Recently, I read the source code of google/btree. But I'am conflused by the struct copyOnWriteContext. It is used in node function mutableFor, like following
func (n *node) mutableFor(cow *copyOnWriteContext) *node {
if n.cow == cow {
return n
}
fmt.Println("new node?")
out := cow.newNode()
if cap(out.items) >= len(n.items) {
out.items = out.items[:len(n.items)]
} else {
out.items = make(items, len(n.items), cap(n.items))
}
copy(out.items, n.items)
// Copy children
if cap(out.children) >= len(n.children) {
out.children = out.children[:len(n.children)]
} else {
out.children = make(children, len(n.children), cap(n.children))
}
copy(out.children, n.children)
return out
}
I review all the code in this module, and found that there is only one place create the copyOnWriteContext's instance. It is when the tree is created.
func New(degree int) *BTree {
return NewWithFreeList(degree, NewFreeList(DefaultFreeListSize))
}
So what is the meaning for mutableFor. Because there is only one copyOnWriteContext in the entire code. The n.cow always equals param cow.
n.cow can be nil.
// freeNode frees a node within a given COW context, if it's owned by that
// context. It returns what happened to the node (see freeType const
// documentation).
func (c *copyOnWriteContext) freeNode(n *node) freeType {
if n.cow == c {
// clear to allow GC
n.items.truncate(0)
n.children.truncate(0)
n.cow = nil
if c.freelist.freeNode(n) {
return ftStored
} else {
return ftFreelistFull
}
} else {
return ftNotOwned
}
}
I'm just trying different things to learn go and understand the structures of how it works. Currently playing around with slices and custom types.
I have the following code which works fine and as expected.
package imgslice
import (
"fmt"
"image"
)
type imageData struct {
position int // Image Number
image *image.RGBA // Image Store
height int // Height In Pixels
width int // Width In Pixels
}
func init() {
fmt.Println("Starting")
lbl := &[]imageData{}
println(lbl)
InitImage(lbl, 3)
fmt.Printf("%#v\n", lbl)
}
// Initalise the imageData arrray
func InitImage(l *[]imageData, images int) {
var i int
var newRecord imageData
for i = 0; i < images; i++ {
newRecord = imageData{position: i}
*l = append(*l, newRecord)
}
return
}
I'm trying to re-write the InitImage function to work as a method (i think that is the correct term). But I get the error:
invalid receiver type *[]imageData ([]imageData is not a defined type)
(edit: Error is on the func (l *[]imageData) InitImageNew(images int) { line)
The only reason I want to do this is a) learning to see if it can be done and b) stylistically I think I prefer this do having the slice as an extra argument.
func (l *[]imageData) InitImageNew(images int) {
var i int
var newRecord imageData
for i = 0; i < images; i++ {
newRecord = imageData{position: i}
*l = append(*l, newRecord)
}
return
}
Looking at this answer: Function declaration syntax: things in parenthesis before function name
It seems like it should be possible - however this answer seems to say it's not possible: Golang monkey patching.
Is it possible to rewrite this so I could do
lbl := &[]imageData{}
lbl.InitImageNew(4)
You can only defined methods on named types (or pointers to named types). []Type is a compound type. You could make it a named type by defining it:
type TypeSlice []Type
And then define methods on it.
This is covered in the spec section on types.
