FS-1135 - Random functions for collections

RFCs/FS-1135-random-functions-for-collections.md

@@ -0,0 +1,219 @@
+# F# RFC FS-1135 - Random functions for collections (List, Array, Seq)
+
+The design suggestion [Add shuffle, sample etc. methods for lists, arrays etc.](https://github.com/fsharp/fslang-suggestions/issues/508) has been marked "approved in principle".
+
+This RFC covers the detailed proposal for this suggestion.
+
+- [x] [Suggestion](https://github.com/fsharp/fslang-suggestions/issues/508)
+- [x] Approved in principle
+- [ ] [Implementation]() (no implementation yet)
+- [ ] Design Review Meeting(s) with @dsyme and others invitees
+
+[Discussion](https://github.com/fsharp/fslang-design/discussions/731)
+
+# Summary
+
+This feature extends the collection apis with functions for random sampling and shuffling built-in fsharp collections.
+
+# Motivation
+
+This feature is motivated by the following use cases:
+ - Using F# for data science and machine learning (like building a neural network), where data shuffling plays important role
+ - Building games, where random sampling is used for generating random levels, random decks, etc.
+ - Building simulations, where random sampling is used for generating random input data
+
+# Detailed design
+
+### General
+
+The following general rules are applied to all functions
+ - New functions should be implemented in `List`, `Array`, `Seq` modules
+ - Each function should have a variant that takes a [Random](https://learn.microsoft.com/en-us/dotnet/api/system.random) argument
+ - Each function should have a variant that takes a custom `randomizer` function. This function should return a `float` value that is greater than or equal to `0.0`, and less than `1.0`, like [Random.NextDouble](https://learn.microsoft.com/en-us/dotnet/api/system.random.nextdouble)).
+ - Custom shared thread-safe `Random` instance should be used for function without `Random` argument (since `Random.Shared` is only available since .NET 6)
+
+### Shuffle
+
+The shuffle functions return a new collection of the same collection type and of the same size, with each item in a randomly mixed position. The chance to end up in any position is weighted evenly on the length of the collection.
+
+The following functions will be added to each module.
+
+```fsharp
+// Array module
+val randomShuffle: array:'T[] -> 'T[]
+val randomShuffleWith: random:Random -> array:'T[] -> 'T[]
+val randomShuffleBy: randomizer: (unit -> float) -> array:'T[] -> 'T[]
+val randomShuffleInPlace: array:'T[] -> 'T[]
+val randomShuffleInPlaceWith: random:Random -> array:'T[] -> 'T[]
+val randomShuffleInPlaceBy: randomizer: (unit -> float) -> array:'T[] -> 'T[]
+// List module
+val randomShuffle: list:'T list -> 'T list
+val randomShuffleWith: random:Random -> list:'T list -> 'T list
+val randomShuffleBy: randomizer: (unit -> float) -> list:'T list -> 'T list
+// Seq module
+val randomShuffle: source:'T seq -> 'T seq
+val randomShuffleWith: random:Random -> source:'T seq -> 'T seq
+val randomShuffleBy: randomizer: (unit -> float) -> source:'T seq -> 'T seq
+```
+[ArgumentNullException](https://learn.microsoft.com/en-us/dotnet/api/system.argumentnullexception) is raised if collection is `null`, or if the `random` argument is `null`.
+
+[ArgumentOutOfRangeException](https://learn.microsoft.com/en-us/dotnet/api/system.argumentoutofrangeexception) is raised if the `randomizer` function returns a `float` value that is less than `0.0` or greater or equal to `1.0`.
+
+Example:
+```fsharp
+let allPlayers = [ "Alice"; "Bob"; "Charlie"; "Dave" ]
+let round1Order = allPlayers |> List.randomShuffle // [ "Charlie"; "Dave"; "Alice"; "Bob" ]
+```
+
+### Choice
+
+The choice functions return a single random element from the given collection. The random choice is weighted evenly on the size of the collection.
+
+The following functions will be added to each module.
+
+```fsharp
+// Array module
+val randomChoice: array:'T[] -> 'T
+val randomChoiceWith: random:Random -> array:'T[] -> 'T
+val randomChoiceBy: randomizer: (unit -> float) -> array:'T[] -> 'T
+// List module
+val randomChoice: list:'T list -> 'T
+val randomChoiceWith: random:Random -> list:'T list -> 'T
+val randomChoiceBy: randomizer: (unit -> float) -> list:'T list -> 'T
+// Seq module
+val randomChoice: source:'T seq -> 'T
+val randomChoiceWith: random:Random -> source:'T seq -> 'T
+val randomChoiceBy: randomizer: (unit -> float) -> source:'T seq -> 'T
+```
+[ArgumentNullException](https://learn.microsoft.com/en-us/dotnet/api/system.argumentnullexception) is raised if collection is `null`, or if the `random` argument is `null`.
+
+[ArgumentException](https://learn.microsoft.com/en-us/dotnet/api/system.argumentexception) is raised if collection is empty.
+
+[ArgumentOutOfRangeException](https://learn.microsoft.com/en-us/dotnet/api/system.argumentoutofrangeexception) is raised if the `randomizer` function returns a `float` value that is less than `0.0` or greater or equal to `1.0`.
+
+Example:
+```fsharp
+let allPlayers = [ "Alice"; "Bob"; "Charlie"; "Dave" ]
+let round1Order = allPlayers |> List.randomChoice // "Charlie"
+```
+
+### Choices
+
+Choices should select N elements from input collection in random order, once element is taken it can be selected again.
