std/src/runtime/char.nv - main - nv-lang/nova

std/src/runtime/char.nv
60 строк3 KB
Evgeniy Golovin
style(225.1): вертикальный ритм — sweep классов a-e (std/src)
26 июл 2026, 02:46
26 июл 2026, 02:4638ecf16
// std/runtime/char.nv — char ↔ str (UTF-8 encode/decode) + int/char/u8
// scalar conversions (D54/D77).
//
// [M-compiler-nv-porting-wave] (2026-07-07) item A: NOT auto-gen anymore.
// Was mixed (str.from/str.from_codepoint auto-gen'd by `nova-codegen
// emit-runtime-stubs` from runtime_registry.rs + char.from/u8.try_from
// Nova-implemented via a `nova_body: Some("...")` Rust string literal
// rendered verbatim into this file). Auto-gen writes exactly ONE file per
// module (single_file form, `module_to_path`) — it cannot coexist with a
// hand-maintained sibling for the SAME module (Nova import resolution is
// single-file XOR folder per module path; it does NOT scan a directory
// for arbitrary same-module sibling files — `sync.nv`+`sync_test.nv` only
// merge via the `*_test.nv` test-peer special case). So: same path
// StringBuilder/WriteBuffer/ReadBuffer already took (Plan 91.12/109) —
// `char_runtime()` in runtime_registry.rs now returns `vec![]`, and this
// whole file — all four declarations — is hand-maintained. Single file,
// single source of truth.

module runtime.char

// ─── str ───

// Plan 174.2 (str.from-ретракция): `export extern "nova" fn str.from(c char)`
// — the compiler-generated engine hook that used to back `str.from(char)` —
// is RETRACTED. `char` needs no concrete `@to_str()` override either: the
// bare-T blanket (`fn[T] T @to_str() -> str => "${@}"`,
// std/runtime/string/core.nv) covers it — interpolating a `char` value
// lowers DIRECTLY to `nova_char_to_str` in emit_c.rs::emit_interpolated_str
// (primitive dispatch table, same C function the retracted extern used to
// name), so `c.to_str()` is byte-identical to the old `str.from(c)`. UTF-8
// encode codepoint в 1-4 байта — used by interpolation (C-level) and the
// Display protocol default body (`w.write(@to_str())`, protocols.nv).

// Unchecked UTF-8 encode codepoint в 1-4 байта. Caller гарантирует cp ∈ [0, 0x10FFFF]. Invalid codepoint → empty str (silent — predictable degradation). Используется JSON unicode escape parser-ами + другими decoders.
export extern "nova" fn str.from_codepoint(cp int) -> str

// ─── char ───

// D54: int → char, validates unicode scalar value [0, 0x10FFFF] excluding
// surrogates. Ресивер-форма `@to_char()` (решение владельца 2026-07-09,
// цепочечность: `(32 + off).to_char()?`) — бывший `int.to_char()`
// ретрактирован по §22 (одно имя = одна дверь); каст `int as char` запрещён D54.
export fn int @to_char() -> Result[char, CharFromError] {
    if @ < 0 || @ > 0x10FFFF || (@ >= 0xD800 && @ <= 0xDFFF) {
        Err(CharFromError)
    } else {
        Ok(unsafe { @ as char })
    }
}

// ─── u8 ───

// D54/D77: char → u8, fails if codepoint > 0xFF (non-Latin-1).
export fn u8.try_from(c char) -> Result[u8, TryFromCharError] {
    if c as int > 0xFF {
        Err(TryFromCharError)
    } else {
        Ok(c as int as u8)
    }
}