I'm hoping someone may be able to help i'm using Xcode 8 and swift 3
I have a playground file Xcode 7 swift 2 that involves a Midi callback for Midi Input everything works fine in 7
I tried a conversion to 8 and it brought up errors regarding memory and a few name changes mostly of what i believe to be non serious i also redefined the infinite loop using PlaygroundSupport
However the error i cannot get over involves MyMIDIReadProc at
MIDIInputPortCreate(midiClient, "MidiTest_InPort", MyMIDIReadProc, nil, &inPort);
The error says
Cannot convert value of type '(pktList: UnsafePointer, readProcRefCon: UnsafeMutablePointer, srcConnRefCon: UnsafeMutablePointer) -> Void' to expected argument type 'MIDIReadProc' (aka '#convention(c) (UnsafePointer, Optional>, Optional>) -> ()')
My understanding is that it needs a #convention(c) wrapper of some description inserted. I think i'm on the right track because you can wrap a function but my knowledge of where to put it has run out. Again i was hoping some one might be able to advise
Thanks for reading
apologies for any bad language as i'm self taught
Here is the original Xcode 7 code
import Cocoa
import CoreMIDI
import XCPlayground
func getDisplayName(obj: MIDIObjectRef) -> String
{
var param: Unmanaged<CFString>?
var name: String = "Error";
let err: OSStatus = MIDIObjectGetStringProperty(obj, kMIDIPropertyDisplayName, ¶m)
if err == OSStatus(noErr)
{
name = param!.takeRetainedValue() as String
}
return name;
}
func MyMIDIReadProc(pktList: UnsafePointer<MIDIPacketList>,
readProcRefCon: UnsafeMutablePointer<Void>, srcConnRefCon: UnsafeMutablePointer<Void>) -> Void
{
let packetList:MIDIPacketList = pktList.memory;
let srcRef:MIDIEndpointRef = UnsafeMutablePointer<MIDIEndpointRef>(COpaquePointer(srcConnRefCon)).memory;
print("MIDI Received From Source: \(getDisplayName(srcRef))");
var packet:MIDIPacket = packetList.packet;
for _ in 1...packetList.numPackets
{
let bytes = Mirror(reflecting: packet.data).children;
var dumpStr = "";
// bytes mirror contains all the zero values in the ridiulous packet data tuple
// so use the packet length to iterate.
var i = packet.length;
for (_, attr) in bytes.enumerate()
{
dumpStr += String(format:"$%02X ", attr.value as! UInt8);
--i;
if (i <= 0)
{
break;
}
}
print(dumpStr)
packet = MIDIPacketNext(&packet).memory;
}
}
var midiClient: MIDIClientRef = 0;
var inPort:MIDIPortRef = 0;
var src:MIDIEndpointRef = MIDIGetSource(0);
MIDIClientCreate("MidiTestClient", nil, nil, &midiClient);
MIDIInputPortCreate(midiClient, "MidiTest_InPort", MyMIDIReadProc, nil, &inPort);
MIDIPortConnectSource(inPort, src, &src);
// Keep playground running
XCPlaygroundPage.currentPage.needsIndefiniteExecution = true;
And here is the Xcode 8 code converted
var str = "Hello, playground"
import Cocoa
import CoreMIDI
import XCPlayground
import PlaygroundSupport
func getDisplayName(obj: MIDIObjectRef) -> String
{
var param: Unmanaged<CFString>?
var name: String = "Error";
let err: OSStatus = MIDIObjectGetStringProperty(obj, kMIDIPropertyDisplayName, ¶m)
if err == OSStatus(noErr)
{
name = param!.takeRetainedValue() as String
}
return name;
}
func MyMIDIReadProc(pktList: UnsafePointer<MIDIPacketList>,
readProcRefCon: UnsafeMutablePointer<Void>, srcConnRefCon: UnsafeMutablePointer<Void>) -> Void
{
let packetList:MIDIPacketList = pktList.pointee;
let srcRef:MIDIEndpointRef = UnsafeMutablePointer<MIDIEndpointRef>(OpaquePointer(srcConnRefCon)).pointee;
print("MIDI Received From Source: \(getDisplayName(obj: srcRef))");
var packet:MIDIPacket = packetList.packet;
for _ in 1...packetList.numPackets
{
let bytes = Mirror(reflecting: packet.data).children;
var dumpStr = "";
var i = packet.length;
for (_, attr) in bytes.enumerated()
{
dumpStr += String(format:"$%02X ", attr.value as! UInt8);
i -= 1;
if (i <= 0)
{
break;
}
}
print(dumpStr)
packet = MIDIPacketNext(&packet).pointee;
}
}
var midiClient: MIDIClientRef = 0;
var inPort:MIDIPortRef = 0;
var src:MIDIEndpointRef = MIDIGetSource(0);
MIDIClientCreate("MidiTestClient", nil, nil, &midiClient);
MIDIInputPortCreate(midiClient, "MidiTest_InPort", MyMIDIReadProc, nil, &inPort);
MIDIPortConnectSource(inPort, src, &src);
PlaygroundPage.current.needsIndefiniteExecution = true
Pointer types are drastically changed in Swift 3. Many C-based APIs' signatures are changed accordingly.
