It?s a mistake ofexpression. I mean, 1. readconcurrently indifferent threads. or 2. write in one thread andread inother threads. noconcurrently writing happened.
Sorry about this. I come from a not-English-speaking country, it's hard for me to express myself. ???? ???:Simon Slavinslavins at bigfraud.org ???:General Discussion of SQLite Databasesqlite-users at mailinglists.sqlite.org ????:2015?10?9?(??)?20:38 ??:Re: [sqlite] Is it possible that keep SQLite in normal lockingmodewithout mmap? On 9 Oct 2015, at 1:29pm, sanhua.zh sanhua.zh at foxmail.com wrote: 2. multithread-reading-and-writing, not serial It is not possible for two different operations, whether different threads or different processes, to write to a database at the same time. No matter what system you use this just isn't possible. To prove this to yourself, imagine that you had a table with a UNIQUE index, and that two threads/processes tried to write rows with the same key. The correct way for the API to respond would be to allow one to write its row, and to return an error to the other. This cannot happen if the system accepts both write requests at the same time. Writing to a database /must/ be serial. There is a work-around for this which involve writing new rows by appending the data to a text file (entirely outside SQL) and having another process read its way through the text file updating the database file. Appending to a text file is faster than updating a SQLite database. There are other work-arounds too. But they all involve writing new data to some place which is not the live database. Simon. _______________________________________________ sqlite-users mailing list sqlite-users at mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
