At 16:55 +0100 3/26/02, Roger Baklund wrote:
>* Paul DuBois
>> At 15:23 +0100 3/26/02, Tino Didriksen wrote:
>> >Question about the CHAR to VARCHAR conversion of:
>> >Now, I can see the logic in changing most CHAR to VARCHAR if
>> they are long
>> >and variable sized.
>> >I have a column that is VARCHAR(32). I use it to store MD5 hash
>> values. MD5
>> >hashes are always 32 byte.
>> >In other words, that's 1 byte wasted because it's VARCHAR.
>> >With 2500 entries and more on the way, that's at least 2500 byte wasted.
>> >Is there no way to enforce a column type around the silent changes?
>> No. It makes no sense to do so.
>well... even if 2500 bytes is insignificant, I think his question makes
>The only way I know to enforce CHAR is to make sure all other columns in the
>table are of fixed size types... ie not *TEXT, *BLOB or VARCHAR.
That's why it doesn't make sense to have just a single column be CHAR.
As soon as a table has even one other variable-length column, the entire
table becomes a variable-length-row table and any advantages of having
fixed-length rows are lost. Even in the case that a column might always
have 32 bytes, the single-byte advantage that you might get from using a
CHAR are presumably offset by savings in the other variable-length columns.
If the latter is not true, then those columns should be changed to fix-length