[cknitt]

Fixes #9073.

Analysis didn't record constructors added by type t += ..., so completion offered none for an extensible variant. They are now registered as constructors of an open type. Typed completion of an open type, in both expressions and patterns, offers the ones declared in the type's file and in the current file. The exhaustive-switch snippet for an open type ends with a _ case, since other files can extend it.

New analysis test: CompletionExtensibleVariant.

Constructors are named so that they are in scope where completion runs: unqualified inside their declaring module, Sub.Flag elsewhere in the file, and Module.Flag from another file, using source paths in namespaced projects.

Known limitation: an extension declared inside a module whose signature hides it (module Hidden: {} = { type attr += Flag }) is still offered. Filtering through the visible signature is left for a follow-up.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Kbf3gLMawVgtmiKy3TMBc3

[github-actions]

Changelog

🐛 Bug fix

• Editor completion offers the constructors of an extensible variant declared in the current file or in the type's file, and the exhaustive switch snippet for one ends with a _ case.

Semantic Diff

10 commits · Full page

Complete the constructors of extensible variants

_{e751edb · cknitt + Claude · Oct 27, 2026, 13:06 UTC}

Why

Editor analysis didn't record the constructors added by type t += ..., and an open type has none in its declaration, so completion offered no constructor where a value of the type was expected. They are now registered as constructors of an open type, and typed completion of an open type offers those declared in the type's file and in the current file. Since other files can extend the type, the exhaustive switch snippet for one adds a _ case.

Diagnostic modified (completion of an extensible variant)

type attr = ..
type attr += Str(string) | Num(int)

// let a: attr = Nu

text · tests/analysis_tests/tests/src/CompletionExtensibleVariant.res

-Nullable, Null
+Num(_), Nullable, Null

Diagnostic modified (exhaustive switch snippet)

let value: attr = Str("s")
// switch valu

rescript · After

switch value {
| Str(_) => %todo
| Num(_) => %todo
| Flag => %todo
| _ => %todo
}

Review

• Which extensions are offered? Those in the file that declares the type and in the current file; an extension in a third file is not known to the current file's analysis.

Match extension constructors by type and name them by module path

_{5639dff · cknitt + Claude · Oct 27, 2026, 17:50 UTC}

Codex review on #9085: extensions were matched to an open type by its last name only, so A.t got the constructors of B.t, and a constructor declared in a nested module was inserted without its module, which doesn't compile because an extension constructor can't be found from the expected type. An extension now records the type it extends and is matched by that type's file and stamp, and it is named by its module path, for example Sub.Flag; a prefix may match the last part.

Name extension constructors relative to where completion runs

_{f2fd425 · cknitt + Claude · Oct 27, 2026, 20:29 UTC}

Codex review on #9085: extension constructors were named relative to the root of their file and to the file of the completed value. Inside the declaring module that inserted Sub.Flag, which is not in scope there, and for a value imported from another file the constructors were not qualified with its module, while extensions in the edited file were missed. A completion request now records the edited file and the module path at the cursor; constructors of that file drop the module path shared with the cursor, constructors of other files get their module's name, and extensions are collected from the edited file.

Qualify extension constructors of namespaced modules by source path

_{a62bfe5 · cknitt + Claude · Oct 27, 2026, 21:45 UTC}

Codex review on #9085: the module name of another file comes from its compiled name, which in a namespaced project is Base-MyNamespace, so completion inserted Base-MyNamespace.Flag. The name is turned into its source path MyNamespace.Base first.

Offer extensions in scope at the cursor and from interface modules

_{e6a28af · cknitt + Claude · Oct 28, 2026, 07:47 UTC}

Codex review on #9085: an extension of the edited file declared after the cursor was offered although it isn't in scope there, and an extension inside a module of an interface (module Sub: { type attr += Flag }) was never registered, because only typed-tree extensions were. The completion scope now records the cursor position and skips later extensions of the edited file, and extensions in signatures are registered too.

Qualify extensions of nested interface modules and complete their payloads

_{52ee62d · cknitt + Claude · Oct 29, 2026, 06:34 UTC}

Codex review on #9085: an extension two modules deep in an interface got only the outer module in its path, because the signature walk didn't extend the module path when entering a module, and completion inside the payload of a qualified constructor such as Sub.Toggle(...) found no constructor, because the payload position keeps only Toggle. Signature modules now extend the path, and payload lookup also matches the last part of a qualified constructor name.

Type extension payloads in their own file and match qualified constructors

_{b03db7b · cknitt + Claude · Oct 29, 2026, 10:10 UTC}

Codex review on #9085: a payload type declared next to the extension, as in type Base.attr += Color(color) with a local color, was resolved in the open type's file, so payload completion lost it; and of two same-named extensions, B.Value(...) completed the payload of the first one, A.Value, because the payload position kept only Value. Each extension constructor now carries the environment of its file, and the payload position keeps the written path: the constructor sharing the longest suffix with it is meant, ties going to the most recent open, then to the last declared one.

Look up unqualified extension payloads in the scope at the cursor

_{d56bea1 · cknitt + Claude · Oct 29, 2026, 10:33 UTC}

Codex review on #9085: ranking same-named extension constructors let an open outrank a local extension declared after it, so after open A and a later local Value(...), completion used A.Value's payload; and an extension inside a module type, module type S = { type attr += Phantom }, was offered as S.Phantom and added to exhaustive switches. A qualified constructor is now the one its written path names, and an unqualified one is looked up in the completion scope as the compiler does, which now records extension constructors, the most recent declaration or open winning. Extensions in module types are skipped.

