Question

In the example below I have a bibliographic table with authors and their papers. For example, authors ‘001’ and ‘003’ wrote article ‘678’ together.

articleId | authorId
123 | 001
123 | 002
345 | 002
345 | 003
345 | 004
678 | 001
678 | 003

I need to select co-occurrences between authors based on their common authorship. For example, for table above I need to construct the following table:

AuthorA | AuthorB
001 | 002
002 | 003
002 | 004
003 | 004
001 | 003

First table is very large (approx. 1.800.000 rows). When I first try with MS SQL Server 2008, construction of the second table was fast, but I’m stuck with MySQL now. I use the following query:

SELECT foo.authorId AS authorA, bar.authorId AS authorB
  FROM
    (SELECT * FROM tblAuthorHasBib) AS foo,
    (SELECT * FROM tblAuthorHasBib) AS bar
  WHERE
    foo.articleId = bar.articleId
  AND
    foo.authorId <> bar.authorId
  GROUP BY foo.authorId, bar.authorId

How to rewrite my query to be as fast as with MS SQL? Thanks in advance for any pointer.

Answer 1

You could write your query as shown below, which will avoid any using the GROUP BY clause and any in-line views.

SELECT foo.authorId AS authorA, bar.authorId AS authorB
FROM tblAuthorHasBib foo
JOIN tblAuthorHasBib bar
   ON foo.articleId = bar.articleId 
WHERE foo.authorId != bar.authorId

Alternatively, as per @1osmi’s comment, if you wanted only unique permutations of authors then you could replace the != with < as shown below

SELECT foo.authorId AS authorA, bar.authorId AS authorB
FROM tblAuthorHasBib foo
JOIN tblAuthorHasBib bar
   ON foo.articleId = bar.articleId 
WHERE foo.authorId < bar.authorId