@@ -0,0 +1,113 @@ |
| 1 | +"""UTF-8 text I/O with BOM and CRLF hygiene (audit F15). |
| 2 | + |
| 3 | +Every sprint that reads or writes a `.dlm` file or any plain-text store |
| 4 | +artifact must route through these helpers. The contract: |
| 5 | + |
| 6 | +- UTF-8 strict: invalid bytes raise `DlmEncodingError` with the offending |
| 7 | + byte offset. |
| 8 | +- A leading UTF-8 BOM is stripped and a warning is emitted (some Windows |
| 9 | + editors add it; YAML parsers choke on it). |
| 10 | +- CRLF is normalized to LF before hashing or parsing, so section IDs are |
| 11 | + stable across Windows and Unix edits of the same content. |
| 12 | + |
| 13 | +Binary I/O uses the normal `open(..., "rb")` path; this module is only for |
| 14 | +text. |
| 15 | +""" |
| 16 | + |
| 17 | +from __future__ import annotations |
| 18 | + |
| 19 | +import logging |
| 20 | +from pathlib import Path |
| 21 | +from typing import Final |
| 22 | + |
| 23 | +_LOG: Final = logging.getLogger(__name__) |
| 24 | + |
| 25 | +_UTF8_BOM: Final = "\ufeff" |
| 26 | +_UTF8_BOM_BYTES: Final = b"\xef\xbb\xbf" |
| 27 | + |
| 28 | + |
| 29 | +class DlmEncodingError(ValueError): |
| 30 | + """Raised when a text file is not valid UTF-8. |
| 31 | + |
| 32 | + `byte_offset` is the zero-based position of the first offending byte, |
| 33 | + as reported by the codec. |
| 34 | + """ |
| 35 | + |
| 36 | + def __init__(self, path: Path | None, byte_offset: int, reason: str) -> None: |
| 37 | + self.path = path |
| 38 | + self.byte_offset = byte_offset |
| 39 | + self.reason = reason |
| 40 | + where = str(path) if path is not None else "<text>" |
| 41 | + super().__init__(f"{where}: invalid UTF-8 at byte {byte_offset}: {reason}") |
| 42 | + |
| 43 | + |
| 44 | +def read_text(path: Path) -> str: |
| 45 | + """Read `path` as UTF-8; strip BOM; normalize CRLF → LF. |
| 46 | + |
| 47 | + Raises `DlmEncodingError` on invalid UTF-8. Emits a warning via the |
| 48 | + `dlm.io.text` logger when a BOM is present so pipelines surface the |
| 49 | + (usually Windows-editor) provenance. |
| 50 | + """ |
| 51 | + raw = path.read_bytes() |
| 52 | + return _decode(raw, path=path) |
| 53 | + |
| 54 | + |
| 55 | +def read_text_str(raw: bytes, *, source: str = "<bytes>") -> str: |
| 56 | + """In-memory variant of `read_text` for tests and streamed input.""" |
| 57 | + # Decode with path=None; `source` is only for the error message. |
| 58 | + try: |
| 59 | + return _decode(raw, path=None) |
| 60 | + except DlmEncodingError as exc: |
| 61 | + # Rewrite to include source in the message. |
| 62 | + raise DlmEncodingError(None, exc.byte_offset, f"{source}: {exc.reason}") from exc |
| 63 | + |
| 64 | + |
| 65 | +def write_text(path: Path, content: str) -> None: |
| 66 | + """Write `content` as UTF-8 with LF line endings, no BOM. |
| 67 | + |
| 68 | + Writes atomically: writes to a temp sibling file then `os.replace`s. |
| 69 | + """ |
| 70 | + # Normalize line endings on the way out too, belt-and-braces. |
| 71 | + normalized = content.replace("\r\n", "\n").replace("\r", "\n") |
| 72 | + tmp = path.with_suffix(path.suffix + f".tmp.{_pid()}") |
| 73 | + tmp.write_bytes(normalized.encode("utf-8")) |
| 74 | + tmp.replace(path) |
| 75 | + |
| 76 | + |
| 77 | +def normalize_for_hashing(content: str) -> str: |
| 78 | + """Produce the canonical form used when hashing text content. |
| 79 | + |
| 80 | + - BOM removed (if any) |
| 81 | + - CRLF → LF |
| 82 | + - CR alone → LF |
| 83 | + |
| 84 | + This must exactly mirror what `_decode` produces so that |
| 85 | + `hash(read_text(path)) == hash(normalize_for_hashing(serialize(...)))`. |
| 86 | + """ |
| 87 | + if content.startswith(_UTF8_BOM): |
| 88 | + content = content[1:] |
| 89 | + return content.replace("\r\n", "\n").replace("\r", "\n") |
| 90 | + |
| 91 | + |
| 92 | +# --- internals --- |
| 93 | + |
| 94 | + |
| 95 | +def _decode(raw: bytes, *, path: Path | None) -> str: |
| 96 | + try: |
| 97 | + text = raw.decode("utf-8") |
| 98 | + except UnicodeDecodeError as exc: |
| 99 | + raise DlmEncodingError(path, exc.start, exc.reason) from exc |
| 100 | + |
| 101 | + if text.startswith(_UTF8_BOM): |
| 102 | + where = str(path) if path is not None else "<text>" |
| 103 | + _LOG.warning("%s: UTF-8 BOM present; stripped", where) |
| 104 | + text = text[1:] |
| 105 | + |
| 106 | + # CRLF first, then stray CR, to avoid double-replacing \r\n as \r then \n. |
| 107 | + return text.replace("\r\n", "\n").replace("\r", "\n") |
| 108 | + |
| 109 | + |
| 110 | +def _pid() -> int: |
| 111 | + import os as _os |
| 112 | + |
| 113 | + return _os.getpid() |
