Re: [sqlite] Transaction behavior with large numbers of inserts

2009-01-28 Thread vlemaire
Hi, Does your reader threads access to the same table as your writer ? Maybe (I'm not 100% sure it works) you could try : on a single file database, with the PRAGMA read_uncommited=1 on the readers connections (this is not a good idea if the tables are the same between readers and writers, it

Re: [sqlite] Compressed dump SQLite3 database

2009-01-20 Thread vlemaire
Hello, thank you and others for your answers. We are not allowed to use external libraries, because of industrial certification constraints. We can use SQLite because we can not do without database, but it's a big stuff to validate it according to those constraints, so we have to reduce as much

[sqlite] Number precision

2009-01-12 Thread vlemaire
Hello, As described in SQL syntax (http://www.sqlite.org/syntaxdiagrams.html#type-name) a column type may be declared like this : int(x) or float(x, y) ... what does x and y means ? max number of digits ? max value ? max bytes encoding ? ... a subsidiary question is : how to specify a precision

Re: [sqlite] Deadlock on SQLite

2008-11-07 Thread vlemaire
Hi, I've read that there is a compile option to unable connexion being thread-safe : http://www.sqlite.org/compile.html#threadsafe I guess also there could be problems if you have a not commited transaction pending while you are using another for write access, and/or a previous statement not

[sqlite] Minimum write size on disk

2008-11-07 Thread vlemaire
Hello, The atomic write size on the system file hosting my database is limited by design to 32Kbyte. I want to verify that SQLite will never write more than this amount. When I say "atomic write" it concerns calls to the pwrite function of standard library, with buffer size parameter set to up

[sqlite] row count after a select

2008-11-05 Thread vlemaire
Hello, Do you know a way to count the number of rows identified by a select statement, just after the first call to sqlite3_step ? Of course without fetching the entire result set, nor having to systematically add count(*) Thanks ! Vincent ___