Giorgi / DuckDB.NET / 38363493621

93%

push

github

Encode a string in the writer and write a short one into the vector

A string was passed to duckdb_vector_assign_string_element, so the
marshaller of the call encoded it to UTF-8 with a terminator, and DuckDB
measured it again and checked the UTF-8.

The writer now encodes it. A string of up to 12 bytes is written straight
into the vector, in the layout of DuckDB's string_t, with no native call.
A longer one is encoded on the stack, or into a rented array above 170
characters, and handed to duckdb_unsafe_vector_assign_string_element_len
with its length. What the encoder writes is always valid UTF-8, so
DuckDB's check is skipped.

A string still ends at its first NUL character, as it did when DuckDB
measured it as a C string.

Checked: 27,016 strings appended to a VARCHAR, a VARCHAR[] and a BLOB
column and returned from a scalar function are stored byte for byte as by
1.5.6-alpha.74, on net8.0 and net10.0. Not yet benchmarked.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TXdqHuNtM7YXS8B65qdfxX

1936 of 2141 branches covered (90.43%)

Branch coverage included in aggregate %.

22 of 22 new or added lines in 1 file covered. (100.0%)

3632 of 3866 relevant lines covered (93.95%)

1036451.05 hits per line