refactor: two names that meant different things in different files

[productdevbook]

refactor: two names that meant different things in different files

Neither is a bug today. Both are the shape a bug arrives in: a line moved
between two files keeps working and quietly changes contract.

parseCellRef is exported from xlsx/worksheet.ts and is
deliberately lenient — a cell reference that will not parse should cost
one cell, not the read. pivot-writer.ts had a private one under the
same name that throws, which is right for what it does (a targetCell
the caller typed, where a typo has no salvageable reading and pivoting at
A1 silently would be worse) and wrong to reach for anywhere else. It is
now parseCellRefStrict, with the reason for the difference on it.

decodeUtf8 means "decode, and report a part over V8's
MAX_STRING_LENGTH as the #514 ParseError naming it" in reader.ts,
ods/reader.ts and xlsb/reader.ts, and "decode" in stream-reader.ts
and ods/stream.ts. The streaming readers not having the ceiling is
intentional and is what docs/PARITY.md records — streamXlsxRows is
the answer to the ceiling, since it SAX-parses the worksheet off the
decompression stream and never builds one string for it. The two spellings
are now decodeUtf8 and decodeUtf8Unchecked, so the difference is
visible at the call site rather than only in PARITY.

Both notes record what is still true of each streaming path: the parts
stream-reader.ts decodes whole are the small ones, so a
sharedStrings.xml over the ceiling would throw V8's raw error; and
streamOdsRows streams rows out of content.xml but still builds the
whole part as a string to do it.

Names only. pnpm test green — 10,583 tests, 234 files.

🤖 Generated with Claude Code