On 8/12/09 5:11 AM, Martin Hansson wrote:
> Approved for me, but perhaps you could educate me, see below:
>>>
>>>>
>>>> +uint Item_func_get_user_var::decimal_precision() const
>>>> +{
>>>> + uint precision= max_length;
>>>> + Item_result restype= result_type();
>>>> +
>>>> + /* Default to maximum as the precision is unknown a priori. */
>>>> + if ((restype == DECIMAL_RESULT) || (restype == INT_RESULT))
>>>> + precision= DECIMAL_MAX_PRECISION;
>>>> +
>>>> + return precision;
>>>> +}
>>>> +
>>>> +
>>> What made you add this?
>>>
>>
>> Item::decimal_precision calculates the precision from the maximum
>> length, yet the maximum length in the case of a decimal type user
>> variable is the maximum string length of a decimal (see method
>> fix_length_and_dec, class Item_func_get_user_var). Hence, it is not
>> correct (for the decimal type user variable case) to calculate the
>> precision from the length. Furthermore, the precision and scale for
>> user variables are set to the maximum in any case.
> Alright, so max_length (the display length) is 83 after
> fix_length_and_dec(), and 'decimals' is the default 30.
> Item::decimal_precision() returns 81 while
> Item_func_get_user_var::decimal_precision() returns 65 . But the actual
> field created in the below test is DECIMAL(65, 14) without the above fix.
>
> SET @decimal= 1.1;
> CREATE TABLE t1 SELECT @decimal AS c1;
> DESC t1;
>
> Incidentally, the difference is the same: 81 - 65 = 14 = 30 - 16. Where
> is this adjustment done?
>
Old Item::decimal_precision:
uint Item::decimal_precision() const
{
Item_result restype= result_type();
if ((restype == DECIMAL_RESULT) || (restype == INT_RESULT))
{
uint prec=
my_decimal_length_to_precision(max_length, decimals, unsigned_flag);
return min(prec, DECIMAL_MAX_PRECISION);
}
return min(max_length, DECIMAL_MAX_PRECISION);
}
The precision would be capped to DECIMAL_MAX_PRECISION.
