-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Eberhard, Markus (external) wrote: > In other > words: minimize the number of accesses to database file (but I'm not > sure if this is possible).
You could use the SQLite backup API to read the entire file into :memory: (quickly and sequentially) and then issue the query against the memory database. That will obviously only make sense if a substantial portion of the database is returned by the select query. You can also use the operating system file api to read the entire file (discarding what you read) which will warm the operating system file cache meaning less disk accesses during the actual query. (This assumes you aren't doing something like using XP with its default 10MB file cache.) Finally you could access the blobs later on demand. Rather than having the select return the blobs, just get the rowids. Later on you can use the incremental blob I/O api to get the blob contents. This approach is useful if you don't need them all at once (eg you have a user interface with a scroll bar) and that they are non-trivial in size (ie at least several database pages). Roger -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.9 (GNU/Linux) iEYEARECAAYFAkpW9EEACgkQmOOfHg372QR60wCghtye362kXB6ByDfpj2I2rm88 nckAn0LiFPAUkRe7BNN3My92+MNZo0aE =z2GP -----END PGP SIGNATURE----- _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users