> Then I tried in a loop with 2 programs to write / read in parallel and it > seems to work without problems. Can anyone advise if this has any chance to > work (or say it would definitely NOT work)? > > As a short summary: would it be interesting for anyone to enable read-only > open with a special pragma to allow reading without locking (that means, > shared locks being a noop)? > >
Although speaking generally such method could be used in some situations, I don't think it's good to allow to use it even with a "i know what I'm doing" pragma. Any structured file (sqlite is an example) have internal dependencies. One of the reasons to block is to write different parts of structured data together without intervention from other parties in order to keep the data integral. Imagine writing cache that kept changes for your writer and finally it needed to flush the data and at the same time your "anytime" reader started to perform some query in the middle of this multiply pages writing process. I can't predict whether the reader will end with some valid result or it will return with "database malformed" error. Instead consider changing your own logic. You wrote "without* any possibility to be blocked". I suppose you already have a perfect writer that fits your needs, but if you post some info about the nature of your writer and reader (records per second and something like this), it would help to be more specific with answers. Max _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users