pm-lang: type checking (v1)

[sean-parent]

Summary

First of three Phase 3 sub-plans for the pm-lang language server design
(docs/superpowers/specs/2026-07-17-pm-lang-language-server-design.md), split out per
writing-plans' scope-check guidance rather than one plan spanning the type checker, the new
pm-lsp crate, and the VS Code extension.

• cel_parser::Ty — a minimal static type model (built-in primitives + Ty::Any), with
  Literal/TypeId conversions and a unifies_with predicate where Any unifies silently with
  everything in both directions.
• cel_parser::op_table::{OperandTypes, builtin_operand_types} — projects the existing runtime
  operator-dispatch tables into a checker-usable shape, so the static checker and the runtime
  dispatcher read one shared source of truth instead of a second, hand-duplicated table.
• cel_parser::ty::check_expr — checks Expr::Op (via builtin_operand_types) and Expr::Logical
  (fixed &&/|| bool semantics) directly; recurses into Apply/Tuple/TupleIndex/If so a
  nested Op error still surfaces, but infers those nodes themselves as Ty::Any — checking
  call/tuple/if shapes is deferred to a later phase.
• pm_lang::typecheck::check_sheet — checks each cell's literal initializer against its
  : type_name annotation, and each relationship/conditional method's body against its
  declared outputs (arity: single value vs. n-tuple, including a tuple-shaped body for a
  single-output method; and per-output type).

Plan doc: docs/superpowers/plans/2026-07-18-pm-lang-type-checking-v1.md.

Built via subagent-driven development: each of the 4 tasks was implemented and independently
reviewed (spec compliance + code quality), followed by a final whole-branch review. One Minor
polish commit addresses that review's findings (missing # Examples doc blocks, a
ParseError::new/new_range inconsistency, the single-output tuple-body gap, and unifying the
operator-routing logic between the checker and the runtime dispatcher).

Test plan

• cargo build --workspace — zero warnings
• cargo test --workspace and cargo test --doc --workspace — all passing
• cargo clippy --workspace --exclude begin --all-targets -- -D warnings
• cargo clippy -p begin --no-default-features --all-targets -- -D warnings
• cargo clippy -p begin --all-targets -- -D warnings
• cargo fmt --all — no diff

🤖 Generated with Claude Code

---

Note

Medium Risk
New diagnostic semantics can disagree with runtime until LSP integration is designed (ParseError is !Send), but changes are additive and do not alter PmParser execution; incorrect Ty::Any unification would cause false negatives on operator/cell checks.

Overview
Adds Phase 3 sub-plan 1: a best-effort static type checker on top of existing PmAstParser ASTs, intended for upcoming pm-lsp diagnostics (not wired here).

In cel-parser, introduces Ty (built-in primitives + Ty::Any with bidirectional silent unification) and ty::check_expr, which validates Expr::Op against builtin_operand_types and Expr::Logical for bool operands while recursing through apply/tuple/if nodes but inferring those as Any. op_table gains OperandTypes / builtin_operand_types and a shared signatures_for helper so runtime BuiltinScope::lookup and the checker read the same overload tables (no duplicated signature knowledge).

In pm-lang, check_sheet maps declared cell types via TypeRegistry, checks cell literal initializers against annotations (aligned with runtime parse_literal_as rules, including unsuffixed ints on unsigned types), and checks relationship/conditional method bodies against declared output arity and per-output types—including rejecting a tuple body when only one output is declared.

Docs/handoff and the implementation plan are updated to mark this sub-plan complete and note follow-ons (pm-lsp, VS Code extension, ParseError → CELError for async LSP).

^{Reviewed by Cursor Bugbot for commit 64dbb20. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 949032e. Configure here.}

[Copilot]

Pull request overview

Adds pm-lang type checking (v1) across pm-lang sheets and CEL expression ASTs, providing best-effort static diagnostics (intended groundwork for a future pm-lsp integration) while preserving CEL/pm-lang extensibility via Ty::Any.

Changes:

• Introduces cel_parser::Ty plus cel_parser::ty::check_expr to infer/check types for Expr::Op (builtin overload tables) and Expr::Logical (&&/|| bool semantics), while recursing through other nodes to surface nested operator errors.
• Exposes builtin operand type signatures from cel-parser’s runtime dispatch tables via OperandTypes + builtin_operand_types, and shares routing logic between the checker and runtime lookup.
• Adds pm_lang::check_sheet to validate cell literal initializers vs annotations and method bodies vs declared outputs (arity + per-output type), plus a detailed implementation plan doc.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
pm-lang/src/typecheck.rs	New check_sheet entry point + tests for pm-lang sheet-level type diagnostics.
pm-lang/src/lib.rs	Wires typecheck module and re-exports check_sheet.
cel-parser/src/ty.rs	Adds Ty model and check_expr checker over Expr trees (with doctests + unit tests).
cel-parser/src/op_table.rs	Exposes OperandTypes/builtin_operand_types and factors shared operator→signature routing for checker + runtime.
cel-parser/src/lib.rs	Exports ty module and re-exports Ty, OperandTypes, builtin_operand_types.
docs/superpowers/plans/2026-07-18-pm-lang-type-checking-v1.md	Adds the Phase 3 sub-plan writeup for implementing type checking v1.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.