集团的毗连已经给出奇怪的结果

Question

我有这样的SQL查询：

SELECT v.*, group_concat(distinct(vi.interest_id)) as interests, group_concat(distinct(vs.skill_id)) as skills, vc.date_from 
FROM `vacancies` as v 
LEFT JOIN `vacancy_interests` as vi on v.vacancy_id = vi.vacancy_id 
LEFT JOIN `vacancy_skills` as vs on v.vacancy_id = vs.vacancy_id 
LEFT JOIN `vacancy_calendar` as vc on v.vacancy_id = vc.vacancy_id 
WHERE v.vacancy_visibility_end_date >= CURDATE() 
GROUP BY v.vacancy_id 

考虑该查询返回的结果如下（最后3列是那些在这个问题讨论）：

vacancy_id,org_id,name,description,number_required,occupancy_kind,website,offer,logo,banner,address_country,address_city,address_postal_code,address_line_1,address_line_2,vacancy_visibility_start_date,vacancy_visibility_end_date,engagement,interests,skills,date_from 

    "2","1","test123","aze<sdgqswdfg","1","1","","blabla",NULL,"12049394_10208129537615226_4853636504350654671_n.jpg","Belgie","Brussel","1000","Brusselsestraat 15",NULL,"2016-09-02 00:00:00","2016-09-19 00:00:00","3","13,6,1","4,3","2016-09-13 00:00:00" 
    "3","1","blablabla","lkpjoip","1","2","","blabla",NULL,NULL,"Belgie","Antwerpen","2000","Antwerpsestraat 16",NULL,"2016-09-02 00:00:00","2016-09-29 00:00:00","3","28","7,8,5","2016-09-01 00:00:00" 
    "4","1","hahaha","14556dsf","1","3","","blabla",NULL,NULL,"Belgie","Mechelen","2800","Mechelsesteenweg 17",NULL,"2016-09-02 00:00:00","2016-09-28 00:00:00","3",NULL,NULL,"2016-09-26 00:00:00" 
    "5","1","omggg","45sdfdj5","1","1","","blabla",NULL,NULL,"Belgie","Gent","3000","Gentsesteenweg 18",NULL,"2016-09-02 00:00:00","2016-09-30 00:00:00","3","17,11","4,1","2016-09-19 00:00:00" 
    "6","1","this is a test","wauhiufdsq","1","2","","blabla",NULL,NULL,"Belgie","Luik","4000","Luikseweg 19",NULL,"2016-09-02 00:00:00","2016-09-30 00:00:00","3","19,17,22","6","2016-08-10 00:00:00" 

请注意，空缺利益和空缺技能表可以包含单个空缺的多个记录。例如。这可能是空缺3有3行与所有不同的interest_id's。 group_concat在这里解决了我的问题。 所以这个查询工作正常，因为它应该。

不过，我遇到两个问题有以下几点：

1）当我通过ID在利益HAVING添加过滤器，这只能返回我而不是预期的两排一排。

SELECT v.*, group_concat(distinct(vi.interest_id)) as interests, group_concat(distinct(vs.skill_id)) as skills, vc.date_from 
FROM `vacancies` as v 
LEFT JOIN `vacancy_interests` as vi on v.vacancy_id = vi.vacancy_id 
LEFT JOIN `vacancy_skills` as vs on v.vacancy_id = vs.vacancy_id 
LEFT JOIN `vacancy_calendar` as vc on v.vacancy_id = vc.vacancy_id 
WHERE v.vacancy_visibility_end_date >= CURDATE() 
GROUP BY v.vacancy_id 
HAVING interests IN (17) 

这只会返回一行。与vacanacy_id 5即记录，同时它也应该显然也返回vacancy_id = 6

的事情，是最奇怪的对我来说，如果我做同样的事情，但对于技能（有技巧（4）） ，这会返回多行并显示正确的结果。

2）当我想在date_from过滤（用在具有兴趣和技能一起，我请执行下列操作：

SELECT v.*, group_concat(distinct(vi.interest_id)) as interests, group_concat(distinct(vs.skill_id)) as skills, vc.date_from 
FROM `vacancies` as v 
LEFT JOIN `vacancy_interests` as vi on v.vacancy_id = vi.vacancy_id 
LEFT JOIN `vacancy_skills` as vs on v.vacancy_id = vs.vacancy_id 
LEFT JOIN `vacancy_calendar` as vc on v.vacancy_id = vc.vacancy_id 
WHERE v.vacancy_visibility_end_date >= CURDATE() AND date(vc.date_from) > '2016-09-10' 
GROUP BY v.vacancy_id 
HAVING skills IN (4) 

这只会回到我的空位数5，而显然也是空房数2的日期比2016年9月10日（2016年9月13日00:00:00）大....

我在做什么错在这里？

Answer 1

你HAVING条款是错误的方法来检查是否存在条件使用级联值，只需使用：

HAVING MAX(vi.interest_id IN (17)) > 0 

当你这样做：

HAVING interests IN (17) 

然后您比较字符串到数字。该字符串被悄悄地转换为数字。在这种情况下，只有第一个元素被转换。所以，如果interests以“17”开始，那么它匹配，否则它不匹配。

此外，请注意，只要没有太多兴趣和技能，您在group_concat()中使用distinct的方法就没有问题。如果每个空缺有100个，那么中间结果将有10,000行 - 并且需要更长的时间来处理。然而，每种方法只有少数几种，方法很好。