Skip to content

v35.3.2 — Analyzer fix: char_at returns Type::Char (was Str)

Choose a tag to compare

View all tags
@fajarkraton fajarkraton released this 08 May 23:02
· 92 commits to main since this release
v35.3.2
3dfa24d

[v35.3.2] — 2026-05-09 🐛 Analyzer fix: String::char_at returns Type::Char (was incorrectly grouped as Type::Str) — patch

Bugfix: s.char_at(i) == 'X' was rejected by the analyzer with
SE004: type mismatch: expected str, found char even though the
interpreter correctly returned Value::Char. Pre-v35.3.2 the
analyzer grouped substring | char_at together both returning
Type::Str (src/analyzer/type_check/check.rs:2773); v35.3.2 splits
the grouping so char_at correctly returns Type::Char.

Pre-fix vs post-fix

fn main() {
    let s: str = "hello"
    let c = s.char_at(0)
    if c == 'h' { println("OK") } else { println("FAIL") }
}
  • Pre-v35.3.2: SE004: type mismatch: expected str, found char
  • Post-v35.3.2: OK

What's NOT in this patch (deferred)

The original "language fix #4 lexer perf" pending item assumed a
deeper migration (~2-4h, daily-impact 24× speedup for self-host
lexer). The B0 audit (docs/V35_3_2_LEXER_PERF_B0_FINDINGS.md)
revealed:

  • char_at IS implemented; the analyzer mistype was the only blocker
  • Stdlib cascade migration would touch 119 sites (43 lexer.fj
    • 76 parser_ast.fj) — 3× the original "41 sites" memory estimate
  • True 10-20× perf gain needs a NEW byte_at(s, i) -> i64 builtin
    (numeric compare, no allocation)

Three deferred options for future ships:

  • Option A (v35.3.3 or v35.4.0): cascade-migrate stdlib to use
    char_at + char literals. ~1.5-2h. ~5-10× lexer perf gain.
  • Option B (v35.4.0 minor): add byte_at builtin + cascade-migrate
    to numeric compare. ~2-3h. ~10-20× lexer perf gain.

v35.3.2 unblocks future perf work by making char_at correctly
usable for new code; deeper cascade scope-decisions deferred to
their own design pass.

Risks (per CLAUDE.md §6.8) — all NONE realized

  • No stdlib/*.fj nor codegen touched → Stage 2 byte-equality unaffected
  • Lib tests: 7,633 PASS (no regression; no-op for code that didn't compare char_at result)
  • Existing char_at callers: 0 usage in stdlib pre-fix (verified via grep) → no compatibility risk

Source of truth

  • docs/V35_3_2_LEXER_PERF_B0_FINDINGS.md — B0 audit + 4 scope options + Option C rationale + 119-site cascade scope for future Options A/B
Assets 2
Loading