@ and:= 运算符在 SQL

Question

I am working on a table where I have to find average time difference between sessions.

I came across 2 posts where they gave the following solutions. Get average time between times in SQL

    SELECT  AVG(period)
FROM    (
        SELECT  TIME_TO_SEC(TIMEDIFF(@date_posted, date_posted)) AS period,
                @date_posted := date_posted
        FROM    (
                SELECT @date_posted := NULL
                ) vars,
                messages
        ORDER BY
                date_posted
        ) q

and

How to calculate the smallest period of time between consecutive events?

SELECT  (temperature - @r) AS diff,
        @r := temperature
FROM    (
        SELECT  @r := 0
        ) vars,
        temperatures
ORDER BY
        time

I am finding it difficult to understand the query structure to modify it to fit my purposes. Can someone please break it down into simpler terms. With basic googling, I think @var is some variable where you insert values afterwards. I have no idea about:= operator.

I thank you for your time.

Answer 1

As others described about the @ is a variable and := is for assignment, let me try to clarify the what and how things are going.

Query starts with the From clause. The first ( wrapped in parens ) is like a select query with an alias result. This is basically doing an inline query declaration of a "variable" called @date_posted and assigning it a value null. It uses "vars" as just a simple alias name for sql from table/alias requirement. It will only ever return a single row no matter how many @variables you include (comma separated), so you'll never have an issue with a Cartesian Join. That is why a simple comma before the actual table of messages in the second part.

( SELECT @date_posted := NULL ) vars,

                

Now, the rest of the query is using the messages table, ordered by the date_posted.

Now the select TIME_TO_SEC() component. This is computing whatever the time difference is between the @date_posted (currently null on first record read) and compares to date_posted and stores in expected column "period" as you would expect.

The first time through, being NULL, the period for the first record will be a null answer and should not skew the final average as nulls are left out. Now that it has finished computing the period value for that row, it goes to the second column and says take whatever value is in the "date_posted" on the record and assign it to the @date_posted variable. So now you have the seed value started for all subsequent records. The assignment has to be done AFTER the calculation so it acts like a lag (previous record) value when it starts the next record for processing. Now it will have the date to properly compare against.

Only after all records are processed can it then take the final outer AVG() of the period for your results.

Does that help?

Answer 2

User variables used like this are a workaround to the lack of window functions in Mysql 5.7.
In both examples they refer specifically to the lag feature.

Both SQL sentences are similar so I will be analyzing only the first one.

---

Example from first post

Inner FROM clause

( SELECT @date_posted := NULL) vars, messages

This is the inicialization. @date_posted gets an initial NULL value.

Inner SELECT clause

TIME_TO_SEC(TIMEDIFF(@date_posted, date_posted)) AS period, @date_posted := date_posted

@date_posted inside TIMEDIFF holds previous row's value of date_posted and is compared against date_posted from this row to get the difference. Then is assigned current's row date_posted value to carry on to the next row.
In first row @date_posted gets the NULL value from initialization.

Outer SELECT FROM clause

Applies AVG (Average) function to the calculated period column

---

:= operator

Quoting the manual...

:= Assignment operator. Causes the user variable on the left hand side of the operator to take on the value to its right. The value on the right hand side may be a literal value, another variable storing a value, or any legal expression that yields a scalar value, including the result of a query (provided that this value is a scalar value). You can perform multiple assignments in the same SET statement. You can perform multiple assignments in the same statement.
Unlike =, the:= operator is never interpreted as a comparison operator. This means you can use:= in any valid SQL statement (not just in SET statements) to assign a value to a variable.