I wrote up a quick and dirty example displaying Randomizable in action working alongside Strideable Randomizable Potential Ā· GitHub . This example uses the 32 bit Unix timestamp to determine a Dateās range (0 ā¦ UInt32.max). Colorās range would obviously be 0 ā¦ 0xFFFFFF. As you can see from this example, Randomizable has the potential to provide massive functionality to types that are able to conform to it. Rust also implements a trait called Rand that achieves the same here https://doc.rust-lang.org/rand/rand/trait.Rand.html .
Ā·Ā·Ā·
On Nov 15, 2017, 11:43 AM -0600, Nate Cook <natecook@apple.com>, wrote:
On Nov 12, 2017, at 7:47 PM, Alejandro Alonso <aalonso128@outlook.com<mailto:aalonso128@outlook.com>> wrote:
Sorry Iāve been gone for a while, I had to do a lot of traveling.
1. Initially I made this thinking that developers had the power to determine their own lower bound. The current implementation uses the integerās min value as a lower bound. If it makes sense to only allow unsigned integers from an RNG, then Iām perfectly fine with. I do disagree when you say that it should only generate UInt32s. The current approach allows, lets say mt19337 and mt19337-64, to be used within one generator. So if you wanted a UInt32, mt19337 would be used, and if you asked for a UInt64, mt19337-64 would be used.
2. The Randomizable protocol isnāt always used with integers. Think Date.random or Color.random. These types of values are difficult to express with ranges. Randomizable solves this issue.
I donāt think these examples explain how the Randomizable protocol is useful. As currently proposed, types that are Randomizable can provide a random value, but the details of that are left up to the type, and vary widely. When you use .random on an integer type, you get any value between .min and .max, but on a floating-point type, itās a value between 0.0 and 1.0. What would the range be when you use Date.random? What about Color.random?
Hereās an implementation using the Randomizable protocol as a constraint:
Ā Ā Ā Ā extension Array where Element: Randomizable {
Ā Ā Ā Ā Ā Ā Ā Ā init(randomElements: Int) {
Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā self = (0..<randomElements).map({ _ in Element.random })
Ā Ā Ā Ā Ā Ā Ā Ā }
Ā Ā Ā Ā }
Is this useful? Note that thereās no way to constrain these random values to be in a particular range or any other criteriaātheyāre only going to be random values in whatever the range is that a specific types supplies.
The point Iām trying to make is that although all these types can have random values generated, their similarity is mostly coincidental. As long as we provide good ways of generating random values at the appropriate level, we donāt need the Randomizable protocol. (To me, the appropriate extension points are FixedWidthInteger, BinaryFloatingPoint, and Bool.) Without Randomizable, we can still write things like this, which are the kinds of functionality we want to add anyway:
Ā Ā Ā Ā extension Array where Element: FixedWidthInteger {
Ā Ā Ā Ā Ā Ā Ā Ā init(randomElements: Int, in bounds: Range<Element>) {
Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā Ā self = (0..<randomElements).map({ _ in bounds.random })
Ā Ā Ā Ā Ā Ā Ā Ā }
Ā Ā Ā Ā }
-Nate
3. Iāve made the adjustment necessary for this.
4. So while I can see your point for this, it would break the consistency with Randomizableās random property. You could argue that we could make this property a function itself, but I think most will agree that Int.random is a cleaner api than Int.random().
5. Iāve made the adjustment necessary for this.
6. I actually forgot to implement the random api for the ranges where Bound: BinaryFloatingPoint. While implementing this, I realized these would never fail and would always return a non-optional. So, I decided making the other Countable ranges non-optional. (0 ..< 10).random would return a non-optional, (0.0 ..< 10.0).random would return a non-optional, and Array(0 ..< 10).random would return an optional. I can agree that something like (0 ..< 10).random is hard to discover, so I added Int.random(in: 0 ..< 10) (along with BinaryFloatingPoint). However, these are not requirements of Randomizable. I think these methods would benefit more if they were extension methods:
extension Randomizable where Self: FixedWidthInteger, Self.Stride: SignedInteger {
public static func random(
in range: Countable{Closed}Range,
using generator: RandomNumberGenerator
) -> Self {
return range.random(using: generator)
}
}
extension Randomizable where Self: BinaryFloatingPointer {
public static func random(
in range: {Closed}Range,
using generator: RandomNumberGenerator
) -> Self {
return range.random
}
}
I think external types that wish to do something similar, like Data.random(bytes: 128), could extend Randomizable with their own custom needs. The stdlib would at this point provide all the features needed to make this happen very simply for something like Data.random(bytes: 128).
