> > I don't think that explanation holds water. > I don't have proof that I am correct, but you can try to test my hypothesis the following way:
Let there be 4 databases: -- f1.db create table t(a real unique); insert into t values(9223372036854775807.0); -- f2.db create table t(a real unique); insert into t values(9223372036854775806.0); -- i1.db create table t(a real unique); insert into t values(9223372036854775807); -- i2.db create table t(a real unique); insert into t values(9223372036854775806); Hash all 4 of them: -- sha256sum *.db a50f856aaaacefb0c6cedb3ca66a6539685a2463b47e9a4dc53c174d0a14b6f5 f1.db a50f856aaaacefb0c6cedb3ca66a6539685a2463b47e9a4dc53c174d0a14b6f5 f2.db c12ffcb265433dc61ed848962c4f5a05e215bba81ee6ad6db17b71f22cdbb463 i1.db b2f58d7f372023145537978a0d13296d0bc47b1d75694feabf377908d0a5a7a8 i2.db The ones which insert an integral-looking literal are storing something different from what the ones that insert a fractional-looking literal are. Also see that i1.db and i2.db are different while f1.db and f2.db are equal. Then, one might assume that internally the comparisons executed to ensure uniqueness will behave differently too. _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
