What is easier to read in EXISTS subqueries? [closed]

问题

It\'s a question of readability. There is no difference in performance.
_{Old versions of SQL Server were silly enough to look up meta data, but not any more.}

SELECT foo FROM bar WHERE EXISTS (SELECT * FROM baz WHERE baz.id = bar.id);
SELECT foo FROM bar WHERE EXISTS (SELECT 1 FROM baz WHERE baz.id = bar.id);

I am not considering NULL or \"fun variants\" which don\'t seem intuitive to me.

SELECT foo FROM bar WHERE EXISTS (SELECT NULL FROM baz WHERE baz.id = bar.id); SELECT foo FROM bar WHERE EXISTS (SELECT 1/0 FROM baz WHERE baz.id = bar.id);

The question popped up in comments just now. I researched the manuals of the most popular RDBMS:

• MS SQL seems to favor SELECT * in the manual.
• The example in the PostgreSQL 9.4 manual uses SELECT 1.
• Oracle 11g has SELECT * in the language reference.
• MySQL 5.7 has SELECT * in the reference manual but alsoSELECT 1 in the comments.
• SQLite has no example in the language reference.

A search on SO for code:\"EXISTS (SELECT 1\" yields 5,048 results.
A search on SO for code:\"EXISTS (SELECT *\" yields 5,154 results.
_{Updated links and counts 07.2015.}

So SELECT * has the popular vote and the big commercial RDBMS on its side.
I find SELECT 1 more intuitive. It\'s like saying \"if at least one exists\".
Is SELECT * more intuitive?

回答1:

Intuitive is ...EXISTS (SELECT * .. because you really don't care

• The only keyword of importance is EXISTS
• The choice of ...EXISTS (SELECT 1 .. perpetuates the general myths and superstitions around EXISTS (eg comments on the MySQL docs).
• ANSI standard says "doesn't matter"
• It's more interesting to understand that EXISTS is a semi-join.

回答2:

I still use EXISTS (SELECT * ...), for historical (gbn: should that be hysterical?) reasons. Technically, there is no difference, of course; the optimiser / planner will throw it away and reduce it to one bit of information. For the human reader the * looks more special, it will stand out as a special symbol, and not as a value or constant. Also, I tend to reduce the amount of literals and magic constants in my programs (eventually, only 0 and 1 should remain).

回答3:

In the context of EXISTS the SQL optimizer knows that it doesn't matter what it returns as long as it returns something. So to you it doesn't matter.

For the intuitive part: I don't think * will be right.

It's better to ask in words: "check whether even the slightest part exists" - meaning 1 (or something else).

来源：https://stackoverflow.com/questions/7710153/what-is-easier-to-read-in-exists-subqueries