Following those changes manually would be painful. You can make Swift work for you, with a little modification.
Try changing the function header:
func MyMIDIReadProc(pktList: UnsafePointer<MIDIPacketList>,
readProcRefCon: UnsafeMutablePointer<Void>, srcConnRefCon: UnsafeMutablePointer<Void>) -> Void
{
to a closure declaration:
let MyMIDIReadProc: MIDIReadProc = {pktList, readProcRefCon, srcConnRefCon in
Swift infers argument types perfectly in this style.
You may need to fix pointer type conversion:
let srcRef:MIDIEndpointRef = UnsafeMutablePointer<MIDIEndpointRef>(OpaquePointer(srcConnRefCon)).pointee;
to something like this:
//I'm not sure using `!` is safe here...
let srcRef: MIDIEndpointRef = UnsafeMutablePointer(srcConnRefCon!).pointee
(By the way, the equivalent part in your Xcode 7 code is a little bit redundant. You have no need to use intermediate COpaquePointer there.)
In Swift 3, pointers cannot be nil, and nullable pointers are represented with Optionals. You may need many other fixes to work with C-based APIs in Swift 3.
OOPer is pointing (ahem) you in the right direction. Here is a blog post on using Swift 3 Core MIDI along with a working github repo.
Assuming that you're working with CoreMIDI 1.3 or later, you may have more luck using MIDIInputPortCreateWithBlock instead of MIDIInputPortCreate.
This method takes a Swift block as a parameter instead of requiring an #convention(c) function reference, making it more amenable to use within methods belonging to Swift classes, e.g.:
public func midiReadBlock(ptr: UnsafePointer<MIDIPacketList>, _: UnsafeMutableRawPointer?) -> Void {
let list: MIDIPacketList = ptr.pointee
...
}
You may also find these two extensions useful.
This one (derived from here) allows you to iterate directly over a MIDIPacketList using for pkt in list:
extension MIDIPacketList: Sequence {
public func makeIterator() -> AnyIterator<MIDIPacket> {
var iterator: MIDIPacket?
var nextIndex: UInt32 = 0
return AnyIterator {
nextIndex += 1
if nextIndex > self.numPackets { return nil }
if iterator != nil {
iterator = withUnsafePointer(to: &iterator!) { MIDIPacketNext($0).pointee }
} else {
iterator = self.packet;
}
return iterator
}
}
}
and this one adds a method to a MIDIPacket to extract the contents as a [UInt8] instead of having to use the really broken tuple syntax:
extension MIDIPacket {
public var asArray: [UInt8] {
let mirror = Mirror(reflecting: self.data)
let length = Int(self.length)
var result = [UInt8]()
result.reserveCapacity(length)
for (n, child) in mirror.children.enumerated() {
if n == length {
break
}
result.append(child.value as! UInt8)
}
return result
}
}
I'd like to know how to retrieve the parent struct of an instance.
I have no idea how to implement this.
For instance:
type Hood struct {
name string
houses []House
}
type House struct {
name string
people int16
}
func (h *Hood) addHouse(house House) []House {
h.houses = append(h.houses, house)
return h.houses
}
func (house *House) GetHood() Hood {
//Get hood where the house is situated
return ...?
