Hi, thanks for the clarification. I may need to add a custom implementation of the VFS or a patch for the -shm files to be held on memory-disk then.
Regarding the question, why I want to wrap reads in a transaction is, because I don't know if there'll be a write as part of the transaction in before. That partly depends on results returned from the query. Thanks for your answers! Best regards, Martin On Thu, Sep 16, 2010 at 9:14 PM, Richard Hipp <d...@sqlite.org> wrote: > Reads to not write to the WAL. But they might write to the shared-memory. > > On Thu, Sep 16, 2010 at 1:16 PM, Martin C. <mart...@gmx.at> wrote: > >> Hi, >> >> I am using SQLite on a NAND based embedded system, and to ensure the >> lifetime of the NAND I need to optimize writes to the NAND. >> >> Can you tell me, if a read-only transaction, by which I mean >> BEGIN TRANSACTION >> SELECT * from SomeTable; >> COMMIT >> will cause any writes to the WAL? >> >> What about the -shm file, is it likely that the transaction will cause >> a write to the NAND in this case? >> >> Is there any statistics about actual writes I can query to analyze >> write-behavior of my software? >> >> Thanks in advance! >> >> Best regards, >> Martin >> _______________________________________________ >> sqlite-users mailing list >> sqlite-users@sqlite.org >> http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users >> > > > > -- > D. Richard Hipp > d...@sqlite.org > _______________________________________________ > sqlite-users mailing list > sqlite-users@sqlite.org > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users > _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users