> Le 8 déc. 2014 à 14:39, Simon Slavin <[email protected]> a écrit : > > On 8 Dec 2014, at 1:31pm, Gwendal Roué <[email protected]> wrote: > >> We share the same conclusion. I even tried to decorate the update query with >> "ORDER" clauses, in a foolish attempt to reverse the ordering of row >> updates, and circumvent the issue. > > A way to solve this is to use REAL for page numbers instead of INTEGER. To > insert a page between two existing ones, give it a number which is the mean > of the two pages you're inserting it between. Every so often you can run a > maintenance routine which renumbers all pages to integers. > > Alternatively, store your pages as a linked list.
Polluting my database schema around such a bug is not an option for me, as long as I can find a work around that is good enough and leaves my intent intact. The one I chose involves destroying the unique index before running the failing update query, and then recreating it. All I look for is this issue to enter the ticket list of sqlite at http://www.sqlite.org/src/reportlist, so that this fantastic embeddable database gets better. Gwendal Roué _______________________________________________ sqlite-users mailing list [email protected] http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