+
+The following functions will be added to each module.
+
+```fsharp
+// Array module
+val randomChoices: count:int -> array:'T[] -> 'T[]
+val randomChoicesWith: random:Random -> count:int -> array:'T[] -> 'T[]
+val randomChoicesBy: randomizer: (unit -> float) -> count:int -> array:'T[] -> 'T[]
+// List module
+val randomChoices: count:int -> list:'T list -> 'T list
+val randomChoicesWith: random:Random -> count:int -> list:'T list -> 'T list
+val randomChoicesBy: randomizer: (unit -> float) -> count:int -> list:'T list -> 'T list
+// Seq module
+val randomChoices: count:int -> source:'T seq -> 'T seq
+val randomChoicesWith: random:Random -> count:int -> source:'T seq -> 'T seq
+val randomChoicesBy: randomizer: (unit -> float) -> count:int -> source:'T seq -> 'T seq
+```
+[ArgumentNullException](https://learn.microsoft.com/en-us/dotnet/api/system.argumentnullexception) is raised if collection is `null`, or if the `random` argument is `null`.
+
+[ArgumentException](https://learn.microsoft.com/en-us/dotnet/api/system.argumentoutofrangeexception) is raised if N is negative.
+
+[ArgumentException](https://learn.microsoft.com/en-us/dotnet/api/system.argumentexception) is raised if collection is empty.
+
+[ArgumentOutOfRangeException](https://learn.microsoft.com/en-us/dotnet/api/system.argumentoutofrangeexception) is raised if the `randomizer` function returns a `float` value that is less than `0.0` or greater or equal to `1.0`.
+
+Example:
+```fsharp
+let allPlayers = [ "Alice"; "Bob"; "Charlie"; "Dave" ]
+let round1Order = allPlayers |> List.randomChoices 3 // ["Bob", "Dave", "Bob"]
+```
+
+### Sample
+
+Sample should select N elements from input collection in random order, once element is taken it won't be selected again. N can't be greater than collection length
+
+The following functions will be added to each module.
+
+```fsharp
+// Array module
+val randomSample: count:int -> array:'T[] -> 'T[]
+val randomSampleWith: random:Random -> count:int -> array:'T[] -> 'T[]
+val randomSampleBy: randomizer: (unit -> float) -> count:int -> array:'T[] -> 'T[]
+// List module
+val randomSample: count:int -> list:'T list -> 'T list
+val randomSampleWith: random:Random -> count:int -> list:'T list -> 'T list
+// Seq module
+val randomSample: count:int -> source:'T seq -> 'T seq
+val randomSampleWith: random:Random -> count:int -> source:'T seq -> 'T seq
+val randomSampleBy: randomizer: (unit -> float) -> count:int -> source:'T seq -> 'T seq
+```
+[ArgumentNullException](https://learn.microsoft.com/en-us/dotnet/api/system.argumentnullexception) is raised if collection is `null`, or if the `random` argument is `null`.
+
+[ArgumentException](https://learn.microsoft.com/en-us/dotnet/api/system.argumentoutofrangeexception) is raised if N is greater than collection length or is negative.
+
+[ArgumentException](https://learn.microsoft.com/en-us/dotnet/api/system.argumentexception) is raised if collection is empty.
+
+[ArgumentOutOfRangeException](https://learn.microsoft.com/en-us/dotnet/api/system.argumentoutofrangeexception) is raised if the `randomizer` function returns a `float` value that is less than `0.0` or greater or equal to `1.0`.
+
+Example:
+```fsharp
+let allPlayers = [ "Alice"; "Bob"; "Charlie"; "Dave" ]
+let round1Order = allPlayers |> List.randomSample 3 // ["Charlie", "Dave", "Alice"]
+```
+
+# Drawbacks
+
+Users may be tempted to use some of the recently added method of `System.Random` that also apply to collections instead of the ones we add in FSharp.Core. It may also be confusing to some, especially since the naming over there is slightly different. See [.NET 8 What's New](https://learn.microsoft.com/en-us/dotnet/core/whats-new/dotnet-8/runtime#methods-for-working-with-randomness).
+
+# Alternatives
+
+Not doing this.
+
+# Compatibility
+
+* Is this a breaking change? **No**
+* What happens when previous versions of the F# compiler encounter this design addition as source code? **Library function, not applicable**
+* What happens when previous versions of the F# compiler encounter this design addition in compiled binaries? **Library function, not applicable**
+* If this is a change or extension to FSharp.Core, what happens when previous versions of the F# compiler encounter this construct? **Will work as usual**
+
+# Pragmatics
+
+## Diagnostics
+
+Please list the reasonable expectations for diagnostics for misuse of this feature. **I don't see a way to misuse it**
+
+## Tooling
+
+Please list the reasonable expectations for tooling for this feature, including any of these:
+
+* Debugging
+    * Breakpoints/stepping
+    * Expression evaluator
+    * Data displays for locals and hover tips
+* Auto-complete
+* Tooltips
+* Navigation and Go To Definition
+* Colorization
+* Brace/parenthesis matching
+
+**These will work just like for other collections functions**
+
+## Performance
+
+* For existing code **Existing code won't be affected**
+* For the new features **Performance should be respected when implementing this feature, since it can be used in performance-sensitive scenarios**
+
+## Scaling
+
+Algorithmic complexity of the new features should be O(n) for list and seq functions. O(1) for most array functions (except shuffle).
+
+## Culture-aware formatting/parsing
+
+N/A
+
+# Unresolved questions
+
+N/A