Incorrect representation of NUMERIC fields

[mlazdans]

CREATE TABLE TESTN18
(
    ID INTEGER,
    NUMERIC_18 NUMERIC(18,18)
);
INSERT INTO TESTN18 (ID, NUMERIC_18) VALUES (100, 9.223372036854775807);
SELECT NUMERIC_18 FROM TESTN18 WHERE ID = 100;

Expected:

           NUMERIC_18
=====================
 9.223372036854775807

Actual:

Looks like internal conversion to float / double somewhere.

[mlazdans]

When you edit a row and paste 9.223372036854775807 into NUMERIC_18 this will throw an error:

Notice the 9.223372036854776468 in error.

[arvanus]

I can't see this in the next weeks, but asap I'll take a look By the way, it's strange that the simple update raised a error, can you copy the update and try in isql too?

[mlazdans]

9.223372036854776468 via isql also raises an error:

Statement failed, SQLSTATE = 22003 arithmetic exception, numeric overflow, or string truncation -numeric value is out of range

which is expected because 9.223372036854775807 is maximum that will fit in numeric(18,18) which is signed int64: eg max 2^63-1 and min is -(2^63)

Not sure but it feels like there might be some float/double conversation somewhere in the middle. If so this is bad because one might lose precision even not at the edge cases like this one.