}
Cheers
You should retain a pointer to the hood.
type House struct {
hood *Hood
name string
people int16
}
and when you append the house
func (h *Hood) addHouse(house House) []House {
house.hood = h
h.houses = append(h.houses, house)
return h.houses
}
then you can easily change the GetHood, although a getter may not be required at that point.
func (house *House) GetHood() Hood {
return *house.hood
}
How should bit fields be declared and used in Swift?
Declaring an enum like this does work, but trying to OR 2 values together fails to compile:
enum MyEnum: Int
{
case One = 0x01
case Two = 0x02
case Four = 0x04
case Eight = 0x08
}
// This works as expected
let m1: MyEnum = .One
// Compiler error: "Could not find an overload for '|' that accepts the supplied arguments"
let combined: MyEnum = MyEnum.One | MyEnum.Four
I looked at how Swift imports Foundation enum types, and it does so by defining a struct that conforms to the RawOptionSet protocol:
struct NSCalendarUnit : RawOptionSet {
init(_ value: UInt)
var value: UInt
static var CalendarUnitEra: NSCalendarUnit { get }
static var CalendarUnitYear: NSCalendarUnit { get }
// ...
}
And the RawOptionSet protocol is:
protocol RawOptionSet : LogicValue, Equatable {
class func fromMask(raw: Self.RawType) -> Self
}
However, there is no documentation on this protocol and I can't figure out how to implement it myself. Moreover, it's not clear if this is the official Swift way of implementing bit fields or if this is only how the Objective-C bridge represents them.
You can build a struct that conforms to the RawOptionSet protocol, and you'll be able to use it like the built-in enum type but with bitmask functionality as well. The answer here shows how:
Swift NS_OPTIONS-style bitmask enumerations.
Updated for Swift 2/3
Since swift 2, a new solution has been added as "raw option set" (see: Documentation), which is essentially the same as my original response, but using structs that allow arbitrary values.
This is the original question rewritten as an OptionSet:
struct MyOptions: OptionSet
{
let rawValue: UInt8
static let One = MyOptions(rawValue: 0x01)
static let Two = MyOptions(rawValue: 0x02)
static let Four = MyOptions(rawValue: 0x04)
static let Eight = MyOptions(rawValue: 0x08)
}
let m1 : MyOptions = .One
let combined : MyOptions = [MyOptions.One, MyOptions.Four]
Combining with new values can be done exactly as Set operations (thus the OptionSet part), .union, likewise:
m1.union(.Four).rawValue // Produces 5
Same as doing One | Four in its C-equivalent. As for One & Mask != 0, can be specified as a non-empty intersection
// Equivalent of A & B != 0
if !m1.intersection(combined).isEmpty
{
// m1 belongs is in combined
}
Weirdly enough, most of the C-style bitwise enums have been converted to their OptionSet equivalent on Swift 3, but Calendar.Compontents does away with a Set<Enum>:
let compontentKeys : Set<Calendar.Component> = [.day, .month, .year]
Whereas the original NSCalendarUnit was a bitwise enum. So both approaches are usable (thus the original response remains valid)
Original Response
I think the best thing to do, is to simply avoid the bitmask syntax until the Swift devs figure out a better way.
Most of the times, the problem can be solved using an enum and and a Set
enum Options
{
case A, B, C, D
}
var options = Set<Options>(arrayLiteral: .A, .D)
An and check (options & .A) could be defined as:
options.contains(.A)
Or for multiple "flags" could be:
options.isSupersetOf(Set<Options>(arrayLiteral: .A, .D))
Adding new flags (options |= .C):
options.insert(.C)
This also allows for using all the new stuff with enums: custom types, pattern matching with switch case, etc.
Of course, it doesn't have the efficiency of bitwise operations, nor it would be compatible with low level things (like sending bluetooth commands), but it's useful for UI elements that the overhead of the UI outweighs the cost of the Set operations.
