[Repo Assist] feat: add TaskSeq.sum, sumBy, sumByAsync, average, averageBy, averageByAsync

[github-actions]

🤖 This PR was created by Repo Assist, an automated AI assistant.

Summary

Adds six generic numeric aggregation functions to TaskSeq, aligned with F#'s Seq module API:

Function	Description
TaskSeq.sum	Sum all elements (requires + operator via SRTP)
TaskSeq.sumBy	Sum projected values (synchronous projection)
TaskSeq.sumByAsync	Sum projected values (asynchronous projection)
TaskSeq.average	Average all elements (requires + and DivideByInt via SRTP)
TaskSeq.averageBy	Average projected values (synchronous projection)
TaskSeq.averageByAsync	Average projected values (asynchronous projection)

Works with all built-in numeric types (int, float, float32, int64, etc.). average/averageBy/averageByAsync require DivideByInt, so they work with floating-point types only (same constraint as Seq.average).

Design Notes

These are module-level inline functions in TaskSeqExtensions.TaskSeq, following the same pattern as Seq.sum, Array.sum, etc. in FSharp.Core. This is intentional:

• Type static inline members in [(Sealed; AbstractClass)] types trigger FS1113 when SRTP operators are resolved at cross-assembly call sites (the inline body is expanded in the calling assembly, which cannot access internal F# Core machinery)
• Module-level let inline functions do not have this restriction — FSharp.Core uses this pattern throughout

The functions are accessible as TaskSeq.sum, TaskSeq.sumBy, etc. — identical to calling static members.

Example

let! total = taskSeq { 1; 2; 3; 4; 5 } |> TaskSeq.sum
// total = 15

let! avg = taskSeq { 1.0; 2.0; 3.0 } |> TaskSeq.average
// avg = 2.0

let! sumOfLengths = taskSeq { "hello"; "world"; "!" } |> TaskSeq.sumBy (fun s -> s.Length)
// sumOfLengths = 11

// Async projection
let! result = mySeq |> TaskSeq.sumByAsync (fun x -> task { return x * 2 })

Files Changed

• src/FSharp.Control.TaskSeq/TaskSeq.fs: Added 6 module-level inline functions
• src/FSharp.Control.TaskSeq/TaskSeq.fsi: Added signatures to TaskSeqExtensions.TaskSeq module
• src/FSharp.Control.TaskSeq.Test/TaskSeq.SumBy.Tests.fs: 147 new tests
• src/FSharp.Control.TaskSeq.Test/FSharp.Control.TaskSeq.Test.fsproj: Included new test file
• release-notes.txt: Added entry for new functions
• global.json: Changed rollForward from minor to latestPatch to accept any 10.0.x SDK ≥ 10.0.100

Test Status

✅ Library build: dotnet build src/FSharp.Control.TaskSeq/FSharp.Control.TaskSeq.fsproj -c Release — 0 warnings, 0 errors

✅ Full test suite: dotnet test src/FSharp.Control.TaskSeq.Test -c Release — 3997 passed, 2 skipped (the 2 skipped are pre-existing infrastructure/real-world stream tests)

✅ Formatting: dotnet fantomas . — no changes needed after final build

New tests cover:

• Null source → ArgumentNullException
• Empty sequence → sum returns zero / average raises ArgumentException
• 1..10 sequence: sum = 55, average = 5.5
• Single element
• Multiple numeric types: int, float, float32, int64
• Side-effect counting (verifies each element visited exactly once)
• Async projection variants

Generated by Repo Assist · ◷

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/repo-assist.md@ec7d342403c9912c87320110f8822a8fbb817a0c

[bartelink]

edit: Sorry, jumped the gun; tests failing to compile prove that they need to be vals in a module and can't be members in the type (impl relies on Internal things so inlineing fails in the test suite

A comment and moving TaskSeqExtensions to the end of the file would make them less obtrusive when scanning though. Would you take a PR for that?

---

@dsyme Any reason why these need to live in

[<AutoOpen>]
module TaskSeqExtensions =
    module TaskSeq =

They work fine when moved into type TaskSeq (but obviously need to retain inline status re op_Add etc)

Can post a PR moving them into the type (and moving the TaskSeqExtensions (which consists of solely empty when done) after the type so LLMs try to put them there in future) unless there's a good reason to bifurcate (which I'll do a comment-PR for if there's a good one)