Record constructors (FS-1073) by charlesroddie · Pull Request #19974 · dotnet/fsharp

charlesroddie · 2026-06-19T10:47:23Z

Implements FS-1073. F# records compile to a class whose all-fields constructor is callable from C# (new MyRecord(a, b)) but not from F#, which only permits { Field = … }. This adds a RecordConstructorSyntax preview language feature that surfaces that constructor to F#, with positional and named arguments.

Notes:

Accessibility mirrors { } construction, so a private/internal record's representation is not bypassed (unlike C#'s public IL constructor).
An existing same-named value binding takes precedence over the constructor (FS-1073 scope rule).
Behind --langversion:preview.
Pattern matching is out of scope.

Records compile to a class whose all-fields constructor is callable from C# (new MyRecord(a, b)) but not from F#, which only allows { Field = ... } syntax. This adds a RecordConstructorSyntax preview feature that surfaces that constructor to F# too, supporting positional and named arguments. Implemented via a new MethInfo.RecdAllFieldsCtor case surfaced by InfoReader for record tycons; it elaborates through the existing mkRecordExpr path, so there is no codegen or overload-resolution change. Accessibility mirrors { } construction (the ctor is no more accessible than the record's representation/fields), so the C# behaviour of a public IL constructor bypassing a private record is not inherited. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

github-actions · 2026-06-19T10:48:03Z

❗ Release notes required

You can open this PR in browser to add release notes: open in github.dev

✅ Found changes and release notes in following paths:

Change path Release notes path Description

`src/Compiler` docs/release-notes/.FSharp.Compiler.Service/11.0.100.md

`src/Compiler/Facilities/LanguageFeatures.fsi` docs/release-notes/.Language/preview.md

Shorter name; a record has exactly one synthesized constructor, so the 'AllFields' qualifier is not needed to disambiguate. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

The feature surfaces only the all-fields constructor. A struct record's zero-init default and a [<CLIMutable>] record's IL parameterless .ctor must remain non-callable from F#; both 'Point()' and 'R()' are rejected (FS0501). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…terless ctor A struct record's 'Point()' is default (zero) initialization, not a real constructor; only [<CLIMutable>] emits an actual parameterless .ctor. Name the two tests accordingly. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

rangeOfMethInfo had no RecdCtor case, so it fell through to ArbitraryValRef (None) and GoToDefinition on a positional record-constructor call navigated nowhere. Add a RecdCtor arm returning the record type's range, mirroring DefaultStructCtor. Adds FSharp.Compiler.Service.Tests smoke tests: go-to-definition lands on the record type, the tooltip mentions the type, and find-all-references links the constructor call to the type declaration. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Regression guard: the positional constructor has no field labels, so a [<RequireQualifiedAccess>] record must construct without a spurious diagnostic. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…073) Library exposes a record and an inline constructor function; the app constructs records positionally and via the inline function. The new syntax is gated behind RECORD_CTOR_FEATURE / --langversion:preview for local builds, with a classic { } fallback so SDK-compiler scenarios still build. Both branches elaborate to the same record-allocation node, so the pickled representation is unchanged and the matrix exercises both directions (feature-enabled consumer, older consumer). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

The constructor is not a declared member, so it rides on the record's representation visibility through a signature: available when the .fsi exposes the representation, rejected (FS1133, like { }) when the .fsi hides it. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…ents Per review: rename the per-feature RECORD_CTOR_FEATURE define to the reusable USES_PREVIEW_COMPILER, and remove the explanatory comments so the addition is compact. No behavior change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

A positional record constructor quotes as NewRecord (R, Value (1), Value (2)), identical to { A = 1; B = 2 } - both lower to the same node before quotation translation, so no constructor call appears in the quotation. Uses Console.WriteLine + Expr.ToString() to avoid sprintf/printf. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…tional-ctor # Conflicts: # src/Compiler/Facilities/LanguageFeatures.fs # src/Compiler/Facilities/LanguageFeatures.fsi

main added a CSharpExtensionTypeDisplay parameter to layoutMethInfoCSharpStyle and updated its callers; the RecdCtor arm (new in this branch) was auto-merged with the old call. Pass extTypeDisplay to match the sibling arms. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

charlesroddie · 2026-07-20T18:08:00Z

This PR OK @T-Gro ? Seems straightforward to me.