They showed how to do this in one of the WWDC videos.
let combined = MyEnum.One.toRaw() | MyEnum.Four.toRaw()
Note that combined will be Int type and will actually get a compiler error if you specify let combined: MyEnum. That is because there is no enum value for 0x05 which is the result of the expression.
I think maybe some of the answers here are outdated with overcomplicated solutions? This works fine for me..
enum MyEnum: Int {
case One = 0
case Two = 1
case Three = 2
case Four = 4
case Five = 8
case Six = 16
}
let enumCombined = MyEnum.Five.rawValue | MyEnum.Six.rawValue
if enumCombined & MyEnum.Six.rawValue != 0 {
println("yay") // prints
}
if enumCombined & MyEnum.Five.rawValue != 0 {
println("yay again") // prints
}
if enumCombined & MyEnum.Two.rawValue != 0 {
println("shouldn't print") // doesn't print
}
If you don't need to interoperate with Objective-C and just want the syntax of bit masks in Swift, I've written a simple "library" called BitwiseOptions that can do this with regular Swift enumerations, e.g.:
enum Animal: BitwiseOptionsType {
case Chicken
case Cow
case Goat
static let allOptions = [.Chicken, .Cow, .Goat]
}
var animals = Animal.Chicken | Animal.Goat
animals ^= .Goat
if animals & .Chicken == .Chicken {
println("Chick-Fil-A!")
}
and so on. No actual bits are being flipped here. These are set operations on opaque values. You can find the gist here.
#Mattt's very famous "NSHipster" has an extensive detailed description of the RawOptionsSetType : http://nshipster.com/rawoptionsettype/
It includes a handy Xcode snipped:
struct <# Options #> : RawOptionSetType, BooleanType {
private var value: UInt = 0
init(_ value: UInt) { self.value = value }
var boolValue: Bool { return value != 0 }
static func fromMask(raw: UInt) -> <# Options #> { return self(raw) }
static func fromRaw(raw: UInt) -> <# Options #>? { return self(raw) }
func toRaw() -> UInt { return value }
static var allZeros: <# Options #> { return self(0) }
static func convertFromNilLiteral() -> <# Options #> { return self(0) }
static var None: <# Options #> { return self(0b0000) }
static var <# Option #>: <# Options #> { return self(0b0001) }
// ...
}
You have to use .toRaw() after each member:
let combined: Int = MyEnum.One.toRaw() | MyEnum.Four.toRaw()
will work. Because as it is you're just trying to assign "One" which is a MyEnum type, not an integer. As Apple's documentation says:
“Unlike C and Objective-C, Swift enumeration members are not assigned a default integer value when they are created. In the CompassPoints example, North, South, East and West do not implicitly equal 0, 1, 2 and 3. Instead, the different enumeration members are fully-fledged values in their own right, with an explicitly-defined type of CompassPoint.”
so you have to use raw values if you want the members to represent some other type, as described here:
Enumeration members can come prepopulated with default values (called raw values), which are all of the same type. The raw value for a particular enumeration member is always the same. Raw values can be strings, characters, or any of the integer or floating-point number types. Each raw value must be unique within its enumeration declaration. When integers are used for raw values, they auto-increment if no value is specified for some of the enumeration members. Access the raw value of an enumeration member with its toRaw method.
I use the following I need the both values I can get, rawValue for indexing arrays and value for flags.
enum MyEnum: Int {
case one
case two
case four
case eight
var value: UInt8 {
return UInt8(1 << self.rawValue)
}
}
let flags: UInt8 = MyEnum.one.value ^ MyEnum.eight.value
(flags & MyEnum.eight.value) > 0 // true
(flags & MyEnum.four.value) > 0 // false
(flags & MyEnum.two.value) > 0 // false
(flags & MyEnum.one.value) > 0 // true
MyEnum.eight.rawValue // 3
MyEnum.four.rawValue // 2
This worked for me.
