feat: specialize constant matches in generic procedures

Reuse checked match selection for generic constant subjects and explicit match $(). Preserve owned captures, runtime scope and control transfers; keep ordinary calls, mutable subjects and potential runtime defaults unprobed. Update README and TODO checkpoint boundaries.

Verified: bro build; bro test (250 passed); python3 tests/native.py (555 interpreter/native comparisons at -O0/-O2, allocator, LLVM and CLI checks); direct interpreter/native constant-match smoke (42).
This commit is contained in:
2026-10-27 12:53:19 +01:00
parent 8d1a98f982
commit c3e10b3bed
5 changed files with 345 additions and 10 deletions
+4 -2
View File
@@ -102,16 +102,18 @@ type-producing bodies support mutable type and value locals, direct binding assi
type-producing bodies also execute scalar/enum `match` statements and type-valued match expressions in initializers, assignments and returns, such as `return match wide { true: i16, false: i8 }`. selected expression arms may themselves be matches. an ordinary checked IR match returns the selected arm index, preserving subject/pattern typing, range boundaries, overlap, unreachable-arm and exhaustiveness diagnostics without checking unexecuted bodies. statement arms retain scoped locals, assignments, early returns and enclosing-loop break/continue; expression selection shares the type-control step bound.
matches also select nested types, array lengths, generic arguments, callback signatures and field defaults in type-producing bodies. for example, `return struct { values [2](match wide { true: i16, false: i8 }) }` generates concrete storage without elaborating the unselected arm. snapshots preserve loop-local type/value bindings, selected literals retain their spelling and source provenance, and nominal results are cached before their recursive fields are expanded. deferred type selection shares the producer's execution bound; failed elaboration retains its original diagnostic without leaving unchecked generated types behind. general type-valued storage and broader generic-body match specialization remain open.
matches also select nested types, array lengths, generic arguments, callback signatures and field defaults in type-producing bodies. for example, `return struct { values [2](match wide { true: i16, false: i8 }) }` generates concrete storage without elaborating the unselected arm. snapshots preserve loop-local type/value bindings, selected literals retain their spelling and source provenance, and nominal results are cached before their recursive fields are expanded. deferred type selection shares the producer's execution bound; failed elaboration retains its original diagnostic without leaving unchecked generated types behind. general type-valued storage and inline expansion remain open.
tagged-union matches in type-producing bodies support immutable by-value payload captures in statements, expressions and nested type/storage positions. for example, `match choice { .count |length|: return [length]i16, .empty: return bool }` uses a captured `usize` to generate storage. grouped variants require identical nonvoid payload types; invalid captures in unselected arms are still diagnosed. selected payloads become owned typed constants, preserving narrow integers, nested aggregates and strings across generated lengths/defaults, loop snapshots and repeatable compilation. pointer/slice/callback-bearing active payloads retain the constant-materialization escape boundary below.
ordinary procedures may compute integer, boolean, float and string compile-time arguments and return nested records/tuples, fixed arrays, tagged unions and explicit fallible results from constant inputs, such as `Fixed(i16, length(2))`. calls reuse the ordinary checker, typed IR and interpreter, preserving local type inference, checked arithmetic, mutable locals, range/while loops, break/continue and recursive or specialized helper calls. aggregate and fallible arguments receive their declared context, including dependent `[N]T` parameters; cached specializations compare normalized values, not literal spelling or evaluation identity. returned constants become owned source literals before temporary interpreter storage is released, preserving by-value copies, exact integer types, string bytes, enum/optional fields and IEEE float bits. tagged unions retain their actual discriminant and only the active payload, including nested aggregates and void/zero-cell variants. fallible constants retain success/error identity, composed or payload-bearing errors and void success; `catch` can consume them in generated lengths, defaults and runtime expressions.
