"The abs(X) function returns the absolute value of the numeric argument X. Abs(X) returns NULL if X is NULL. Abs(X) return 0.0 if X is a string or blob that cannot be converted to a numeric value. If X is the integer -9223372036854775807 then abs(X) throws an integer overflow error since there is no equivalent positive 64-bit two complement value."
* The int should be -9223372036854775808. * Also, "return 0.0" should be "returns 0.0". * Also, "two complement" should be "two's complement". Minor typos aside, it's interesting that SQLite is nice enough to convert integers strictly less than -1 * 2^63 to floats and then take the floating absolute value. This means that the only gap in the function's domain is at the sole integer value -1 * 2^63. One greater and SQLite gives you a value; one less and SQLite gives you a value: sqlite> select abs(-9223372036854775807); 9223372036854775807 sqlite> select abs(-9223372036854775808); Error: integer overflow sqlite> select abs(-9223372036854775809); 9.22337203685478e+18 Given that, and given SQLite's general philosophy of being flexible about its inputs, would it make sense for SQLite to convert -1 * 2^63 to a float and return the floating absolute value, as it does for (-1 * 2^63) - 1? I can also see the arguments against that choice: users who were careful to store a 64-bit integer into a field may be surprised when it gets converted to a float when passed through abs(). You also have backward compatibility considerations. Thanks as always to drh and team for the awesome tool. Eric _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