- Alejandro
On Nov 5, 2017, 10:44 PM -0600, Nate Cook <natecook@apple.com<mailto:natecook@apple.com>>, wrote:
Thanks for continuing to push this forward, Alejandro! Iām excited about the potential of having access to these APIs as part of the standard library. Here are a few comments on some different parts of the proposal:
1) For your RandomGenerator protocol, Iām not totally clear on the semantics of the next(_:) and next(_:upperBound:) methods. Do they both have zero as their lower bound, for example? Iām not sure it makes sense to have signed integers generated directly by an RNGāperhaps T: FixedWidthInteger & UnsignedInteger would be a more useful constraint. (Does it even need to be generic? What if RNGs just generate UInt32s?)
2) Can you say more about the purpose of the Randomizable protocol? How would we use that protocol in useful ways that we wouldnāt get from being able to select random values from ranges (half-open and closed) of FixedWidthInteger / BinaryFloatingPoint? My experience has been that a full-width random value is rarely what a user needs.
3) I agree with Xiaodi that Random should probably be a struct with a single shared instance, but I donāt think it should be internal. Hiding that shared RNG would make it hard for non-stdlib additions to have the same usage, as they would need to have completely separate implementations for the ādefaultā and custom RNG versions.
4) I would also still suggest that the simplest version of random (that you use to get a value from a range or an element from a collection) should be a function, not a property. Collection properties like first, last, and count all represent facts that already exist about a collection, and donāt change unless the collection itself changes. Choosing a random element, on the other hand, is clearly going to be freshly performed on each call. In addition, with the notable exception of count, we try to ensure O(1) performance for properties, while random will be O(n) except in random-access collections. Finally, if it is a method, we can unify the two versions by providing a single method with the shared RNG as the default parameter.
5) To match the sorted() method, shuffled() should be on Sequence instead of Collection. I donāt think either shuffled() or shuffle() needs to be a protocol requirement, since there isnāt really any kind of customization necessary for different kinds of collections. Like the sorting algorithms, both could be regular extension methods.
6) I donāt know whether or not a consensus has formed around the correct spelling of the APIs for generating random values. From the proposal it looks like the preferred ways of getting a random value in a range would be to use the random property (or method) on a range or closed range:
Ā Ā Ā Ā (0..<10).random // 7
Ā Ā Ā Ā (0.0 ... 5.0).random // 4.112312
If thatās the goal, and we donāt want those values to be optional, weāll need an implementation of random for floating-point ranges and an overload for fixed-width integer ranges. That said, I donāt think that style is as discoverable as having static methods or initializers available on the different types:
Ā Ā Ā Ā Int.random(in: 0..<10)
Ā Ā Ā Ā Double.random(in: 0.0 ... 5.0)
Ā Ā Ā Ā // or maybe
Ā Ā Ā Ā Int(randomIn: 0..<10)
Ā Ā Ā Ā Double(randomIn: 0.0 ... 5.0)
(My only quibble with the initializer approach is that Bool would be awkward.)
In addition, this alternative approach could make creating random values more consistent with types that donāt work well in ranges:
Ā Ā Ā Ā Data.random(bytes: 128)
Ā Ā Ā Ā Color.random(r: 0...0, g: 0...1, b: 0...1, a: 1...1)
āāāā
Thanks again!
Nate
On Nov 5, 2017, at 6:33 PM, Alejandro Alonso via swift-evolution <swift-evolution@swift.org<mailto:swift-evolution@swift.org>> wrote:
[Proposal] Random Unification by Azoy Ā· Pull Request #760 Ā· apple/swift-evolution Ā· GitHub is the current API and proposed solution.
- Alejandro
On Nov 5, 2017, 6:18 PM -0600, Xiaodi Wu <xiaodi.wu@gmail.com<mailto:xiaodi.wu@gmail.com>>, wrote:
My comments are directed to the "more up-to-date" document that you just linked to in your reply to Jon. Is that one outdated? If so, can you send a link to the updated proposal and implementation for which you're soliciting feedback?
On Sun, Nov 5, 2017 at 6:12 PM, Alejandro Alonso <aalonso128@outlook.com<mailto:aalonso128@outlook.com>> wrote:
The proposal and implementation have the current updated API. The link I sent Jon was the one I brought up a few weeks ago which is outdated now. The proposal answers all of your questions. As for `.random` being a function, some would argue that it behaves in the same way as `.first` and `.last` which are properties.