…tional-ctor # Conflicts: # src/Compiler/Facilities/LanguageFeatures.fs # src/Compiler/Facilities/LanguageFeatures.fsi

charlesroddie · 2026-07-29T08:33:20Z

PR updated @abonie

…tional-ctor

…tional-ctor # Conflicts: # src/Compiler/xlf/FSComp.txt.cs.xlf # src/Compiler/xlf/FSComp.txt.de.xlf # src/Compiler/xlf/FSComp.txt.es.xlf # src/Compiler/xlf/FSComp.txt.fr.xlf # src/Compiler/xlf/FSComp.txt.it.xlf # src/Compiler/xlf/FSComp.txt.ja.xlf # src/Compiler/xlf/FSComp.txt.ko.xlf # src/Compiler/xlf/FSComp.txt.pl.xlf # src/Compiler/xlf/FSComp.txt.pt-BR.xlf # src/Compiler/xlf/FSComp.txt.ru.xlf # src/Compiler/xlf/FSComp.txt.tr.xlf # src/Compiler/xlf/FSComp.txt.zh-Hans.xlf # src/Compiler/xlf/FSComp.txt.zh-Hant.xlf

T-Gro · 2026-08-04T09:00:51Z

🤖🕵️
@charlesroddie Re the synthesized record constructor surfacing: it's emitted unconditionally, so it leaks into signature-file generation (SigGenerationRoundTripTests fail with non-compiling .fsi — FS0193/FS0501) and flips FS1133→FS0501 under the default langversion. Exclude RecdCtor from .fsi/NicePrint output and gate its surfacing behind the RecordConstructorSyntax langversion, not just the call site.

The synthesized all-fields constructor was surfaced unconditionally, leaking into signature-file generation (SigGenerationRoundTripTests produced non-compiling .fsi) and changing name resolution on released langversions. Gate the surfacing on the RecordConstructorSyntax langversion in InfoReader (not just at the call site), exclude RecdCtor from NicePrint/.fsi output, and drop the now-redundant call-site feature check. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

…tional-ctor # Conflicts: # src/Compiler/Facilities/LanguageFeatures.fs

github-actions · 2026-08-04T19:00:30Z

🔍 Tooling Safety Check — Affects-Build-Infra, Affects-Compiler-Output
Affects-Build-Infra: CompilerCompat .fsproj files add PropertyGroup with LangVersion/DefineConstants
Affects-Compiler-Output: Modifies compiler checking/name resolution to add record constructor syntax

Generated by PR Tooling Safety Check · opus46 3.9M · ◷

…eLine - infos.fs: build the single tupled parameter group with List.singleton instead of a two-line list literal - NicePrint.fs: split the constructor filter one conjunct per line - CompilerCompat Program.fs: use Console.WriteLine in the record-constructor block (printn is not yet available) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

- InfoReader.fs: match ValueNone rather than a wildcard in the RecdCtor XML-doc-signature arm - Trim the RecdCtor comments in InfoReader/NicePrint/MethodCalls to the non-obvious rationale Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

charlesroddie · 2026-08-06T07:54:07Z

🤖🕵️ @charlesroddie Re the synthesized record constructor surfacing: it's emitted unconditionally, so it leaks into signature-file generation (SigGenerationRoundTripTests fail with non-compiling .fsi — FS0193/FS0501) and flips FS1133→FS0501 under the default langversion. Exclude RecdCtor from .fsi/NicePrint output and gate its surfacing behind the RecordConstructorSyntax langversion, not just the call site.

Thanks for the advice! Was a bit stuck on this. Done now.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

charlesroddie · 2026-08-07T10:53:17Z

Fixed spacing so needs approval again @T-Gro @abonie thanks!

charlesroddie requested a review from a team as a code owner June 19, 2026 10:47

github-project-automation Bot added this to F# Compiler and Tooling Jun 19, 2026

github-project-automation Bot moved this to New in F# Compiler and Tooling Jun 19, 2026

charlesroddie and others added 2 commits June 19, 2026 11:58

Rename MethInfo.RecdAllFieldsCtor to RecdCtor

d3e1804

Shorter name; a record has exactly one synthesized constructor, so the 'AllFields' qualifier is not needed to disambiguate. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Add release notes for record constructors (FS-1073)

402baa2

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

github-actions Bot added the AI-Tooling-Check-Scanned-Clean Tooling check: diff analyzed, no interesting infrastructure files label Jun 19, 2026