sqlite3.Row is ~1.09× slower than tuple (default)

Setting `row_factory=Row` makes fetching 20,000 rows ~9% slower — the cost of name-based access.

Result

Spread over 3 runs: sqlite3.Row CV 1.6% / tuple (default) CV 1.6% / Ratio CV 2.4%

The statement

con.row_factory = sqlite3.Row
rows = con.execute('SELECT id,name,score FROM items LIMIT 20000').fetchall()

rows = con.execute('SELECT id,name,score FROM items LIMIT 20000').fetchall()

Conditions

20,000 rows fetched with `fetchall()`, three columns.

Where this number does not apply

With few rows the difference is lost in noise.

Reproduce

cd bench && python sqbench.py --samples 9 --out ../data/run.json