Primary Key which exists Multiple Times - sql

Question

I want to create a Table "News" as Example. In this Table I have a group_id and a lang_id. The group_id stays as for the ID for the "News" but as group_id because it has a Translation in some Languages. This group_id shall be autoIncrement if I insert a new "News" with different lang_id but the same for every lang_id I attach to the Insert Command.
For Example a Table what I want to Achieve.
I say here new News in German(1),English(2),French(3) it shall Create new Rows for this Langs with the same group_id but itself auto Increment.
Table: News
x-----------------------------x
| group_id| lang_id | news |
x---------|-----------|-------x
| 1 | 1 | Hallo |
| 1 | 2 | Hello |
| 1 | 3 | Holla |
| 2 | 1 | bye |
| 2 | 2 | byebye|
| 2 | 3 | Ciao |
x-----------------------------x
How could I achieve that with an Primary Key, group_id would be the Primary Key??

Answer 1

The primary key must be unique by definition.
This means that in your case, the primary key of the table should be the combination of group_id and lang_id.
As to the question if your group_id can be an identity column, That's possible, but you will have to use SET IDENTITY_INSERT for every new language for an existing content.
example:
CREATE TABLE tblNews
(
group_id int identity(1,1) not null,
lang_id int not null,
news varchar(20),
primary key(group_id, lang_id)
)
INSERT INTO tblNews VALUES(1, 'hello')
SET IDENTITY_INSERT tblNews ON
INSERT INTO tblNews (group_id, lang_id, news) VALUES(1, 2, 'hello')
INSERT INTO tblNews (group_id, lang_id, news) VALUES(1, 3, 'holla')
SET IDENTITY_INSERT tblNews OFF
INSERT INTO tblNews VALUES(1, 'good bye')
SET IDENTITY_INSERT tblNews ON
INSERT INTO tblNews (group_id, lang_id, news) VALUES(2, 2, 'byebye')
INSERT INTO tblNews (group_id, lang_id, news) VALUES(2, 3, 'Ciao')
SET IDENTITY_INSERT tblNews OFF
SELECT *
FROM tblNews
results:
group_id lang_id news
1 1 hello
1 2 hello
1 3 holla
2 1 good bye
2 2 byebye
2 3 Ciao
Though this design is clearly possible, I'm having a hard time recommending it. I would probably use a different approach for a multilingual database - one table for the news, one table for the languages, and one table for the translations:
tblNews
newsId int identity(1,1) primary key,
-- and whatever else content that is not language-dependent
tblLanguages
languageId int identity(1,1),
languageName nvarchar
tblNewsContent
NewsId (pk)
LanguageId (pk)
Contant nvarchar

Answer 2

The three table method is the best option, but I would say it needs something unique that isn't dependent on the translation.
If you are on sql server 2012+ you can use a sequence for this situation, though I can't say it is the best practice.
rextester: http://rextester.com/RRTJ4439
create sequence dbo.NewsIdSequence as int start with 1 increment by 1;
create table Lang (
id int not null
, name nvarchar(64)
, alias nvarchar(64)
, constraint pk_Lang primary key clustered (id)
, constraint uq_Lang_Name unique (name)
);
create table NewsLanguage (
news_id int not null
constraint df_NewsLanguage_news_id default (next value for dbo.NewsIdSequence)
, lang_id int not null
, title nvarchar(256) not null
, article nvarchar(max) not null
, constraint pk_NewsLanguage primary key clustered (news_id, lang_id)
, constraint fk_langLanguage_lang_id foreign key (lang_id) references lang(id)
);
insert into Lang (id, Name, alias)
select top 3 langid, name, alias
from syslanguages
order by langid;
declare #NextNewsId int;
set #NextNewsId = next value for dbo.NewsIdSequence;
insert into NewsLanguage( news_id, lang_id, title, article)
select #NextNewsId, 0, 'Hello', 'Hello ... '
union all select #NextNewsId, 1, 'Hallo', 'Hallo ... '
union all select #NextNewsId, 2, 'Bonjour', 'Bonjour ...';
set #NextNewsId = next value for dbo.NewsIdSequence;
insert into NewsLanguage( news_id, lang_id, title, article) values
(#NextNewsId, 0, 'Goodbye','Goodbye ...' )
, (#NextNewsId, 1, 'Auf Wiedersehen', 'Auf Wiedersehen ...' )
, (#NextNewsId, 2, 'Au Revoir', 'Au Revoir ...');
select *
from dbo.NewsLanguage nl
inner join dbo.Lang l on nl.lang_id = l.id
The three table method would be better, as Zohar Peled explained. Here's a version without hungarian notation:
create table Lang (
id int not null identity (1,1)
, name nvarchar(64)
, alias nvarchar(64)
, constraint pk_Lang primary key clustered (id)
, constraint uq_Lang_Name unique (name)
);
create table News (
id int not null identity (1,1)
, unique_column_of_importance nvarchar(64)
, constraint pk_News primary key clustered (id)
, constraint uq_News_Title unique (unique_column_of_importance)
);
create table NewsLanguage (
news_id int not null
, lang_id int not null
, title nvarchar(256) not null
, article nvarchar(max) not null
, constraint pk_NewsLanguage primary key clustered (news_id, lang_id)
, constraint fk_NewsLanguage_news_id foreign key (news_id) references news(id)
, constraint fk_NewsLanguage_lang_id foreign key (lang_id) references lang(id)
);