1 << 0 //0000
1 << 1 //0010
1 << 2 //0100
1 << 3 //1000
enum Collision: Int {
case Enemy, Projectile, Debris, Ground
func bitmask() -> UInt32 {
return 1 << self.rawValue
}
}
I'm taking a guess that something like this is how they are modeling enum options in Foundation:
struct TestOptions: RawOptionSet {
// conform to RawOptionSet
static func fromMask(raw: UInt) -> TestOptions {
return TestOptions(raw)
}
// conform to LogicValue
func getLogicValue() -> Bool {
if contains([1, 2, 4], value) {
return true
}
return false
}
// conform to RawRepresentable
static func fromRaw(raw: UInt) -> TestOptions? {
if contains([1, 2, 4], raw) {
return TestOptions(raw)
}
return nil
}
func toRaw() -> UInt {
return value
}
// options and value
var value: UInt
init(_ value: UInt) {
self.value = value
}
static var OptionOne: TestOptions {
return TestOptions(1)
}
static var OptionTwo: TestOptions {
return TestOptions(2)
}
static var OptionThree: TestOptions {
return TestOptions(4)
}
}
let myOptions = TestOptions.OptionOne | TestOptions.OptionThree
println("myOptions: \(myOptions.toRaw())")
if (myOptions & TestOptions.OptionOne) {
println("OPTION ONE is in there")
} else {
println("nope, no ONE")
}
if (myOptions & TestOptions.OptionTwo) {
println("OPTION TWO is in there")
} else {
println("nope, no TWO")
}
if (myOptions & TestOptions.OptionThree) {
println("OPTION THREE is in there")
} else {
println("nope, no THREE")
}
let nextOptions = myOptions | TestOptions.OptionTwo
println("options: \(nextOptions.toRaw())")
if (nextOptions & TestOptions.OptionOne) {
println("OPTION ONE is in there")
} else {
println("nope, no ONE")
}
if (nextOptions & TestOptions.OptionTwo) {
println("OPTION TWO is in there")
} else {
println("nope, no TWO")
}
if (nextOptions & TestOptions.OptionThree) {
println("OPTION THREE is in there")
} else {
println("nope, no THREE")
}
...where myOptions and nextOptions are of type TestOptions - I'm not exactly sure how fromMask() and getLogicValue() are supposed to act here (I just took some best guesses), maybe somebody could pick this up and work it out?
If you want bitfield in Swift, then enum is the wrong way. Better just do like this
class MyBits {
static let One = 0x01
static let Two = 0x02
static let Four = 0x04
static let Eight = 0x08
}
let m1 = MyBits.One
let combined = MyBits.One | MyBits.Four
You don't really need the class/static wrapper, but I include it as a kind of pseudo namespace.
Do bitwise operation using raw value then create a new enum object using the result.
let mask = UIViewAutoresizing(rawValue: UIViewAutoresizing.FlexibleWidth.rawValue|UIViewAutoresizing.FlexibleHeight.rawValue)
self.view.autoresizingMask = mask
Here's something I put together to try to make a Swift enum that resembles to some extent a C# flags-style enum. But I'm just learning Swift, so this should only be considered to be "proof of concept" code.
/// This EnumBitFlags protocol can be applied to a Swift enum definition, providing a small amount
/// of compatibility with the flags-style enums available in C#.
///
/// The enum should be defined as based on UInt, and enum values should be defined that are powers
/// of two (1, 2, 4, 8, ...). The value zero, if defined, should only be used to indicate a lack of
/// data or an error situation.
