SQL Server 2012 : extract Regex groups

Question

I have text in my database in Markdown format. I'd like to extract links and count the number of matching links I have. I can get a listing of text blocks that contain links using a query similar to this:

SELECT post_text
FROM posts p
WHERE p.body like '%\[%](http%)%' ESCAPE '\'

How do I go to the next step though, and just extract the link portion of the text (the part that is in the parenthesis)? If I can get this, I can count the number of times this specific link is in my dataset.

Some sample data:

"Visit [Google](http://google.com)"    -> Should return "http://google.com"
"Get an [iPhone](http://www.apple.com) (I like it better than Android)"   -> Should return "http://www.apple.com"
"[Example](http://example.com)"    -> Should return "http://example.com"
"This is a message"    -> Nothing to return on this one, no link
"I like cookies (chocolate chip)"  -> Nothing to return on this one, no link
"[Frank] says 'Hello'" -> Nothing to return on this one, no link

I am using SQL Server 2012 (if there are differences between versions in this regard).

Answer 1

Assuming the actual data is no more complex than the stated examples, this should work without resorting to RegEx:

DECLARE @posts TABLE
(
   post_id INT NOT NULL IDENTITY(1, 1),
   post_text NVARCHAR(4000) NOT NULL,
   body NVARCHAR(2048) NULL
);
INSERT INTO @posts (post_text, body) VALUES (N'first',
                                           N'Visit [Google](http://google.com)');
INSERT INTO @posts (post_text, body) VALUES (N'second',
                                           N'Get an [iPhone](http://www.apple.com)');
INSERT INTO @posts (post_text, body) VALUES (N'third',
                                           N'[Example](http://example.com)');
INSERT INTO @posts (post_text, body) VALUES (N'fourth',
                                           N'This is a message');
INSERT INTO @posts (post_text, body) VALUES (N'fifth',
                                           N'I like cookies (chocolate chip)');
INSERT INTO @posts (post_text, body) VALUES (N'sixth',
                                           N'[Frankie] says ''Relax''');
INSERT INTO @posts (post_text, body) VALUES (N'seventh',
                                           NULL);


SELECT p.post_text,
       SUBSTRING(
                  p.body,
                  CHARINDEX(N'](', p.body) + 2,
                  CHARINDEX(N')', p.body) - (CHARINDEX(N'](', p.body) + 2)
                ) AS [URL]
FROM   @posts p
WHERE  p.body like '%\[%](http%)%' ESCAPE '\';

Output:

post_text  URL
first      http://google.com
second     http://www.apple.com
third      http://example.com

PS:
If you really want to use Regular Expressions, they can only be done via SQLCLR. You can write your own or download pre-done libraries. I wrote one such library, SQL#, that has a Free version that includes the RegEx functions. But those should only be used if a T-SQL solution cannot be found, which so far is not the case here.