SQL primary key: integer vs varchar

The primary key is supposed to represent the identity for the row and should not change over time.

I assume that the varchar is some sort of natural key - such as the name of the entity, an email address, or a serial number. If you use a natural key then it can sometimes happen that the key needs to change because for example:

• The data was incorrectly entered and needs to be fixed.
• The user changes their name or email address.
• The management suddenly decide that all customer reference numbers must be changed to another format for reasons that seem completely illogical to you, but they insist on making the change even after you explain the problems it will cause you.
• Maybe even a country or state decides to change the spelling of its name - very unlikely, but not impossible.

By using a surrogate key you avoid problems caused by having to change primary keys.

i was a bit disappointed because i've the habit to create an integer primary key (following what some teacher told me at the university). I've read a lot of documentation on the performance boost using integer primary key.

There is a term for this: confirmation bias:

"also called confirmatory bias or myside bias) is a tendency for people to favor information that confirms their preconceptions or hypotheses, independently of whether they are true. This results in people selectively collecting new evidence, interpreting evidence in a biased way, or selectively recalling information from memory."

Of course, your first reaction will be to say, "But that's not true!" Yeah, you would say that 'cos you're biased ;) [tongue firmly embedded in cheek]

Here's a classic example: say you had been told by your zoology professor that all swans are white and, sure enough, all swans you and your friends have ever encountered are white. Now let's say later in life a colleague expressed the opinion that perhaps there is such creature as a black swan. What?! That's not what your were taught. Your world is rocked! You immediately go out and conduct a swan survey and you count 1,000 white swans and zero black swans. Proof! If you'd found 10,000 white swans then the hypothesis 'All swans are white' would be ten times truer, right?

A different approach would be to forget about white swans for the moment and try to seek out a black swan. Perhaps take a holiday by the sea in sunny Dawlish?

I really don't mean to sound disrespectful; you admit to reading a lot about what you have been told and that indeed earns my respect. So here's a challenge: try to find cases where adding an integer column to a table is unnecessary.

Here are some hints and spoiler