Keep a catch-all when expanding or switching on an open type

_{fb97e50 · cknitt + Claude · Oct 29, 2026, 18:02 UTC}

Why

Codex review on #9085: now that an open type extracts as a variant, the Expand catch-all and Exhaustive switch code actions replaced or built the switch from the constructors known here, dropping the _ that matches constructors added in other files.

Codeaction modified (Expand catch-all)

type attr = ..
type attr += Str(string) | Num(int)
let f = (value: attr) =>
  switch value {
  | Str(_) => "str"
  | _ => "other"
  }

text · the _ is replaced by

-Num(_)
+Num(_) | _

Codeaction modified (Exhaustive switch)

type attr = ..
type attr += Str(string) | Num(int)
let value: attr = Str("s")
value

rescript

 switch value {
 | Str(_) => %todo
 | Num(_) => %todo
+| _ => %todo
 }

Review

• Why keep the _ as an alternative instead of a separate case? Replacing only the pattern needs no copy of the case body, and the expanded or-pattern lists the known constructors for splitting by hand.

Resolve an open's extension constructor only through its module

_{347aa66 · cknitt + Claude · Oct 29, 2026, 18:03 UTC}

Why

Codex review on #9085: an open A after a local extension Value resolved the payload of an unqualified Value(...) to the local constructor. Matching A.Value against the constructors also accepted the bare local Value as a suffix, and the later declaration won. A constructor named without a module is declared around the cursor, so an open no longer brings it into scope.

Completion modified (extension payload after an open)

type attr = ..
module A = {
  type attr += Value(bool)
}
type attr += Value(option<int>)
open A
let x: attr = Value()

text · payload completion inside Value() completes

-option<int>, the payload of the local Value
+bool, the payload of A.Value: true, false

[codecov]

Codecov Report

❌ Patch coverage is 89.88764% with 18 lines in your changes missing coverage. Please review.
✅ Project coverage is 81.76%. Comparing base (25e3dfb) to head (347aa66).

Files with missing lines	Patch %	Lines
analysis/src/type_utils.ml	86.11%	15 Missing ⚠️
analysis/src/xform.ml	80.00%	2 Missing ⚠️
analysis/src/completion_back_end.ml	96.42%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #9085      +/-   ##
==========================================
+ Coverage   81.72%   81.76%   +0.03%     
==========================================
  Files         460      460              
  Lines       64817    64973     +156     
==========================================
+ Hits        52969    53122     +153     
- Misses      11848    11851       +3     

Files with missing lines	Coverage Δ
analysis/src/completion_expressions.ml	86.88% <100.00%> (+0.33%)	⬆️
analysis/src/completion_front_end.ml	81.83% <100.00%> (+0.22%)	⬆️
analysis/src/completion_patterns.ml	88.88% <100.00%> (+0.31%)	⬆️
analysis/src/process_cmt.ml	85.67% <100.00%> (+1.25%)	⬆️
analysis/src/shared_types.ml	80.68% <ø> (ø)
analysis/src/completion_back_end.ml	81.83% <96.42%> (+0.34%)	⬆️
analysis/src/xform.ml	88.39% <80.00%> (-0.05%)	⬇️
analysis/src/type_utils.ml	68.43% <86.11%> (+2.26%)	⬆️

... and 5 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[pkg-pr-new]

Open in StackBlitz

rescript

npm i https://pkg.pr.new/rescript@9085

@rescript/belt

npm i https://pkg.pr.new/@rescript/belt@9085

@rescript/darwin-arm64

npm i https://pkg.pr.new/@rescript/darwin-arm64@9085

@rescript/linux-arm64

npm i https://pkg.pr.new/@rescript/linux-arm64@9085

@rescript/linux-x64

npm i https://pkg.pr.new/@rescript/linux-x64@9085

@rescript/runtime

npm i https://pkg.pr.new/@rescript/runtime@9085

@rescript/win32-x64

npm i https://pkg.pr.new/@rescript/win32-x64@9085

commit: 347aa66

[github-actions]

Developer playground preview: https://rescript-lang.github.io/rescript/dev-playground/?version=pr-9085

[cknitt]

@codex review

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-29T18:07:23.772194Z	417ad95	Manual request

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 98ecc79d93

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cknitt]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f678bb32e5

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cknitt]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 56cff12ba2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cknitt]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3408b27a6e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cknitt]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 84e1386175

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cknitt]

@codex review

[cknitt]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3cb8834566

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cknitt]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 462f92c4a4

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cknitt]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b1191669a3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cknitt]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 417ad95d74

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Resolve module aliases when matching payload constructors

When an extension constructor is accessed through a module alias, textual path matching cannot find its payload. For example, with module B = { type attr += Value(bool) } and module Alias = B, completing inside Alias.Value(...) compares ["Alias"; "Value"] against the recorded ["B"; "Value"]; neither comparison succeeds, so boolean payload suggestions disappear in both expressions and patterns. Opening Alias has the same problem. Resolve the written module path before selecting the constructor. The fresh counterexample uses a module alias, which the existing direct-qualified-path tests do not cover.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Keep the completion scope active during code-action extraction

For code actions inside a nested module, this wrapper restores completion_scope before extension constructors are collected: Xform.extract_type_from_expr calls process_completable at lines 14–17, but calls Type_utils.extract_type afterward at lines 31–36. Consequently, an “Exhaustive switch” or “Expand catch-all” action inside module Sub uses an empty scope path and inserts Sub.Flag for an extension declared in Sub, where that qualification is not accessible. Constructors declared after the action position are also no longer filtered. Extend the scope lifetime through the code action's type extraction, or pass the context explicitly. This is fresh evidence from the code-action consumer, rather than the completion paths covered by the earlier fixes.

Useful? React with 👍 / 👎.