v0.0.24

0.0.24

Features

• Everyday filesystem work stays static across more of the file lifecycle. fs.rename, fs.renameSync, and fs/promises.rename lower with Node's callback, promise, replacement, and platform error behavior; fs.writeSync covers the Buffer-window and UTF-8 string overloads with current-offset and positioned writes; and fs/promises.open returns a native FileHandle with fd, read, write, readFile, writeFile, appendFile, stat, and idempotent close. The three-argument fs/promises.writeFile accepts the supported UTF-8 and creation-mode options, while filesystem stats expose blocks, nlink, and atimeMs through path, promise, file-handle, and dynamic-bridge results.
• Native text decoding covers the WHATWG encoding families. A statically labelled TextDecoder recognizes the standard aliases for the legacy single-byte, UTF-16, Chinese, Japanese, and Korean encodings, including BOM handling and malformed-input recovery matching the pinned Node runtime. Runtime-valued BufferEncoding arguments also work in Buffer.toString, with case-insensitive aliases, optional defaults, range behavior, and Node's unknown-encoding errors.
• util.parseArgs compiles statically. Boolean and string options, short and clustered spellings, repeated values, defaults, negative options, positionals, token output, permissive parsing, the live default process.argv, and Node's coded validation errors lower through both backends.
• Array mutation and bounded string splitting grow their static surface. Array.prototype.unshift and reverse preserve mutation, spread, evaluation order, and reference identity, while String.prototype.split accepts JavaScript limits for string and literal-RegExp separators with ToUint32 semantics.
• process.stdout.write and process.stderr.write accept encoding and completion arguments. String chunks honor supported encoding aliases, byte chunks retain their bytes, and successful completion callbacks join the next-tick queue in Node's order on both backends.

Performance

• Canonical typed-array loops use native integer induction. Both backends recognize semantics-safe byte loops and emit integer counters and direct byte indexes, avoiding repeated floating-point index conversion while preserving the general fallback whenever the loop shape cannot prove the rewrite.