Giorgi / DuckDB.NET / 26975611933

89%

develop: 89%

push

github

Fix IndexOutOfRangeException in reader metadata after exhausting result. Fixes #332

DataTable.Load (and similar consumers) read all rows and then call
NextResult() to probe for more result sets. InitNextReader() disposed the
vector readers and set vectorReaders=[] at the top of the method, before
checking whether another result set existed. When none did, it returned
false with vectorReaders emptied but fieldCount still holding the previous
result's column count. A subsequent GetSchemaTable()/GetName()/GetFieldType()
then indexed past the emptied array and threw IndexOutOfRangeException.

This regressed in 1.5.0 with the cross-chunk reader-reuse refactor; in 1.4.4
the readers were only replaced when a new result set was found, so an
exhausted reader kept its column metadata.

Move the dispose + clear inside the QueryResult branch so they run only when
switching to a new result set. An exhausted reader now retains its column
metadata, restoring 1.4.4 behavior.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

1327 of 1541 branches covered (86.11%)

Branch coverage included in aggregate %.

3 of 3 new or added lines in 1 file covered. (100.0%)

2770 of 3038 relevant lines covered (91.18%)

434723.3 hits per line