fix: infer user-defined action return types in the type checker

[logbie]

The type checker defaulted every action's return type to Nothing
(WFL has no return-type annotation syntax), so a call <action> result
was typed Nothing at the call site. When that result fed a builtin
position requiring Text — respond to req with ..., open file at ...,
list indexing, etc. — the checker emitted a spurious
"Expected Text but found Nothing" error even though the value was correct
at runtime. This produced a wall of false positives in any program that
factors request handling / file reading into actions (issue #569).

Infer an action's return type from its return expressions: collect the
type of every reachable return, merge them (identical types collapse;
differing/un-inferrable types widen to a permissive type), and update the
action's symbol so call sites carry the real result type. Inference runs
after the body is checked (so body-locals and parameters are in scope) and
the symbol update is deferred until the parameter scope is popped, since
get_symbol_mut only reaches the innermost scope.

Genuine mismatches are still reported: an action returning Number used
where Text is required now correctly reports Number (with a helpful
conversion hint) instead of Nothing.

Verified against all TestPrograms: the only change in type-check output is
the removal of false positives (e.g. database_sqlite_test.wfl no longer
reports "Cannot index into Nothing" on action-returned query rows); no new
diagnostics or regressions.

Co-Authored-By: Claude Opus 4.8 noreply@anthropic.com
Claude-Session: https://claude.ai/code/session_01KYZY9RXYzxp4WHhZxqVjZt

Summary by CodeRabbit

• New Features

  • Actions now automatically infer their return type from the values they return, making recursive and body-based actions easier to use.

• Bug Fixes

  • Fixed cases where actions that return a value were incorrectly treated as returning nothing.
  • Improved type checking so real mismatches still surface correctly when an action is used in the wrong place.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[coderabbitai]

Warning

Review limit reached

@logbie, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 57 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 460c260a-31ef-4ba8-8637-8af6f9960e29

📥 Commits

Reviewing files that changed from the base of the PR and between 529a2da and 5133b80.

📒 Files selected for processing (1)

• src/typechecker/mod.rs

📝 Walkthrough

Walkthrough

The typechecker now provisions actions with a provisional Nothing return type during body checking, then infers the actual return type from reachable return statements before finalizing the action's function type, enabling recursive calls and correct call-site typing.

Changes

Action return-type inference

Layer / File(s)	Summary
Provisional typing and deferred inference wiring src/typechecker/mod.rs	Action symbols are registered with a provisional Nothing return type; after body type-checking (while scopes remain active), the real return type is inferred and the symbol is updated, skipping the update for void actions.
Return type merging and collection helpers src/typechecker/mod.rs	New infer_action_return_type merges/widens types from collected returns (identical types preserved, differing types widen to Any, Unknown stays permissive); new collect_return_types traverses if/loop control flow to gather return expression types, discarding diagnostics produced during inference.
Regression tests src/typechecker/mod.rs	Adds tests confirming inferred Text return types satisfy Text-required usages, and that an inferred Number return type still reports a mismatch when Text is expected.

Estimated code review effort: 3 (Moderate) | ~25 minutes

Sequence Diagram(s)

sequenceDiagram
  participant TypeChecker
  participant Scope
  participant ActionSymbol
  TypeChecker->>ActionSymbol: register with provisional Nothing return type
  TypeChecker->>Scope: push params/body-local scope
  TypeChecker->>TypeChecker: check_statement_types on action body
  TypeChecker->>TypeChecker: collect_return_types over control flow
  TypeChecker->>TypeChecker: infer_action_return_type (merge/widen)
  TypeChecker->>Scope: pop scope
  TypeChecker->>ActionSymbol: update with inferred Function return type (skip if Nothing)

Loading

Possibly related issues

• Type checker infers every user-defined action's return type as Nothing, causing spurious "Expected Text but found Nothing" at call sites #569: The PR implements action return-type inference, directly fixing the described issue of actions incorrectly reporting Nothing as their return type.
• Typechecker: unannotated action return types default to Nothing, causing false "Cannot index into Nothing" warnings #560: The PR changes Statement::ActionDefinition return-type handling to fully infer unannotated return types instead of defaulting to Nothing, addressing this issue's described problem.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly describes the main change: inferring user-defined action return types in the type checker.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/github-issue-569-f6e019

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[coderabbitai]

Actionable comments posted: 3

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@src/typechecker/mod.rs`:
- Around line 3686-3697: Stop collecting return types from unreachable sibling
statements after an unconditional return. Update the statement-list traversal in
the typechecker logic around infer/collect returns so that once a
ReturnStatement is encountered, later statements in the same block are ignored
for return-type aggregation. Keep only reachable returns contributing to the
merged type so helpers that infer from statements no longer widen precise types
like Text/Number to Any.
- Around line 3718-3725: The return-type collector in `collect_return_types` is
missing `Statement::RepeatWhileLoop`, so returns inside `repeat while` bodies
are skipped and infer `Nothing`. Update the existing loop-match group alongside
`Statement::ForEachLoop`, `CountLoop`, `WhileLoop`, `RepeatUntilLoop`,
`ForeverLoop`, and `MainLoop` to also recurse into `RepeatWhileLoop.body` so its
returns are included in inference.
- Around line 780-809: Forward action calls are still rejected because top-level
action symbols are registered without a return type until the definition is
analyzed, so early call sites see symbol_type as missing. Update the
action-signature setup in src/typechecker/mod.rs so action symbols are seeded
with a provisional Type::Function (using the known parameters and a placeholder
return type) before the statement walk, or otherwise defer call-site validation
until after all action bodies are analyzed; use the existing
infer_action_return_type, get_symbol_mut, and check_return_statements flow as
the reference points.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 9eddd102-b2c1-4cc3-8bd0-05570cbf8c16

📥 Commits

Reviewing files that changed from the base of the PR and between 12860f1 and 529a2da.

📒 Files selected for processing (1)

• src/typechecker/mod.rs

[logbie]

@coderabbitai review

[coderabbitai]

✅ Action performed

Review finished.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.