///
/// Note that with C# the enum may contain a value that does not correspond to the defined enum
/// constants. This is not possible with Swift, it enforces that only valid values can be set.
public protocol EnumBitFlags : RawRepresentable, BitwiseOperations {
var rawValue : UInt { get } // This provided automatically by enum
static func createNew(_ rawValue : UInt) -> Self // Must be defined as some boiler-plate code
}
/// Extension methods for enums that implement the EnumBitFlags protocol.
public extension EnumBitFlags {
// Implement protocol BitwiseOperations. But note that some of these operators, especially ~,
// will almost certainly result in an invalid (nil) enum object, resulting in a crash.
public static func & (leftSide: Self, rightSide: Self) -> Self {
return self.createNew(leftSide.rawValue & rightSide.rawValue)
}
public static func | (leftSide: Self, rightSide: Self) -> Self {
return self.createNew(leftSide.rawValue | rightSide.rawValue)
}
public static func ^ (leftSide: Self, rightSide: Self) -> Self {
return self.createNew(leftSide.rawValue ^ rightSide.rawValue)
}
public static prefix func ~ (x: Self) -> Self {
return self.createNew(~x.rawValue)
}
public static var allZeros: Self {
get {
return self.createNew(0)
}
}
// Method hasFlag() for compatibility with C#
func hasFlag<T : EnumBitFlags>(_ flagToTest : T) -> Bool {
return (self.rawValue & flagToTest.rawValue) != 0
}
}
This shows how it can be used:
class TestEnumBitFlags {
// Flags-style enum specifying where to write the log messages
public enum LogDestination : UInt, EnumBitFlags {
case none = 0 // Error condition
case systemOutput = 0b01 // Logging messages written to system output file
case sdCard = 0b10 // Logging messages written to SD card (or similar storage)
case both = 0b11 // Both of the above options
// Implement EnumBitFlags protocol
public static func createNew(_ rawValue : UInt) -> LogDestination {
return LogDestination(rawValue: rawValue)!
}
}
private var _logDestination : LogDestination = .none
private var _anotherEnum : LogDestination = .none
func doTest() {
_logDestination = .systemOutput
assert(_logDestination.hasFlag(LogDestination.systemOutput))
assert(!_logDestination.hasFlag(LogDestination.sdCard))
_anotherEnum = _logDestination
assert(_logDestination == _anotherEnum)
_logDestination = .systemOutput | .sdCard
assert(_logDestination.hasFlag(LogDestination.systemOutput) &&
_logDestination.hasFlag(LogDestination.sdCard))
/* don't do this, it results in a crash
_logDestination = _logDestination & ~.systemOutput
assert(_logDestination == .sdCard)
*/
_logDestination = .sdCard
_logDestination |= .systemOutput
assert(_logDestination == .both)
}
}
Suggestions for improvement are welcome.
EDIT: I've given up on this technique myself, and therefore obviously can't recommend it anymore.
The big problem is that Swift demands that rawValue must match one of the defined enum values. This is OK if there are only 2 or 3 or maybe even 4 flag bits - just define all of the combination values in order to make Swift happy. But for 5 or more flag bits it becomes totally crazy.
I'll leave this posted in case someone finds it useful, or maybe as a warning of how NOT to do it.
My current solution to this situation is based on using a struct instead of enum, together with a protocol and some extension methods. This works much better. Maybe I'll post it someday when I'm more sure that that isn't also isn't going to backfire on me.
Task
Get all flags from flags_combination. Each flag and flags_combination are integers. flags_combination = flag_1 | flags_2
Details
Xcode 11.2.1 (11B500), Swift 5.1
Solution
import Foundation
protocol FlagPrototype: CaseIterable, RawRepresentable where RawValue == Int {}
extension FlagPrototype {
init?(rawValue: Int) {
for flag in Self.allCases where flag.rawValue == rawValue {
self = flag
return
}
return nil
}
static func all(from combination: Int) -> [Self] {
return Self.allCases.filter { return combination | $0.rawValue == combination }
}
}
Usage
enum Flag { case one, two, three }
extension Flag: FlagPrototype {
var rawValue: Int {
switch self {
case .one: return 0x1
case .two: return 0x2
case .three: return 0x4
}
}
}
var flags = Flag.two.rawValue | Flag.three.rawValue
let selectedFlags = Flag.all(from: flags)
print(selectedFlags)
if selectedFlags == [.two, .three] { print("two | three") }