mysql hierarki selvtilslut, hent alle underkategorier

Læs Please Note i bunden først. Okay, du er tilbage.

Oprettelse af en lagret procedure til rekursiv-lignende hierarki-hentning.

Bemærk, du ville ikke have det efter niveauer, men det kan nemt gøres.

Skema:

create table category
(   category_id int not null auto_increment primary key,
    category_name varchar(40) not null,
    parent_id int null,  -- index on this column not a shabby idea
    unique key (category_name)
);

insert category(category_name,parent_id) values ('car',null),('food',null); -- 1,2
insert category(category_name,parent_id) values ('ford',1),('chevy',1),('fruit',2); -- 3,4,5
insert category(category_name,parent_id) values ('economy',3),('escort',6),('exhaust',7); -- 6,7,8
insert category(category_name,parent_id) values ('chassis',7),('loud',8),('banana',5); -- 9,10,11
-- ok granted I could have explicity inserted category_id to make it more obvious

Opret lagret proc:

-- drop procedure showHierarchyBelow;
delimiter $$
create procedure showHierarchyBelow
(
catname varchar(40)
)
BEGIN
    -- deleteMe parameter means i am anywhere in hierarchy of role
    -- and i want me and all my offspring deleted (no orphaning of children or theirs)
    declare bDoneYet boolean default false;
    declare working_on int;
    declare theCount int;
    declare findFirst int;

    select ifnull(category_id,0) into findFirst from category where category_name=catname;

    CREATE TABLE xx_RecursishHelper_xx
    (   -- it's recurshish, not recursive
        category_id int not null,
        processed int not null
    );
    if isnull(findFirst) then
        set findFirst=0;
    end if;
    insert into xx_RecursishHelper_xx (category_id,processed) select findFirst,0;
    if (findFirst=0) then
        set bDoneYet=true;
    else
        set bDoneYet=false;
    end if;

    while (!bDoneYet) do
        -- I am not proud of this next line, but oh well
        select count(*) into theCount from xx_RecursishHelper_xx where processed=0;

        if (theCount=0) then 
            -- found em all
            set bDoneYet=true;
        else
            -- one not processed yet, insert its children for processing
            SELECT category_id INTO working_on FROM xx_RecursishHelper_xx where processed=0 limit 1;
            insert into xx_RecursishHelper_xx (category_id,processed)
            select category_id,0 from category
            where parent_id=working_on;

            -- mark the one we "processed for children" as processed
            update xx_RecursishHelper_xx set processed=1 where category_id=working_on;
        end if;
    end while;

    delete from xx_RecursishHelper_xx where category_id=findFirst;

    select x.category_id,c.category_name
    from xx_RecursishHelper_xx x
    join category c
    on c.category_id=x.category_id;

    drop table xx_RecursishHelper_xx;
END
$$

Test lagret proc:

call showHierarchyBelow('food');
+-------------+---------------+
| category_id | category_name |
+-------------+---------------+
|           5 | fruit         |
|          11 | banana        |
+-------------+---------------+

call showHierarchyBelow('car');
+-------------+---------------+
| category_id | category_name |
+-------------+---------------+
|           3 | ford          |
|           4 | chevy         |
|           6 | economy       |
|           7 | escort        |
|           8 | exhaust       |
|           9 | chassis       |
|          10 | loud          |
+-------------+---------------+

call showHierarchyBelow('ford');
+-------------+---------------+
| category_id | category_name |
+-------------+---------------+
|           6 | economy       |
|           7 | escort        |
|           8 | exhaust       |
|           9 | chassis       |
|          10 | loud          |
+-------------+---------------+

call showHierarchyBelow('xxx');
-- no rows

Bemærk, at jeg kun har ændret dette svar af mine fra et par måneder siden til dine behov.

Bemærk venligst

Ovenstående er kun til illustrative formål. I en situation i den virkelige verden ville jeg aldrig lave tabeller i en lagret proc. DDL-overheaden er betydelig. I stedet ville jeg bruge allerede eksisterende ikke-temp-tabeller med et sessionskoncept. Og rens det ud af rækkerne til den afsluttede session. Så tag ikke ovenstående som mere end en stråmand, der venter på, at du gør det mere performant som sådan. Spørg, om det er forvirrende.