Understanding the execution order of T-SQL query from the execution order of join on and where

Sample code

CREATE TABLE "SCOTT"."A" (

"PERSON_ID" NUMBER(5) NULL ,

"PERSON_NAME" VARCHAR2(255 BYTE) NULL 

);

-- ----------------------------

-- Records of A

-- ----------------------------

INSERT INTO "SCOTT"."A" VALUES ('1', ' Zhang San ');

INSERT INTO "SCOTT"."A" VALUES ('2', ' Li Si ');

INSERT INTO "SCOTT"."A" VALUES ('3', ' Wang Wu ');

INSERT INTO "SCOTT"."A" VALUES ('4', ' Zhao Liu ');

INSERT INTO "SCOTT"."A" VALUES ('5', ' Zhou Qi ');

CREATE TABLE "SCOTT"."B" (

"PERSON_ID" NUMBER(5) NULL ,

"LOVE_FRUIT" VARCHAR2(255 BYTE) NULL 

);

-- ----------------------------

-- Records of B

-- ----------------------------

INSERT INTO "SCOTT"."B" VALUES ('1', ' Banana ');

INSERT INTO "SCOTT"."B" VALUES ('2', ' Apple ');

INSERT INTO "SCOTT"."B" VALUES ('3', ' a mandarin orange ');

INSERT INTO "SCOTT"."B" VALUES ('4', ' pear ');

INSERT INTO "SCOTT"."B" VALUES ('8', ' peach ');

1.join on

SELECT * FROM A LEFT JOIN ORACLE.B ON A.PERSON_ID=B.PERSON_ID AND A.PERSON_ID=1;

2.where

SELECT * FROM A LEFT JOIN ORACLE.B ON A.PERSON_ID=B.PERSON_ID WHERE A.PERSON_ID=1;

Why are the results different ？ It can be explained from the process of query logic processing

---

SQL Execution order

(1)FROM <left_table> <join_type> JOIN <right_table> ON <on_predicate>

(2)WHERE <where_predicate>

(3)GROUP BY <group_by_specification>

(4)HAVING <having_predicate>

(5)SELECT DISTINCT TOP(<top_specification>) <select_list>  

(6)ORDER BY <order_by_list>

• FROM
• ON
• JOIN
• WHERE
• GROUP BY
• WITH CUBE or WITH ROLLUP
• HAVING
• SELECT
• DISTINCT ORDER BY TOP
  ·
  When these steps are carried out , Each step produces a virtual table , The virtual table is used as input for the next step . These virtual tables are for callers （ Client applications or external queries ） Unavailable . Only the table generated in the last step is returned to the caller . If a clause is not specified in the query , The corresponding steps will be skipped

SELECT What has been done at each stage ？

（1）FROM Stage

FROM The stage identifies the source table of the query , And handle table operators . In queries involving join operations （ Various join）, There are mainly the following steps ：

  a. Find the Cartesian product . No matter what kind of join operation it is , The first is to perform cross connect （cross join）, Find the Cartesian product , Generate virtual tables VT1-J1.

  b.ON Filter . This phase is very important for the VT1-J1 Screening , according to ON Clause , Let the predicate take the value true My line passed the test , Insert into VT1-J2.

  c. Add external row . If you specify outer join, Also need to VT1-J2 No matching row found in , Add as an external line to VT1-J2 in , Generate VT1-J3.

 Go through the above steps ,FROM The stage is over . In a nutshell ,FROM The stage is for preprocessing , The tables mentioned in the statement are processed according to the operators provided （ except join, also apply,pivot,unpivot）

（2）WHERE Stage

 WHERE The stage is based on <where_predicate> Medium condition pair VT1 To filter , The line that makes the condition true will be inserted into VT2 in .

（3）GROUP BY Stage

  GROUP The stage is listed according to the specified column name , take VT2 Group the rows in , Generate VT3. Finally, there is only one line per group .

（4）HAVING Stage

 This stage is based on HAVING Predicate pair in clause VT3 And then select the groups , And insert the eligible group into VT4 in .

（5）SELECT Stage

 This stage is the process of projection , Handle SELECT The elements mentioned in clause , produce VT5. This step is usually carried out in the following order 

    a. Calculation SELECT Expressions in the list , Generate VT5-1.

    b. If you have any DISTINCT, Delete VT5-1 Duplicate lines in , Generate VT5-2

    c. If you have any TOP, According to ORDER BY The logical order in which clauses are defined , from VT5-2 Select the number or percentage of rows in the signature , Generate VT5-3

（6）ORDER BY Stage

 according to ORDER BY Clause , Yes VT5-3 The lines in the , Sort , Generate cursor VC6.

---

Example explanation

1.join on Execution process

SELECT * FROM  A LEFT JOIN ORACLE.B ON A.PERSON_ID=B.PERSON_ID AND A.PERSON_ID=1;

Find the Cartesian product , produce 5*5=25 Bar record

ON Filter (A.PERSON_ID=B.PERSON_ID AND A.PERSON_ID=1)

Add external row

2.where Execution process

SELECT * FROM  A LEFT JOIN ORACLE.B ON A.PERSON_ID=B.PERSON_ID WHERE A.PERSON_ID=1;

Find the Cartesian product , produce 5*5=25 Bar record

ON Filter (A.PERSON_ID=B.PERSON_ID )

Add external row

WHERE Stage (A.PERSON_ID=1)

3. The next one sql The results of the implementation of

SELECT * FROM  A LEFT JOIN ORACLE.B ON A.PERSON_ID=B.PERSON_ID