`$(expression)` explicitly evaluates a constant expression, including an ordinary procedure call, even in a package without templates. use a typed expression when its shape cannot be inferred independently, such as `$(Point { x = 40 })`; integer literals retain contextual typing. the resulting value works in runtime expressions, bindings, expression statements and array lengths, while the erased expression still receives ordinary checking. each interpreter evaluation is isolated and limited to 100,000 IR instructions/terminators and 128 active frames; failures retain their originating source locations. runtime parameters and mutable local/global dependencies cannot supply compile-time inputs. data pointers, slices, callbacks and untagged unions cannot escape temporary evaluation as generated constants; active payloads or aggregate fields containing unsupported values are diagnosed rather than retaining interpreter handles. untagged unions lack an active-member identity, so selecting a source variant could discard observable bytes. executed foreign calls remain interpreter traps. general type-valued execution, broader compile-time match/`inline` expansion, reflection, generated type/default arrays and existing generic-library acceptance remain open.
`$(expression)` explicitly evaluates a constant expression, including an ordinary procedure call, even in a package without templates. use a typed expression when its shape cannot be inferred independently, such as `$(Point { x = 40 })`; integer literals retain contextual typing. the resulting value works in runtime expressions, bindings, expression statements and array lengths, while the erased expression still receives ordinary checking. each interpreter evaluation is isolated and limited to 100,000 IR instructions/terminators and 128 active frames; failures retain their originating source locations. runtime parameters and mutable local/global dependencies cannot supply compile-time inputs. data pointers, slices, callbacks and untagged unions cannot escape temporary evaluation as generated constants; active payloads or aggregate fields containing unsupported values are diagnosed rather than retaining interpreter handles. untagged unions lack an active-member identity, so selecting a source variant could discard observable bytes. executed foreign calls remain interpreter traps. general type-valued execution, inline match/loop expansion, reflection, generated type/default arrays and existing generic-library acceptance remain open.
`if $(condition) { ... } else { ... }` evaluates a boolean guard before expanding or checking either body, including calls to ordinary constant-input procedures. concrete generic specializations also select branches whose guards use constants, immutable constant bindings, type comparisons or substituted compile-time parameters. only the selected body is specialized and checked; both bodies are still parsed and name-resolved. selected bodies retain their scope, cleanup and return behavior, and erased guards retain ordinary type checking and source diagnostics. ordinary calls in an implicit guard stay runtime; use `$()` to request their evaluation. optional captures remain runtime, and ordinary nongeneric `if (true)` still checks both bodies.
`match $(subject) { ... }` selects an expression or statement arm before specializing or checking its body, including ordinary constant-input calls. concrete generic procedures also select matches on side-effect-free constant subjects, including substituted scalar/enum/tagged-union arguments and pure field/index projections. ordinary checked IR validates every pattern and capture declaration before selection; only the selected body is elaborated, while all bodies remain parsed and name-resolved. by-value payload captures retain exact types and owned strings, and selected bodies may use runtime parameters, nested matches, cleanup, returns and enclosing-loop transfers. ordinary call subjects, runtime parameters and mutable locals stay runtime, as do contextual aggregate bindings that might omit runtime defaults; ordinary nongeneric literal matches still check all arms. use `$()` to request explicit evaluation. pointer/slice/callback-bearing payloads retain the escape boundary above.
type `==` and `!=` comparisons in compile-time expressions and specializations preserve aliases, nominal declaration identity and structural pointer/array/callback signatures. composed error-set identity and specialization caching ignore member ordering, grouping and repetition. for example, `read proc($T type, value T) i16 { if (T == i16) return value else return value.count }` checks the scalar or record path for each concrete call without requiring both operations on every `T`.
callback signatures use `@proc(value i16) i16`, including grouped parameter names and trailing commas. compatible named functions can initialize callbacks, which may be passed, returned, replaced and stored in records/arrays; `Callback :: alias @proc(value i16) i16` names a reusable signature. parameter/result types match structurally, retaining nominal record identity and exact pointer mutability. callback fields call directly, such as `allocator.vtable.alloc(context, size, alignment)` once the remaining argument types are supported. `@mut proc` and `*proc` are invalid; `@mut Callback` instead points to writable callback storage.