Is it better to use default null or default "" for text fields in MySQL?
Why?
Update: I know what means each of them. I am interested what is better to use considering disk space and performance.
Update 2: Hey ppl! The question was "what is better to use" not "what each means" or "how to check them"...
Use default null. In SQL, null is very different from the empty string (""). The empty string specifically means that the value was set to be empty; null means that the value was not set, or was set to null. Different meanings, you see.
The different meanings and their different usages are why it's important to use each of them as appropriate; the amount of space potentially saved by using default null as opposed to default "" is so small that it approaches negligibility; however, the potential value of using the proper defaults as convention dictates is quite high.
Use whatever makes sense. NULL means "no value available/specified", "" means "empty string."
If you don't allow empty strings, but the user does not have to enter a value, then NULL makes sense. If you require a value, but it can be empty, NOT NULL and a value of "" makes sense.
And, of course, if you don't require a value, but an empty value can be specified, then NULL makes sense.
Looking at an efficiency point of view, an extra bit is used to determine whether the field is NULL or not, but don't bother about such micro-optimization until you have millions of rows.
"" is like an empty box... null is like no box at all.
It's a difficult concept to grasp initially, but as the answers here plainly state - there is a big difference.
In general, NULL should indicate data that is not present or has not been supplied, and therefore is a better default value than the empty string.
Sometimes the empty string is what you need as a data value, but it should almost never be a default value.
NULL means 'there is no value' and is treated especially by RDBMSs regarding where clauses and joins.
"" means 'empty string' and is not treated especially.
It depends on what does the text represent and how will it actually be used in queries.
For example, you can have a questionnaire with some obligatory questions and some optional questions.
'' = '' yields TRUE which satisfies WHERE condition
NULL = NULL yields NULL which doesn't satisfy WHERE condition
Which is better to use depends on what result you want to get.
If your values default to NULL, no query like this:
SELECT *
FROM mytable
WHERE col1 = ?
will ever return these values, even if you pass the NULL for the bound parameter, while this query:
SELECT *
FROM mytable
WHERE col1 = ''
will return you the rows that you set to an empty string.
This is true for MySQL, but not for Oracle, which does not distinguish between empty string and a NULL.
In Oracle, the latter query will never return anything.
A lot of folks are answering the what is the difference between null and '', but the OP has requested what takes up less space/is faster, so here's my stab at it:
The answer is that it depends. If your field is a char(10), it will always take 10 bytes if not set to null, and therefore, null will take up less space. Minute on a row-by-row basis, but over millions and millions of rows, this could add up. I believe even a varchar(10) will store one byte (\0) as an empty string, so again this could add up over huge tables.
In terms of performance in queries, null is in theory quicker to test, but I haven't seen able to come up with any appreciable difference on a well indexed table. Keep in mind though, that you may have to convert null to '' on the application side if this is the desired return. Again, row-by-row, the difference is minute, but it could potentially add up.
All in all it's a micro-optimization, so it boils down to preference. My preference is to use null because I like to know that there's no value there, and not guess if it's a blank string ('') or a bunch of spaces (' '). null is explicit in its nature. '' is not. Therefore, I go with null because I'm an explicit kind of guy.
For MyISAM tables, NULL creates an extra bit for each NULLABLE column (the null bit) for each row. If the column is not NULLABLE, the extra bit of information is never needed. However, that is padded out to 8 bit bytes so you always gain 1 + mod 8 bytes for the count of NULLABLE columns. [1]
Text columns are a little different from other datatypes. First, for "" the table entry holds the two byte length of the string followed by the bytes of the string and is a variant length structure. In the case of NULL, there's no need for the length information but it's included anyways as part of the column structure.
In InnoDB, NULLS take no space: They simply don't exist in the data set. The same is true for the empty string as the data offsets don't exist either. The only difference is that the NULLs will have the NULL bit set while the empty strings won't. [2]
When the data is actually laid out on disk, NULL and '' take up EXACTLY THE SAME SPACE in both data types. However, when the value is searched, checking for NULL is slightly faster then checking for '' as you don't have to consider the data length in your calculations: you only check the null bit.
As a result of the NULL and '' space differences, NULL and '' have NO SIZE IMPACT unless the column is specified to be NULLable or not. If the column is NOT NULL, only in MyISAM tables will you see any peformance difference (and then, obviously, default NULL can't be used so it's a moot question).
The real question then boils down to the application interpretation of "no value set here" columns. If the "" is a valid value meaning "the user entered nothing here" or somesuch, then default NULL is preferable as you want to distinguish between NULL and "" when a record is entered that has no data in it.
Generally though, default is really only useful for refactoring a database, when new values need to come into effect on old data. In that case, again, the choice depends upon how the application data is interpreted. For some old data, NULL is perfectly appropriate and the best fit (the column didn't exist before so it has NULL value now!). For others, "" is more appropriate (often when the queries use SELECT * and NULL causes crash problems).
In ULTRA-GENERAL TERMS (and from a philosophical standpoint) default NULL for NULLABLE columns is preferred as it gives the best semantic interpretation of "No Value Specified".
I prefer null when it is semantically correct. If there is an address field available and the user did not fill in, I give it a "". However if there in an address attribute to in the users table yet I did not offer the user a chance to fill it in, I give it a NULL.
I doubt (but I can't verify) that NULL and "" makes much of a difference.