Peter van Dijk wrote:
every time you open an sqlite database file, the sqlite library has to
parse all table structures. It is much better to keep your
connection/handle open for longer periods of time.
On my XP box it takes about 220us to connect to an SQLite database from
Python, whether there is 1 table or 1000.
A "connect" here is "C=sqlite3.Connection('tmp.db'); C.close()"
0.121: To create 1 tables with 1000 rows each
0.223: To connect 1000 times (222us per connect)
0.651: To create 10 tables with 1000 rows each
0.236: To connect 1000 times (234us per connect)
5.776: To create 100 tables with 1000 rows each
0.224: To connect 1000 times (223us per connect)
58.393: To create 1000 tables with 1000 rows each
0.219: To connect 1000 times (218us per connect)
I guess the parsing is delayed until necessary because what takes the
time is, say, the first select:
A "connect" here is "C=sqlite3.Connection('tmp.db'); C.execute('select a
from t0 limit 1'); C.close()"
0.119: To create 1 tables with 1000 rows each
1.008: To connect 1000 times (1007us per connect)
0.638: To create 10 tables with 1000 rows each
1.264: To connect 1000 times (1263us per connect)
5.801: To create 100 tables with 1000 rows each
4.193: To connect 1000 times (4192us per connect)
58.419: To create 1000 tables with 1000 rows each
32.468: To connect 1000 times (32466us per connect)
So the difference in connect times between a database with 1 table and
10 tables is about 25%, fairly trivial 250us. With 100 tables connect
time is 3ms, worse by a factor of 4 but still not bad. With 1000 tables
the 33ms connect time becomes significant, but I would imagine that 1000
table databases must be pretty unusual.
Martin
-----------------------------------------------------------------------------
To unsubscribe, send email to [EMAIL PROTECTED]
-----------------------------------------------------------------------------
