.

v.note = { [weak self] n 
    switch n {
    case .interaction(let ix):
    let ids = Array(ix.selectedIdentities)
    print("selected: \(ids)")
}

You gonna see something like this.

selected: [111]
selected: [222]
selected: [444]

For Snapshot to Snapshot Rendering

OT4Source based rendering requires you to keep once created value and mutate it in-place to keep internally generated diff informations. But sometimes you can't always build your data, and need only snapshots without intermediate operations. In that case, you can use OT4SnapshotView instead of for snapshot to snapshot rendering.

let v = OT4SnapshotView<OT4KeyValueTree>
var s = OT4KeyValueTree(key: 111, value: "top") 
s.subtrees.append(OT4KeyValueTree(key: 222, value: "first")
s.subtrees.append(OT4KeyValueTree(key: 333, value: "second")
s.subtrees.append(OT4KeyValueTree(key: 444, value: "third")
v.control(.render(s))

// Done!

But rendering with OT4SnapshotView always takes O(n log n) where n is total number of nodes in tree. Choose if you need something quick and dirty working stuff.

Or if you're willing to write more code, you can implement OT4SourceProtocol yourself. All data consumption is protocol based. You can wrap your data to feed it to OT4View.

Safety

OT4SnapshotView and OT4View provide full type-safety. You don't need to worry and check types becuase compiler will do the job for you. Also as it is fully type annotated, your IDE can provide nice auto-completion support.

Performance

In my benchmark, OT4View performs at 1,000 ops/s with 100,000 existing items with a few seconds of initial dealy. Single operation takes about 1ms to finish. It seems major bottleneck is in my HAMT implementation.

But It Does Not Cover Everything

Some features of NSOutlineView is context-sensitive, therefore cannot be represented as value. For example, there's clickedRow property. When you open a context-menu on a node, this property provides end-user clicked-row, but only at the moment, and disappears immediately. Therefore, I cannot support the feature. Therefore, OT4View also provides some members to access such features synchronously.

License & Credits

This code is licensed under "MIT License". Contributions will also be under "MIT License". Copyright Eonil, Hoon H.. 2019.