- Alejandro
On Nov 5, 2017, 6:07 PM -0600, Xiaodi Wu <xiaodi.wu@gmail.com<mailto:xiaodi.wu@gmail.com>>, wrote:
A few quick thoughts:
I know that there's been some discussion that `(1...10).random` is the best spelling, but I'd like to push back on that suggestion. When I want a random number, I tend to think of the type I want first ("I want a random integer") and then a range ("I want a random integer between a and b"), not the other way around. My intuition is that `Int.random(in:)` will be more discoverable, both on that basis and because it is more similar to other languages' syntax (`Math.random` in JavaScript and `randint` in NumPy,
for example). It also has the advantage that the type is explicit, which I think is particularly useful in this case because the value itself is, well, random.
I would also argue that, `random` is most appropriately a method and not a property; there's no hard and fast rule for this, but the fact that the result is stochastic suggests (to me) that it's not a "property" of the range (or, for that matter, of the type).
I would reiterate here my qualms about `Source` being the term used for a generator. These types are not a _source_ of entropy but rather a _consumer_ of entropy.
`UnsafeRandomSource` needs to be renamed; "unsafe" has a specific meaning in Swift--that is, memory safety, and this is not it. Moreover, it's questionable whether this protocol is useful in any sense. What useful generic algorithms can one write with such a protocol?
`XoroshiroRandom` cannot be seeded by any `Numeric` value; depending on the specific algorithm it needs a seed of a specific bit width. If you default the shared instance to being seeded with an `Int` then you will have to have distinct implementations for 32-bit and 64-bit platforms. This is unadvisable. On that note, your `UnsafeRandomSource` needs to have an associated type and not a generic `<T : Numeric>` for the seed.
The default random number generator should be cryptographically secure; however, it's not clear to me that it should be device random.
I agree with others that alternative random number generators other than the default RNG (and, if not default, possibly also the device RNG) should be accommodated by the protocol hierarchy but not necessarily supplied in the stdlib.
The term `Randomizable` means something specific which is not how it's used in your proposed protocol.
There's still the open question, not answered, about how requesting an instance of the hardware RNG behaves when there's insufficient or no entropy. Does it return nil, throw, trap, or wait? The proposed API does not clarify this point, although based on the method signature it cannot return nil or throw. Trapping might be acceptable but I'd be interested to hear your take as to why it is preferable.
On Sun, Nov 5, 2017 at 4:43 PM, Alejandro Alonso via swift-evolution <swift-evolution@swift.org<mailto:swift-evolution@swift.org>> wrote:
For the proof of concept, I had accidentally deleted that one. I have a more up to date one which was discussed a few weeks later. Swift Random Unification Design Ā· GitHub
- Alejandro
On Nov 5, 2017, 4:37 PM -0600, Jonathan Hull <jhull@gbis.com<mailto:jhull@gbis.com>>, wrote:
Is there a link to the writeup? The one in the quote 404s.
Thanks,
Jon
On Nov 5, 2017, at 2:10 PM, Alejandro Alonso via swift-evolution <swift-evolution@swift.org<mailto:swift-evolution@swift.org>> wrote:
Hello once again Swift evolution community. I have taken the time to write up the proposal for this thread, and have provided an implementation for it as well. I hope to once again get good feedback on the overall proposal.
- Alejandro
On Sep 8, 2017, 11:52 AM -0500, Alejandro Alonso via swift-evolution <swift-evolution@swift.org<mailto:swift-evolution@swift.org>>, wrote:
Hello swift evolution, I would like to propose a unified approach to `random()` in Swift. I have a simple implementation here https://gist.github.com/Azoy/5d294148c8b97d20b96ee64f434bb4f5\. This implementation is a simple wrapper over existing random functions so existing code bases will not be affected. Also, this approach introduces a new random feature for Linux users that give them access to upper bounds, as well as a lower bound for both Glibc and Darwin users. This change would be implemented within Foundation.
I believe this simple change could have a very positive impact on new developers learning Swift and experienced developers being able to write single random declarations.
Iād like to hear about your ideas on this proposal, or any implementation changes if need be.
- Alejando
_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org<mailto:swift-evolution@swift.org>
https://lists.swift.org/mailman/listinfo/swift-evolution
_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org<mailto:swift-evolution@swift.org>
https://lists.swift.org/mailman/listinfo/swift-evolution
_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org<mailto:swift-evolution@swift.org>
https://lists.swift.org/mailman/listinfo/swift-evolution
or otherwise)
