是否总是需要加入一对一表？

Question

I have four MySQL tables where the first three share the first tables PK, and the forth has a FK to the third table (see below for schema).

Given the PK of the forth table, I need data from the first and forth table only.

Is there any need to join the second and third table?

For instance, is:

SELECT t1.*,t4.*
FROM t1
INNER JOIN t2 ON t2.t1_idt1=t1.idt1
INNER JOIN t3 ON t3.t2_idt2=t2.idt2
INNER JOIN t4 ON t4.t3_idt3=t3.idt3
WHERE t4.idt4=123;

better or worse than:

SELECT t1.*,t4.*
FROM t1
INNER JOIN t4 ON t4.t3_idt3=t1.idt1
WHERE t4.idt4=123;

Please explain why one is better than the other.

SCHEMA

-- MySQL Script generated by MySQL Workbench
-- 08/29/14 12:34:46
-- Model: New Model    Version: 1.0
SET @OLD_UNIQUE_CHECKS=@@UNIQUE_CHECKS, UNIQUE_CHECKS=0;
SET @OLD_FOREIGN_KEY_CHECKS=@@FOREIGN_KEY_CHECKS, FOREIGN_KEY_CHECKS=0;
SET @OLD_SQL_MODE=@@SQL_MODE, SQL_MODE='TRADITIONAL,ALLOW_INVALID_DATES';

-- -----------------------------------------------------
-- Schema mydb
-- -----------------------------------------------------
CREATE SCHEMA IF NOT EXISTS `mydb` DEFAULT CHARACTER SET utf8 COLLATE utf8_general_ci ;
USE `mydb` ;

-- -----------------------------------------------------
-- Table `mydb`.`t1`
-- -----------------------------------------------------
CREATE TABLE IF NOT EXISTS `mydb`.`t1` (
  `idt1` INT NOT NULL,
  `data` VARCHAR(45) NULL,
  PRIMARY KEY (`idt1`))
ENGINE = InnoDB;


-- -----------------------------------------------------
-- Table `mydb`.`t2`
-- -----------------------------------------------------
CREATE TABLE IF NOT EXISTS `mydb`.`t2` (
  `t1_idt1` INT NOT NULL,
  `data` VARCHAR(45) NULL,
  PRIMARY KEY (`t1_idt1`),
  CONSTRAINT `fk_t2_t1`
    FOREIGN KEY (`t1_idt1`)
    REFERENCES `mydb`.`t1` (`idt1`)
    ON DELETE NO ACTION
    ON UPDATE NO ACTION)
ENGINE = InnoDB;


-- -----------------------------------------------------
-- Table `mydb`.`t3`
-- -----------------------------------------------------
CREATE TABLE IF NOT EXISTS `mydb`.`t3` (
  `t2_t1_idt1` INT NOT NULL,
  `data` VARCHAR(45) NULL,
  PRIMARY KEY (`t2_t1_idt1`),
  CONSTRAINT `fk_t3_t21`
    FOREIGN KEY (`t2_t1_idt1`)
    REFERENCES `mydb`.`t2` (`t1_idt1`)
    ON DELETE NO ACTION
    ON UPDATE NO ACTION)
ENGINE = InnoDB;


-- -----------------------------------------------------
-- Table `mydb`.`t4`
-- -----------------------------------------------------
CREATE TABLE IF NOT EXISTS `mydb`.`t4` (
  `t3_t2_t1_idt1` INT NOT NULL,
  `data` VARCHAR(45) NULL,
  INDEX `fk_t4_t31_idx` (`t3_t2_t1_idt1` ASC),
  CONSTRAINT `fk_t4_t31`
    FOREIGN KEY (`t3_t2_t1_idt1`)
    REFERENCES `mydb`.`t3` (`t2_t1_idt1`)
    ON DELETE NO ACTION
    ON UPDATE NO ACTION)
ENGINE = InnoDB;


SET SQL_MODE=@OLD_SQL_MODE;
SET FOREIGN_KEY_CHECKS=@OLD_FOREIGN_KEY_CHECKS;
SET UNIQUE_CHECKS=@OLD_UNIQUE_CHECKS;

Answer 1

You don't need to join any tables if they're not relevant to the query. A table is relevant if you need to refer to it columns, or you want the result set limited to rows that have matches in that table. Joining with extra tables will slow down the query, because the query planner cannot determine that these joins are redundant; there's no way to tell the DBMS that there's a 1-to-1 relationship between two tables. The best you can do is declare foreign key relationships -- the foreign key is required to be a subset of the values in the other table; but there's no way to declare exact equivalence. In fact, even if you intend to have 1-to-1 equivalence, you don't actually have it -- when you insert into the three tables, you have to do them sequentially, so for a brief period there's an extra row in one or two of the tables.

You can see that the query planner makes use of the extra tables if you compare the result of EXPLAIN between the two queries.

Answer 2

If all IDs are the same in all four tables, and if you don't want to exclude records that don't appear in tables 2 and 3, then there is no reason to include them in the join. It will still use the index on table 1, even without the foreign key relationship.

I would consider renaming all the ID columns to signify the fact that they are the same.

Answer 3

The second solution is much better, the query only get the data from the two required tables instead of applying the jointure between two additional tables that you don't need.

I suggest you to change just the place of your condition:

SELECT t1.*,t4.*
FROM t1
INNER JOIN t4 ON t4.t3_idt3=t1.idt1
                AND t4.idt4=123

By doing so, only the required data are loaded from t4 instead of loading every data and then applying the WHERE condition on the result.

Hope this will help you.