經(jīng)過(guò)百度,google知道hibernate中hql是不支持union的,所以只能借助native sql了.背景如下:一年前寫(xiě)了一個(gè)hql:

　　原來(lái)代碼

　　String countHql2 = "select count

　　(distinct p) from Project as p,CommentSimple as c,ProjectBookmark as b where ("

　　+ "c.owner.id=? and p.id=c.targetId and c.targetType=500) or (b.user.id=? and p.id=b.project.id)";

　　String hql2 = "select distinct p from Project as p,CommentSimple as c,ProjectBookmark as b where ( "+ "c.owner.id=? and p.id=c.targetId and c.targetType=500) or (b.user.id=? and p.id=b.project.id)"; 主要是找出某個(gè)人所有評(píng)論過(guò)或收藏過(guò)的項(xiàng)目.簡(jiǎn)單表結(jié)構(gòu)如下:

　　project:id owner_id(用戶(hù)id)保存項(xiàng)目的基本信息

　　project_bookmark:uid(用戶(hù)id),project_id(收藏的項(xiàng)目的id),owner_id(收藏者的id)

　　comment_simple:target_type(保存對(duì)某種對(duì)象的評(píng)論,值為500時(shí)表示的對(duì)項(xiàng)目的評(píng)論),target_id(保存對(duì)某種對(duì)象的評(píng)論,值為該對(duì)象的id),project_id(項(xiàng)目的id),owner_id(評(píng)論者的id)

　　由于這個(gè)sql執(zhí)行時(shí)所建的索引無(wú)法使用,而且還造成了三個(gè)表連接會(huì)有大量的無(wú)效的查詢(xún)以及重復(fù)結(jié)果,最后還得要distinct可以想象執(zhí)行的效率.

　　只好改用union來(lái)重寫(xiě),需要用到hibernate的native sql,經(jīng)過(guò)努力終于找到可以用union找出整個(gè)對(duì)象以及在配置文件中與該對(duì)象有關(guān)系的對(duì)象的方法.

　　與其說(shuō)是找出來(lái)的,不如說(shuō)是試出來(lái)的,代碼如下:

　　union

　　String sql1 = "SELECT COUNT(*) FROM(SELECT p.id FROM project p,comment_simple c WHERE p.id=c.target_id AND c.target_type=500 AND c.uid=" + userId + " UNION SELECT pr.id FROM project pr,project_bookmark b WHERE pr.id=b.project_id AND b.uid=" + userId + ") AS temp"; String sql2 = "(SELECT {p.*} FROM project p,comment_simple c WHERE p.id=c.target_id AND c.target_type=500 AND c.uid=" + userId + ")" + "UNION" + "(SELECT {p.*} FROM project p,project_bookmark b WHERE p.id=b.project_id AND b.uid=" + userId + ")LIMIT " + (pageIndex - 1) * maxPerPage + "," + maxPerPage; SQLQuery query = this.getSession().createSQLQuery(sql1); Integercount=Integer.valueOf(((BigInteger)query.uniqueResult()).toString()); SQLQuery query2 = this.getSession().createSQLQuery(sql2); query2.addEntity("p", Project.class); List<Project> list = query2.list();

　　sql1符合條件的項(xiàng)目的總數(shù).sql2求出符合條件項(xiàng)目的某一頁(yè).

　　要注意的是:sql2中{p.*}要寫(xiě)成一樣的.

　　簡(jiǎn)而言之:select {a.*} from A a where ... union select {a.*} from A a where...

　　如果還要排序的話(huà)sql2換成sql3:

　　需要order by時(shí)

　　String sql3 = "(SELECT {p.*},p.created FROM project_hz p,comment_simple c WHERE p.id=c.target_id AND c.target_type=500 AND c.uid=" + userId + ")" + "UNION" + "(SELECT {p.*} ,p.created FROM project_hz p,project_bookmark b WHERE p.id=b.project_id AND b.uid=" + userId + ") ORDER BY created LIMIT " + (pageIndex - 1) * maxPerPage + "," + maxPerPage; 要注意的是p.created(需要排序的那個(gè)字段) 要個(gè)別標(biāo)出,因?yàn)閔ibernate在轉(zhuǎn)換為sql是會(huì)寫(xiě)成 select created as ...所以排序時(shí)將不起作用,需要我們自己標(biāo)出.

　　這里只是找出了一個(gè)解決方案,因?yàn)閷?duì)hibernate的調(diào)優(yōu)不是特別熟悉這樣做會(huì)不會(huì)造成另外不好的影響就不清楚了,如果有高手看到有什么問(wèn)題,或者有更好的方法.請(qǐng)不吝賜教!!