Dtyjlui

I often come across columns that are varchar(255), so presumably this is a convention of sorts. 255 is 2^8 - 1. So is there something 'magical' about lengths that are 2^n - 1? Or just specifically 255? Is there a performance or storage optimisation advantage to certain specific varchar lengths? Or is this just old wisdom which is no longer applicable to current versions of MariaDB and MySQL?

I rant against 255 occasionally. Sure, there used to be some reasons for '255', but many are no longer valid, and even counter-productive.

In MySQL, there are reasons to stop at 191, 255, 767, 64K, and probably other values. Some depend on Engine, some on CHARACTER SET, etc.

A VARCHAR is stored as a 1- or 2-byte length plus enough bytes for the current text in whatever charset you have specified. However, the choice of 1 or 2 is not driven only by the individual column; it is driven by the total row size. That is, this is not a valid excuse for using 255.

Use a length that is

• Big enough to conservatively never be exceeded, yet


• As small as seems reasonable.

While I am ranting... CHAR (fixed length) is rarely advised. And almost always it should be CHARACTER SET ascii -- country_code, postal_code, Y/N, M/F, MD5, UUID, base64, etc. (MD5 and UUID should be taken a step further, but that is another rant.)

Potential negative impacts of blindly using '255':

• If you have a lot of columns, you could hit a max row size limit and CREATE TABLE will fail.


• You could overflow a limit on index size.


• Complex SELECTs may need a temp table, and may use MEMORY for it. In this case, VARCHAR(255) becomes CHAR(255) for the temp table. And, if using utf8, that is 755 bytes for every row! (8.0 fixes this design flaw?)

Related: Don't blindly use BIGINT for all numbers. It takes 8 bytes. Even INT is overkill, at 4 bytes. See MEDIUMINT, SMALLINT, and TINYINT. Be aware that the (2) on INT(2) means nothing; it still takes 4 bytes.

At one point in time, that was the max length of a column in Sybase, and the derived SQL Server. I'm sure others copied it to maintain compatibility so that migrating applications / SQL to a new platform wasn't difficult, as everyone was fighting for market share back in the day.

There's no real good reason to perpetuate it, other than compatibility. I've noticed that it seems to have been replaced by varchar(max) these days.

"Standard" column lengths are used all over the place, and invariably indicate there was no effort performed to understand the actual required length of the column.

If you have a column where there will never be more than 24 characters used, why specify the length of the column as 255?

Analyzing and using the correct length for variable-length columns takes a small amount of effort, and ensures side effects of using too large a column length never become a problem. I know your question is about MySQL, however this answer shows how SQL Server uses column lengths in various ways that may not be apparent at first.

We've taken a completely different rule now. We use one of two things:

`NVARCHAR(max)`

if it won't be indexed.

`NVARCHAR(400)`

if it will be indexed.

Generally speaking we believe that limits are obsolete and we really shouldn't care anymore. Disk space is cheap. The 400 ensures the index can always be created on the most restrictive target platform. It is my understanding that the 255 index limit on MySQL is gone else I would have written 255. The thing about VARCHAR is you will not be penalized for allowing it to be bigger than it needs to be so there isn't much downside to being too big, so we just set it big and don't think about it anymore.