Documentation

The text in UTF-16LE, without a byte order mark. The same bytes as encodeUtf16LE.

The length of the text in UTF-16 code units: its characters, plus one for each outside the Basic Multilingual Plane. (Every UTF-8 byte that isn't a continuation byte starts a character, and a four-byte sequence is a surrogate pair.)

Write UTF-16LE into caller-owned memory without checking capacity. Returns the byte count written: 2 * utf16Length text, with no byte order mark or terminator.

The pointer must remain valid and writable for at least that many bytes throughout the call. Twice the text's UTF-8 byte length is a sufficient upper bound when avoiding a length scan. Size and offset arithmetic must fit in Int; source and destination must not overlap, and the caller must respect the lifetime and immutability of any aliases to the buffer. Use encodeUtf16LE or utf16LEBuilder to let the library provide storage.

encodeUtf16LE as a Builder, written straight into the builder's buffer, a piece at a time for text too long for one of 32 KiB.

Decode UTF-16LE text from the server, never failing: each ill-formed code unit (an unpaired surrogate) becomes one U+FFFD, and so does a trailing odd byte. decodeUtf16LE says why.

decodeUtf16LE for a value that arrives in chunks: the result is the same however the value is cut. SQL Server ends PLP chunks at packet ends, which can fall inside a code unit or a surrogate pair. Only the pending character, at most three bytes, is copied between chunks.